diff --git a/README.markdown b/README.markdown new file mode 100644 --- /dev/null +++ b/README.markdown @@ -0,0 +1,23 @@ +# **RegExDot**. + +This is **RegExDot**, a polymorphic regex-engine. + +## Installation + +It can be built and installed using [Cabal](https://www.haskell.org/cabal/users-guide/installing-packages.html). + +## Documentation + +More information about this library can be found at [RegExDot](http://functionalley.eu/RegExDot/regExDot.html). + +## License + +For information on copying and distributing this package, see the file **LICENSE** in this directory. + +## Bug-reporting + +Bug-reports should be emailed to . + +## Author + +This library is written and maintained by Dr. Alistair Ward. diff --git a/changelog b/changelog deleted file mode 100644 --- a/changelog +++ /dev/null @@ -1,47 +0,0 @@ -2010-11-18 Dr. Alistair Ward - -0.9.0.0 - * First version of the package. -0.10.0.0 - * Reacted to major-number changes in the package 'ToolShed-0.10.0.0'. - * Created "src/RegExDot/" sub-directory & then modified module-names accordingly. - * Pacified 'hlint'. - * Added a makefile. - * Removed awkward unary operators from 'Repeatable.hs'. - * Qualified identifiers used in error-messages. -0.10.0.1 - * Ported to ghc-7.0.1: - Used definitions of 'NFData' & 'rnf' from new package 'deepseq', rather than 'parallel'. - Replaced reference to the deprecated function 'Control.Parallel.Strategies.parFlatMap' with @concat . Control.Parallel.Strategies.parMap@. -0.10.1.0 - * Set @threaded@ flag to @True@ by default. - * Replaced use of @threaded@-flag in source-code, with CABAL CPP-macro. - * Relocated "RegExDot.Options" to "ToolShed.Options". -0.10.2.0 - * Renamed package from 'RegExDot' to 'regexdot', for compatibility with Debian's ".deb"-format. -0.10.2.1 - * Added manually controlled "llvm" flag to "regexdot.cabal". - * Changed identifier for type-parameters, to better reflect its role. - * Reacted to the creation of module "ToolShed.Defaultable". - * Uploaded to . -0.10.2.2 - * Amended 'regexdot.cabal' to more correctly specify dependency on package 'toolshed'. - * Used new module "ToolShed.Pair" from package toolshed-0.12.0.0. - * Guarded 'eager-blackholing' flag in cabal file. -0.11.0.0 - * Replaced '(+ 1)' and '(- 1)' with the faster calls 'succ' and 'pred', in "RegExDot.RegEx". - * Reacted to new module-hierarchy, creation of new module "ToolShed.Data.List.Splits" and addition of method 'ToolShed.SelfValidate.getErrors', in toolshed-0.13.0.0, and used it to improved error-reporting in @instance Read RegExDot.RegEx.ExtendedRegEx@. - * Minor reworking of 'RegExDot.Repeatable.repeatableParser'. - * Replaced "System" with "System.Environment" and "System.Exit". -0.11.0.1 - * Added class 'Eq' to the context of 'RegExDot.RegEx.safeReciprocal', for migration to 'ghc-7.4'. -0.11.1.0 - * Removed comments referring to deleted module "ToolShed.Unsafe", from "RegExDot.RegEx". - * Removed 'Show' from the context of functions in "RegExDot.RegEx" - * Exported a new constant 'RegExDot.Anchor.unanchored'. - * Refactored 'RegExDot.ConsumptionProfile.withinConsumptionBounds'. -0.11.1.1 - * Tested with 'haskell-platform-2013.2.0.0'. - * Replaced preprocessor-directives with 'build-depends' constraints in 'regexdot.cabal'. - * In function 'RegExDot.RegEx.findMatch.findMatchSlave.matchPairList', changed 'fromIntegral' (which required a type-signature) to 'toRational'. - * Either replaced instances of '(<$>)' with 'fmap' to avoid ambiguity between "Control.Applicative" & "Prelude" which (from 'base-4.8') also exports this symbol, or hid the symbol when importing the "Prelude".. diff --git a/changelog.markdown b/changelog.markdown new file mode 100644 --- /dev/null +++ b/changelog.markdown @@ -0,0 +1,53 @@ +# 2010-11-18 Dr. Alistair Ward + +## 0.9.0.0 + * First version of the package. +## 0.10.0.0 + * Reacted to major-number changes in the package **ToolShed-0.10.0.0**. + * Created **src/RegExDot/** sub-directory & then modified module-names accordingly. + * Pacified **hlint**. + * Added a makefile. + * Removed awkward unary operators from **Repeatable.hs**. + * Qualified identifiers used in error-messages. +## 0.10.0.1 + * Ported to ghc-7.0.1: + Used definitions of `NFData` & `rnf` from new package **deepseq**, rather than **parallel**. + Replaced reference to the deprecated function `Control.Parallel.Strategies.parFlatMap` with `concat . Control.Parallel.Strategies.parMap`. +## 0.10.1.0 + * Set @threaded@ flag to @True@ by default. + * Replaced use of @threaded@-flag in source-code, with CABAL CPP-macro. + * Relocated **RegExDot.Options** to **ToolShed.Options**. +## 0.10.2.0 + * Renamed package from **RegExDot** to **regexdot**, for compatibility with Debian's *.deb*-format. +## 0.10.2.1 + * Added manually controlled **llvm** flag to the *.cabal*-file. + * Changed identifier for type-parameters, to better reflect its role. + * Reacted to the creation of module **ToolShed.Defaultable**. + * Uploaded to [Hackage](http://hackage.haskell.org/packages/hackage.html). +## 0.10.2.2 + * Amended the *.cabal*-file to more correctly specify dependency on package **toolshed**. + * Used new module **ToolShed.Pair** from package **toolshed-0.12.0.0**. + * Guarded **eager-blackholing** flag in the *.cabal*-file. +## 0.11.0.0 + * Replaced `(+ 1)` and `(- 1)` with the faster calls `succ` and `pred`, in **RegExDot.RegEx**. + * Reacted to new module-hierarchy, creation of new module **ToolShed.Data.List.Splits** and addition of method `ToolShed.SelfValidate.getErrors`, in **toolshed-0.13.0.0**, and used it to improved error-reporting in `instance Read RegExDot.RegEx.ExtendedRegEx`. + * Minor reworking of `RegExDot.Repeatable.repeatableParser`. + * Replaced `System` with `System.Environment` and `System.Exit`. +## 0.11.0.1 + * Added class `Eq` to the context of `RegExDot.RegEx.safeReciprocal`, for migration to **ghc-7.4**. +## 0.11.1.0 + * Removed comments referring to deleted module **ToolShed.Unsafe**, from **RegExDot.RegEx**. + * Removed `Show` from the context of functions in **RegExDot.RegEx** + * Exported a new constant `RegExDot.Anchor.unanchored`. + * Refactored `RegExDot.ConsumptionProfile.withinConsumptionBounds`. +## 0.11.1.1 + * Tested with **haskell-platform-2013.2.0.0**. + * Replaced preprocessor-directives with **build-depends** constraints in the *.cabal*-file. + * In function `RegExDot.RegEx.findMatch.findMatchSlave.matchPairList`, changed `fromIntegral` (which required a type-signature) to `toRational`. + * Either replaced instances of `(<$>)` with `fmap` to avoid ambiguity between **Control.Applicative** & **Prelude** which (from **base-4.8**) also exports this symbol, or hid the symbol when importing the **Prelude**.. +## 0.11.1.2 + * Added **Default-language**-specification to the *.cabal*-file. + * Added file **README.markdown**. + * Converted this file to markdown-format. + * Renamed directory **src/** to **src-lib/** for consistency with other packages. + * Used **CPP** to control the import of symbols from **Control.Applicative**. diff --git a/copyright b/copyright --- a/copyright +++ b/copyright @@ -5,7 +5,7 @@ Copyright (C) 2010-2011 Dr. Alistair Ward. All Rights Reserved. Home-page: - http://functionalley.eu + http://functionalley.eu/RegExDot/regExDot.html License: GNU GENERAL PUBLIC LICENSE Version 3; see '/usr/share/common-licenses/GPL-3' or '/usr/share/doc/licenses/gpl-3.0.txt' where available, or the local packaged file 'LICENSE'. diff --git a/makefile b/makefile deleted file mode 100644 --- a/makefile +++ /dev/null @@ -1,59 +0,0 @@ -# Copyright (C) 2010-2014 Dr. Alistair Ward -# -# This program is free software: you can redistribute it and/or modify -# it under the terms of the GNU General Public License as published by -# the Free Software Foundation, either version 3 of the License, or -# (at your option) any later version. -# -# This program is distributed in the hope that it will be useful, -# but WITHOUT ANY WARRANTY; without even the implied warranty of -# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -# GNU General Public License for more details. -# -# You should have received a copy of the GNU General Public License -# along with this program. If not, see . - -.PHONY: all build check clean configure copy haddock help hlint install prof sdist - -all: install - -install: build haddock - @[ -z "$$CABAL_INSTALL_OPTIONS" ] || echo "INFO: CABAL_INSTALL_OPTIONS='$$CABAL_INSTALL_OPTIONS'" - runhaskell Setup $@ $$CABAL_INSTALL_OPTIONS - -prof: - CABAL_CONFIGURE_OPTIONS="--enable-library-profiling $$CABAL_CONFIGURE_OPTIONS" make install - -copy: build - @[ -z "$$CABAL_COPY_OPTIONS" ] || echo "INFO: CABAL_COPY_OPTIONS='$$CABAL_COPY_OPTIONS'" - runhaskell Setup $@ $$CABAL_COPY_OPTIONS - -build: configure - @[ -z "$$CABAL_BUILD_OPTIONS" ] || echo "INFO: CABAL_BUILD_OPTIONS='$$CABAL_BUILD_OPTIONS'" - runhaskell Setup $@ $$CABAL_BUILD_OPTIONS - -configure: regexdot.cabal Setup.hs - @[ -z "$$CABAL_CONFIGURE_OPTIONS" ] || echo "INFO: CABAL_CONFIGURE_OPTIONS='$$CABAL_CONFIGURE_OPTIONS'" - runhaskell Setup $@ $$CABAL_CONFIGURE_OPTIONS #--user - -haddock: configure - PATH=~/.cabal/bin:$$PATH runhaskell Setup $@ --hyperlink-source #Amend path to find 'HsColour', as required for 'hyperlink-source'. - -hlint: - @$@ -i 'Use &&' src/ -i 'Use ||' +RTS -N - -sdist: - TAR_OPTIONS='--format=ustar' runhaskell Setup $@ - -check: sdist - cabal upload --check --verbose=3 dist/*.tar.gz; - -clean: - runhaskell Setup $@ - find src -type f \( -name '*.hc' -o -name '*.hcr' -o -name '*.hi' -o -name '*.o' \) -delete - -help: - @grep '^[a-zA-Z].*:' makefile | sed -e 's/:.*//' - -rpm: regexdot.rpm build - diff --git a/regexdot.cabal b/regexdot.cabal --- a/regexdot.cabal +++ b/regexdot.cabal @@ -1,30 +1,51 @@ --- Package-properties -Name: regexdot -Version: 0.11.1.1 -Cabal-Version: >= 1.6 -Copyright: (C) 2010-2015 Dr. Alistair Ward -License: GPL -License-file: LICENSE -Author: Dr. Alistair Ward -Stability: Unstable interface, incomplete features. -Synopsis: A polymorphic, POSIX, extended regex-engine. -Build-Type: Simple -Description: Provides a portable, POSIX, extended regex-engine, designed to process a list of /arbitrary/ objects. -Category: Search, Regex -Tested-With: GHC == 6.10, GHC == 6.12, GHC == 7.0, GHC == 7.4, GHC == 7.6, GHC == 7.10 -Homepage: http://functionalley.eu -Maintainer: regexdot functionalley eu -Bug-reports: regexdot functionalley eu -Extra-Source-Files: changelog, copyright, makefile +-- This file is part of RegExDot. +-- +-- RegExDot is free software: you can redistribute it and/or modify +-- it under the terms of the GNU General Public License as published by +-- the Free Software Foundation, either version 3 of the License, or +-- (at your option) any later version. +-- +-- RegExDot is distributed in the hope that it will be useful, +-- but WITHOUT ANY WARRANTY; without even the implied warranty of +-- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +-- GNU General Public License for more details. +-- +-- You should have received a copy of the GNU General Public License +-- along with RegExDot. If not, see . --- Turn on using: 'runhaskell ./Setup.hs configure -f llvm'. +Name: regexdot +Version: 0.11.1.2 +Cabal-version: >= 1.10 +Copyright: (C) 2010-2015 Dr. Alistair Ward +License: GPL +License-file: LICENSE +Author: Dr. Alistair Ward +Stability: stable +Synopsis: A polymorphic, POSIX, extended regex-engine. +Build-type: Simple +Description: Provides a portable, POSIX, extended regex-engine, designed to process a list of /arbitrary/ objects. +Category: Search, Regex +Tested-with: GHC == 6.10, GHC == 6.12, GHC == 7.0, GHC == 7.4, GHC == 7.6, GHC == 7.8, GHC == 7.10 +Homepage: http://functionalley.eu/RegExDot/regExDot.html +Maintainer: mailto regexdot functionalley eu +Bug-reports: mailto regexdot functionalley eu + +-- None of these files are needed at run-time. +Extra-source-files: + changelog.markdown + copyright + README.markdown + +-- Enable using: 'cabal configure -f llvm'. flag llvm Description: Whether the 'llvm' compiler-backend has been installed and is required for code-generation. - manual: True - default: False + Manual: True + Default: False Library - hs-source-dirs: src + Default-language: Haskell2010 + GHC-options: -Wall -O2 -fno-warn-tabs + Hs-source-dirs: src-lib Exposed-modules: RegExDot.Anchor @@ -54,8 +75,6 @@ parallel >= 3.0, parsec == 3.*, toolshed >= 0.13 - - GHC-options: -Wall -O2 -fno-warn-tabs if impl(ghc >= 7.4.1) GHC-prof-options: -prof -fprof-auto -fprof-cafs diff --git a/src-lib/RegExDot/Anchor.hs b/src-lib/RegExDot/Anchor.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/Anchor.hs @@ -0,0 +1,90 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * A type of /zero-width assertion/: + the 'Bow'-predicate is 'True' if no input-data has yet been consumed; the 'Stern'-predicate is 'True' if no input-data remains to be consumed. + + * Whilst at the top-level of the regex, the /absence/ of an anchor is equivalent to a non-capturing @.*?@, + tempting one to implement them using this property, + it doesn't apply to those anchors encountered at either end of a sub-expression. + So in all cases, an anchor constrains the permissible consumption of input-data, + but at the top level it also prevents the regex drifting away from either end of the input-data. +-} + +module RegExDot.Anchor( +-- * Types +-- ** Data-types + Anchor(..), +-- ** Type-synonyms + AnchorSpecification, +-- * Constants + bowToken, + sternToken, + tokens, + unanchored +) where + +import qualified Control.DeepSeq + +-- | Defines the types on /anchor/ by which a /regex/ can be moored to a part of the input-data. +data Anchor = + Bow -- ^ Matches only if no input data has yet been consumed. Can only exist at the start of the entire regex, or (in theory) the start of any /alternative/. + | Stern -- ^ Matches only if no input data remains to be consumed. Can only exist at the end of the entire regex, or (in theory) the end of any /alternative/. + deriving ( + Eq +-- Read, -- See specialisation below. +-- Show -- See specialisation below. + ) + +instance Show Anchor where + showsPrec _ Bow = showChar bowToken + showsPrec _ Stern = showChar sternToken + +instance Read Anchor where + readsPrec _ [] = [] -- No parse. + readsPrec _ (' ' : s) = reads s -- Consume white-space. + readsPrec _ ('\t' : s) = reads s -- Consume white-space. + readsPrec _ (c : s) = case c `lookup` [(bowToken, Bow), (sternToken, Stern)] of + Just anchor -> [(anchor, s)] + _ -> [] -- No parse. + +instance Control.DeepSeq.NFData Anchor where + rnf _ = () + +-- | The conventional token used to denote a 'Bow'-anchor, when in 'String'-form. +bowToken :: Char +bowToken = '^' + +-- | The conventional token used to denote a 'Stern'-anchor, when in 'String'-form. +sternToken :: Char +sternToken = '$' + +-- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. +tokens :: String +tokens = [bowToken, sternToken] + +-- | A specification for the anchors required around the expression +type AnchorSpecification = (Maybe Anchor, Maybe Anchor) + +-- | The specification for an unanchored expression. +unanchored :: AnchorSpecification +unanchored = (Nothing, Nothing) + diff --git a/src-lib/RegExDot/BracketExpression.hs b/src-lib/RegExDot/BracketExpression.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/BracketExpression.hs @@ -0,0 +1,63 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + This concept was traditionally known as a /Character-class/, + but to distinguish it from the more recently introduced & narrower /POSIX Character-class/, it has been renamed /Bracket-expression/. +-} + +module RegExDot.BracketExpression( +-- * Types +-- ** Type-synonyms + BracketExpression, +-- * Constants + delimiterTokens, + negationToken, + tokens, +-- * Functions +-- ** Predicates + containsMatch +) where + +import RegExDot.BracketExpressionMember((=~)) +import qualified RegExDot.BracketExpressionMember as BracketExpressionMember + +-- | A /Bracket-expression/ consists of a list of 'BracketExpressionMember.Member's. +type BracketExpression m = [BracketExpressionMember.Member m] + +-- | True if the specified datum matches any 'BracketExpressionMember.Member' of the 'BracketExpression'. +containsMatch :: Eq m + => m -- ^ The input datum. + -> BracketExpression m -- ^ The list of bracket-expression members within which to find a match. + -> Bool +containsMatch datum = any (datum =~) + +-- | The delimiters of a /bracket-expression/, when in 'String'-form. +delimiterTokens :: (Char, Char) +delimiterTokens = ('[', ']') + +-- | Used to denote a negated /bracket-expression/, when in 'String'-form. +negationToken :: Char +negationToken = '^' + +-- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. +tokens :: String +tokens = [fst delimiterTokens, snd delimiterTokens, negationToken] + diff --git a/src-lib/RegExDot/BracketExpressionMember.hs b/src-lib/RegExDot/BracketExpressionMember.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/BracketExpressionMember.hs @@ -0,0 +1,93 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + Permits /Perl-style shortcut/s to be canned & assigned a single-'Char' mnemonic for subsequent reference; + the implementation of 'Read' looks for a back-slashed 'Char', for which it expects there to be a corresponding canned 'ShowablePredicate.ShowablePredicate'. + + [@CAVEATS@] + + Since the underlying polymorphic data-type isn't required to implement neither 'Enum' nor 'Ord', the implementation of 'Read' can't cope with range-specifications. + Lacking this, Bracket-expression members must be enumerated exhaustively. +-} + +module RegExDot.BracketExpressionMember( +-- * Type-classes + ShortcutExpander(..), +-- * Types +-- ** Data-types + Member(..), +-- * Functions +-- ** Operators + (=~) +) where + +import qualified Control.Arrow +import qualified Control.DeepSeq +import qualified RegExDot.ShowablePredicate as ShowablePredicate + +infix 4 =~ -- Same as (==). + +{- | + * The interface via which /Perl-style shortcut/s are expanded (when they occur within a /bracket-expression/), in a manner appropriate to the chosen type-parameter. + + * The expansion of /Perl-style shortcut/s, is more restricted inside than outside, a /bracket-expression/, + & consequently are merely represented here by a 'ShowablePredicate.ShowablePredicate', rather than providing a more general form suitable also for those /Perl-style shortcuts/ found outside /bracket-expression/s. + + * This interface is implemented elsewhere, where the specific type-parameter & consequently the appropriate set of /Perl-style shortcut/s, are defined. +-} +class ShortcutExpander a where + findPredicate :: Char -> Maybe (ShowablePredicate.ShowablePredicate a) -- ^ Attempt to find the appropriate 'ShowablePredicate.ShowablePredicate' to implement this /Perl-style shortcut/. + +{- | + * A /BracketExpression/ can contain either a literal, a range of literals given @(Enum a, Ord a)@, a /Perl-style shortcut/, or when 'Char' is the type-parameter, a /POSIX Character-class/. + + * This data-type reduces the representation of all these possibilities to either a predicate or a literal. +-} +data Member m = + Predicate (ShowablePredicate.ShowablePredicate m) -- ^ This 'Member' is described using a /predicate/, which is run to determine whether the datum conforms & is a member of the "BracketExpression". + | Literal m -- ^ This 'Member' is defined literally, using an item of the polymorphic type. + deriving Eq + +instance Show m => Show (Member m) where + showsPrec _ (Predicate showablePredicate) = shows showablePredicate + showsPrec _ (Literal literal) = shows literal + +instance (ShortcutExpander m, Read m) => Read (Member m) where + readsPrec _ [] = [] -- No parse. + readsPrec _ (' ' : s) = reads s -- Consume white-space. + readsPrec _ ('\t' : s) = reads s -- Consume white-space. + readsPrec _ ('\\' : shortcut : s) = case findPredicate shortcut of + Just showablePredicate -> [(Predicate showablePredicate, s)] + _ -> error $ "readsPrec RegExDot.BracketExpressionMember.Member:\tfindPredicate failed for shortcut " ++ show shortcut + readsPrec _ literal = Control.Arrow.first Literal `map` reads literal + +instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (Member m) where + rnf (Predicate showablePredicate) = Control.DeepSeq.rnf showablePredicate + rnf (Literal literal) = Control.DeepSeq.rnf literal + +-- | Match-operator. +(=~) :: Eq m + => m -- ^ The input datum. + -> Member m -- ^ The member of the bracket-expression against which the input-datum is to be matched. + -> Bool -- ^ The result of the match-operation. +datum =~ Predicate showablePredicate = ShowablePredicate.predicate showablePredicate datum +datum =~ Literal literal = datum == literal + diff --git a/src-lib/RegExDot/CompilationOptions.hs b/src-lib/RegExDot/CompilationOptions.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/CompilationOptions.hs @@ -0,0 +1,48 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * Defines those options relating to compile-time. + + * These tend to be options which affect the result, rather than the means by which it is achieved. + + * Whilst similar structures are present in other regex-implementations, there's no standardisation of the fields. +-} + +module RegExDot.CompilationOptions( +-- * Types +-- ** Data-types + CompilationOptions(..) +) where + +import qualified ToolShed.Defaultable +import qualified ToolShed.Options + +-- | The switch(es) used to control compilation of the /regex/-engine. +data CompilationOptions = MkCompilationOptions { + complyStrictlyWithPosix :: Bool -- ^ Define the offset of captured data, corresponding to a sub-expression which matched zero times, as the artificial value @-1@ specified by POSIX. +} deriving (Eq, Show) + +instance ToolShed.Defaultable.Defaultable CompilationOptions where + defaultValue = ToolShed.Options.blankValue { complyStrictlyWithPosix = True } + +instance ToolShed.Options.Options CompilationOptions where + blankValue = MkCompilationOptions { complyStrictlyWithPosix = undefined } + diff --git a/src-lib/RegExDot/Consumer.hs b/src-lib/RegExDot/Consumer.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/Consumer.hs @@ -0,0 +1,105 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * Define the requirements & capabilities of a data-type which is designed to consume data. + + * This module also facilitates investigation of group-behaviour, by defining operations on either concatenations or alternations of 'Consumer's. + + * This module is specifically for use in a /regex/-engine, + in which the meta-data from which the /regex/ is composed, are considered to /consume/ input data, + whilst concatenation & alternation of such elements, behave as a different types of groups of consumer. +-} + +module RegExDot.Consumer( +-- * Type-classes + Consumer(..), +-- * Types +-- ** Type-synonyms +-- StarHeight, +-- * Functions + accumulateConsumptionProfiles, +-- accumulateConsumptionProfilesFrom, + aggregateConsumptionProfilesFromAlternatives, + aggregateConsumptionProfilesFromConcatenation, +-- ** Query + getConsumptionBounds, + getFewest, + getHasSpecificRequirement +) where + +import RegExDot.ConsumptionProfile((<>), (|+|)) +import qualified Data.List +import qualified RegExDot.ConsumptionBounds as ConsumptionBounds +import qualified RegExDot.ConsumptionProfile as ConsumptionProfile + +{- | + * A measure of the complexity of a /regex/, which has some baring on either the time-complexity or the space-complexity of the solution. + + * +-} +type StarHeight = Int + +{- | + * The interface, to which types, which have a data-requirement, may conform. + + * Regrettably, methods requiring reference to the type of the consumable, would require multi-parameter type-classes. + + * 'consumptionProfile' defines the quantity-range of data which may be consumed, & any specific data-requirements. + + * 'starHeight' measures the complexity of the 'Consumer'. +-} +class Consumer c where + consumptionProfile :: c -> ConsumptionProfile.ConsumptionProfile + starHeight :: c -> StarHeight + +instance Consumer c => Consumer [c] where + consumptionProfile = undefined + starHeight = Data.List.foldl' (\l -> max l . starHeight) 0 + +-- | Convenience-function, to query the 'ConsumptionBounds.ConsumptionBounds' of a 'Consumer'. +getConsumptionBounds :: Consumer c => c -> ConsumptionBounds.ConsumptionBounds +getConsumptionBounds = ConsumptionProfile.consumptionBounds . consumptionProfile + +-- | Determine the minimum acceptable quantity of data. +getFewest :: Consumer c => c -> ConsumptionBounds.DataLength +getFewest = fst . getConsumptionBounds + +-- | Determine whether the specified 'Consumer', has a specific requirement. +getHasSpecificRequirement :: Consumer c => c -> Bool +getHasSpecificRequirement = ConsumptionProfile.hasSpecificRequirement . consumptionProfile + +-- | The aggregate of the specified alternation of 'ConsumptionProfile.ConsumptionProfile's. +aggregateConsumptionProfilesFromAlternatives :: Consumer c => [c] -> ConsumptionProfile.ConsumptionProfile +aggregateConsumptionProfilesFromAlternatives [] = error "RegExDot.Consumer.aggregateConsumptionProfilesFromAlternatives:\tnull list" +aggregateConsumptionProfilesFromAlternatives l = Data.List.foldl1' (<>) $ consumptionProfile `map` l + +-- | Get the 'ConsumptionProfile.ConsumptionProfile's for the specified list of 'Consumer's, then find the net effect of concatenating them. +aggregateConsumptionProfilesFromConcatenation :: Consumer c => [c] -> ConsumptionProfile.ConsumptionProfile +aggregateConsumptionProfilesFromConcatenation = Data.List.foldl' (\acc -> (acc |+|) . consumptionProfile) ConsumptionProfile.zero {-initial value-} + +-- | Get the 'ConsumptionProfile.ConsumptionProfile' for the specified list of 'Consumer's, then accumulate them, from the specified initial value. +accumulateConsumptionProfilesFrom :: Consumer c => ConsumptionProfile.ConsumptionProfile -> [c] -> ConsumptionProfile.AccumulatedConsumptionProfiles +accumulateConsumptionProfilesFrom = scanr ((|+|) . consumptionProfile) + +-- | Get the 'ConsumptionProfile.ConsumptionProfile' for the specified list of 'Consumer's, then accumulate them. +accumulateConsumptionProfiles :: Consumer c => [c] -> ConsumptionProfile.AccumulatedConsumptionProfiles +accumulateConsumptionProfiles = {-init .-} accumulateConsumptionProfilesFrom ConsumptionProfile.zero -- It's useful to leave the initial value at the end of the list. + diff --git a/src-lib/RegExDot/ConsumptionBounds.hs b/src-lib/RegExDot/ConsumptionBounds.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/ConsumptionBounds.hs @@ -0,0 +1,48 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] Describes the bounds of data-consumption. +-} + +module RegExDot.ConsumptionBounds( +-- * Types +-- ** Type-synonyms + ConsumptionBounds, + DataLength, +-- * Constants + zero, +-- * Functions +-- ** Predicates + isPrecise +) where + +-- | A measure of the quantity of an unspecified type of data. +type DataLength = Int + +-- | The minimum & maximum bounds of potential data-consumption. +type ConsumptionBounds = (DataLength, Maybe DataLength) + +-- | Exactly zero consumption. +zero :: ConsumptionBounds +zero = (0, Just 0) + +-- | Predicate which is 'True' if only a precise quantity of data is consumable; no more, no less. +isPrecise :: ConsumptionBounds -> Bool +isPrecise (fewest, most) = Just fewest == most + diff --git a/src-lib/RegExDot/ConsumptionProfile.hs b/src-lib/RegExDot/ConsumptionProfile.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/ConsumptionProfile.hs @@ -0,0 +1,151 @@ +{-# LANGUAGE CPP #-} +{- + Copyright (C) 2010-2015 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + A 'RegExDot.ConsumptionProfile' is composed from both a capacity to consume, & an ability to discriminate based on the specific data. + The former is described by 'consumptionBounds' & the latter by both 'hasSpecificRequirement' & 'canConsumeAnything'. +-} + +module RegExDot.ConsumptionProfile( +-- * Types +-- ** Type-synonyms + AccumulatedConsumptionProfiles, +-- ** Data-types + ConsumptionProfile(..), +-- * Constants + zero, +-- * Functions + accumulateFrom, + aggregateFromConcatenation, +-- ** Operators + (|+|), + (<>), +-- ** Predicates +-- isPrecise, + withinConsumptionBounds +) where + +import Control.Arrow((***)) +import qualified Data.List +import qualified Data.Maybe +import qualified RegExDot.ConsumptionBounds as ConsumptionBounds +import qualified ToolShed.SelfValidate + +#if !MIN_VERSION_base(4,8,0) +import Control.Applicative((<$>), (<*>)) +#endif + +infixr 5 |+| -- Same as (++). +infixr 2 <> -- Same as (||). + +{- | + * A 'Consumer' is considered to have a 'ConsumptionProfile' composed from both a capacity to consume, & an ability to discriminate. + + * Whilst 'hasSpecificRequirement' & 'canConsumeAnything' look like opposites, they can assume independent values; the instance can be both, but only be neither when empty. +-} +data ConsumptionProfile = MkConsumptionProfile { + consumptionBounds :: ConsumptionBounds.ConsumptionBounds, -- ^ The permissible quantity of data, which can be consumed. + hasSpecificRequirement :: Bool, -- ^ Whether at least one specific input datum is required. + canConsumeAnything :: Bool -- ^ Whether at least one arbitrary input datum can be consumed. +} deriving (Eq, Read, Show) + +instance ToolShed.SelfValidate.SelfValidator ConsumptionProfile where + getErrors c@( + MkConsumptionProfile { + consumptionBounds = (fewest, most), + hasSpecificRequirement = hasSpecificRequirement', + canConsumeAnything = canConsumeAnything' + } + ) = ToolShed.SelfValidate.extractErrors [ + (fewest < 0, "Negative fewest=" ++ show fewest ++ "."), + ( + case most of + Just m -> m < fewest + _ -> False, + "Invalid 'most'; " ++ show c ++ "." + ), + (not $ or [c == zero, hasSpecificRequirement', canConsumeAnything'], "Invalid " ++ show c ++ ".") + ] + +{- +-- | True if there's no choice in the quantity of data to consume. +isPrecise :: ConsumptionProfile -> Bool +isPrecise = ConsumptionBounds.isPrecise . consumptionBounds +-} + +-- | Predicate, which is 'True' if the specified data-length, falls within the specified 'ConsumptionBounds.ConsumptionBounds'. +withinConsumptionBounds + :: ConsumptionBounds.DataLength -- ^ The actual quantity of data consumed. + -> ConsumptionProfile -- ^ The bounds within which data-consumption is required to fall. + -> Bool +withinConsumptionBounds dataLength = uncurry (&&) . ((dataLength >=) *** Data.Maybe.maybe True (dataLength <=)) . consumptionBounds + +-- | Both minimum & maximum set to zero; which can be used as the initial value when accumulating the sum of a list. +zero :: ConsumptionProfile +zero = MkConsumptionProfile { + consumptionBounds = ConsumptionBounds.zero, + hasSpecificRequirement = False, + canConsumeAnything = False +} + +-- | The net effect of two concatenated 'ConsumptionProfile's. +(|+|) :: ConsumptionProfile -> ConsumptionProfile -> ConsumptionProfile +MkConsumptionProfile { + consumptionBounds = (lf, ls), + hasSpecificRequirement = lh, + canConsumeAnything = lc +} |+| MkConsumptionProfile { + consumptionBounds = (rf, rs), + hasSpecificRequirement = rh, + canConsumeAnything = rc +} = MkConsumptionProfile { + consumptionBounds = (lf + rf, (+) <$> ls <*> rs), -- The sum of those of the concatenation. + hasSpecificRequirement = lh || rh, -- The concatenation mandates consumption of at least one specific input datum, if either 'ConsumptionProfile' does. + canConsumeAnything = lc || rc -- The concatenation can consume at least one arbitrary input datum, if either 'ConsumptionProfile' can. +} + +-- | The net effect of two alternative 'ConsumptionProfile's. +(<>) :: ConsumptionProfile -> ConsumptionProfile -> ConsumptionProfile +MkConsumptionProfile { + consumptionBounds = (lf, ls), + hasSpecificRequirement = lh, + canConsumeAnything = lc +} <> MkConsumptionProfile { + consumptionBounds = (rf, rs), + hasSpecificRequirement = rh, + canConsumeAnything = rc +} = MkConsumptionProfile { + consumptionBounds = (lf `min` rf, max <$> ls <*> rs), -- Stretched to envelope alternatives. + hasSpecificRequirement = lh && rh, -- The alternation mandates consumption of at least one specific input datum, if both 'ConsumptionProfile's do. + canConsumeAnything = lc || rc -- The alternation can consume at least one arbitrary input datum, if either 'ConsumptionProfile' can. +} + +-- | The aggregate of the specified concatenation of 'ConsumptionProfile's. +aggregateFromConcatenation :: [ConsumptionProfile] -> ConsumptionProfile +aggregateFromConcatenation = Data.List.foldl' (|+|) zero {-initial value-} + +-- | The 'ConsumptionProfile's of all the aggregated tails of a list of 'Consumer's. +type AccumulatedConsumptionProfiles = [ConsumptionProfile] + +-- | Accumulate the specified list of 'ConsumptionProfile's, from the specified initial value. +accumulateFrom :: ConsumptionProfile -> [ConsumptionProfile] -> AccumulatedConsumptionProfiles +accumulateFrom = scanr (|+|) + diff --git a/src-lib/RegExDot/DSL.hs b/src-lib/RegExDot/DSL.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/DSL.hs @@ -0,0 +1,117 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * Defines /binary/ operators, to form a /Domain-specific Language/, by which to compose 'RegEx.Concatenations' from 'RegEx.Pattern's. + + * Each operator quantifies the specified 'RegEx.Pattern', & prepends it to the specified 'RegEx.Concatenation'. + + [@TODO@] Could @[]@ & 'Meta' be instances of a type-class which has methods (-:) etc., + allowing the former to prepend 'RegEx.captureGroup' & the latter 'RegEx.Require' ? + +-} + +module RegExDot.DSL( +-- * Functions +-- ** Operators + (-:), + (?:), + (??:), + (*:), + (*?:), + (+:), + (+?:), + ( #->#:), + ( #->#?:), + ( #->:), + ( #->?:), + ( #:), + (<~>) +) where + +import RegExDot.Repeatable((^#->#), (^#->#?), (^#->), (^#->?), (^#)) +import qualified RegExDot.Anchor as Anchor +import qualified RegExDot.RegEx as RegEx +import qualified RegExDot.Repeatable as Repeatable + +infixr 5 -:, ?:, ??:, *:, *?:, +:, +?:,#->#:, #->#?:, #->:, #->?:, #:, <~> -- Same as for ':', & lower than Repeatable's operators. + +-- | Prepend an unrepeated 'RegEx.Pattern', to the specified 'RegEx.Concatenation'. +(-:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a +(-:) pattern = (Repeatable.one pattern :) + +-- | Prepend an optional 'RegEx.Pattern', to the specified 'RegEx.Concatenation'. +(?:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a +(?:) pattern = (Repeatable.zeroOrOne pattern :) + +-- | A /non-greedy/ version of '?:'. +(??:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a +(??:) pattern = (Repeatable.zeroOrOne' pattern :) + +-- | Prepend a 'RegEx.Pattern', repeatable zero or more times, to the specified 'RegEx.Concatenation'. +(*:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a +(*:) pattern = (Repeatable.zeroOrMore pattern :) + +-- | A /non-greedy/ version of '*:'. +(*?:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a +(*?:) pattern = (Repeatable.zeroOrMore' pattern :) + +-- | Prepend a 'RegEx.Pattern', repeatable one or more times, to the specified 'RegEx.Concatenation'. +(+:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a +(+:) pattern = (Repeatable.oneOrMore pattern :) + +-- | A /non-greedy/ version of '+:'. +(+?:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a +(+?:) pattern = (Repeatable.oneOrMore' pattern :) + +-- | Prepend a 'RegEx.Pattern', repeated a range of times, to the specified 'RegEx.Concatenation'. +( #->#:) :: (RegEx.Pattern a, Repeatable.RepetitionBounds) -> RegEx.Concatenation a -> RegEx.Concatenation a +( #->#:) (pattern, bounds) = (pattern ^#-># bounds :) + +-- | A /non-greedy/ version of '#->#:'. +( #->#?:) :: (RegEx.Pattern a, Repeatable.RepetitionBounds) -> RegEx.Concatenation a -> RegEx.Concatenation a +( #->#?:) (pattern, bounds) = (pattern ^#->#? bounds :) + +-- | Prepend a 'RegEx.Pattern', repeated at least a specified number of times, to the specified 'RegEx.Concatenation'. +( #->:) :: (RegEx.Pattern a, Repeatable.Repetitions) -> RegEx.Concatenation a -> RegEx.Concatenation a +( #->:) (pattern, fewest) = (pattern ^#-> fewest :) + +-- | A /non-greedy/ version of '#->:'. +( #->?:) :: (RegEx.Pattern a, Repeatable.Repetitions) -> RegEx.Concatenation a -> RegEx.Concatenation a +( #->?:) (pattern, fewest) = (pattern ^#->? fewest :) + +-- | Prepend a 'RegEx.Pattern', repeated a precise number of times, to the specified 'RegEx.Concatenation'. +( #:) :: (RegEx.Pattern a, Repeatable.Repetitions) -> RegEx.Concatenation a -> RegEx.Concatenation a +( #:) (pattern, r) = (pattern ^# r :) + +{- | + * Sandwiches a 'RegEx.Concatenation' between optional 'Anchor.Anchor's to construct a 'RegEx.ExtendedRegEx'. + + * Accounts for the unusual requirement to place an 'Anchor.Stern' at the start of the 'RegEx.Concatenation', or an 'Anchor.Bow' at the end. +-} +(<~>) + :: Anchor.AnchorSpecification -- ^ A specification for the anchors required around the expression. + -> RegEx.Concatenation a -- ^ The bare list of repeatable patterns to anchor. + -> RegEx.ExtendedRegEx a -- ^ The completed regex. +(bow, stern) <~> concatenation = RegEx.MkExtendedRegEx { + RegEx.bowAnchor = bow, + RegEx.concatenation = concatenation, + RegEx.sternAnchor = stern +} diff --git a/src-lib/RegExDot/DataSpan.hs b/src-lib/RegExDot/DataSpan.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/DataSpan.hs @@ -0,0 +1,61 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] Augments a 'Span.Span', with the specific data concerned. +-} + +module RegExDot.DataSpan( +-- * Types +-- ** Type-synonyms + DataSpan, +-- * Functions + after, + empty, + join +) where + +import Control.Arrow((***)) +import qualified RegExDot.ConsumptionBounds as ConsumptionBounds +import qualified RegExDot.Span as Span + +{- | + * Augment 'Span.Span' with a copy of the data to which it refers. + + * Similar to . +-} +type DataSpan a = ([a], Span.Span) + +-- | Constructs an empty instance, located at the specified offset. +empty :: ConsumptionBounds.DataLength -> DataSpan a +empty offset = ([], Span.empty offset) + +-- | Returns the furthest extent, i.e. the offset just after the end. +after :: DataSpan a -> ConsumptionBounds.DataLength +after = Span.after . snd + +{- | + * Condenses a list into a single value. + + * Expects a gap-free, sorted list. +-} +join + :: ConsumptionBounds.DataLength -- ^ The offset at which the concatenated match is considered to have occurred. + -> [DataSpan a] -- ^ The list to be concatenated. + -> DataSpan a +join offset = (concat *** Span.join offset) . unzip diff --git a/src-lib/RegExDot/DataSpanTree.hs b/src-lib/RegExDot/DataSpanTree.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/DataSpanTree.hs @@ -0,0 +1,108 @@ +{- + Copyright (C) 2010-2015 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] Permits transformation of 'RegEx.MatchList', to facilitate standardisation. +-} + +module RegExDot.DataSpanTree( +-- * Types +-- ** Type-synonyms +-- DataSpanTree, +-- DataSpanTreeList, +-- * Functions + extractCaptureGroups, + flattenTreeList, + toTreeList +) where + +import qualified Data.Foldable +import qualified RegExDot.ConsumptionBounds as ConsumptionBounds +import qualified RegExDot.DataSpan as DataSpan +import qualified RegExDot.RegEx as RegEx +import qualified RegExDot.Tree as Tree + +-- | Defines a similar 'Tree.Tree' to 'RegEx.Match', but with different 'Tree.Leaf's. +type DataSpanTree a = Tree.Tree (DataSpan.DataSpan a) + +-- | Defines a similar structure to 'RegEx.MatchList'. +type DataSpanTreeList a = [DataSpanTree a] + +-- | Converts a 'RegEx.MatchList' into a 'DataSpanTreeList', by transforming the 'Tree.Leaf's. +toTreeList :: RegEx.MatchList a -> DataSpanTreeList a +toTreeList = map toTree where + toTree :: RegEx.Match a -> DataSpanTree a + toTree = fmap toDataSpan where + toDataSpan :: RegEx.MatchedData a -> DataSpan.DataSpan a + toDataSpan (_, inputDataOffset, inputData) = (inputData, (inputDataOffset, length inputData)) + +-- | Condenses a 'DataSpanTreeList's into a list of 'DataSpan.DataSpan's, using 'DataSpan.join'. +flattenTreeList + :: ConsumptionBounds.DataLength -- ^ The offset into the input-data at which a match occurred. + -> DataSpanTreeList a -- ^ The tree to flatten. + -> [DataSpan.DataSpan a] +flattenTreeList _ [] = [] +flattenTreeList offset (tree : treeList) = flattenedTree : flattenTreeList (DataSpan.after flattenedTree) treeList where +-- flattenedTree :: DataSpan.DataSpan a + flattenedTree = DataSpan.join offset $ Data.Foldable.toList tree + +{- | + * POSIX describes the contents of /capture-groups/, as summarised in . + + * 'RegEx.Result', is a complete description of the match between 'RegEx.InputData' & RegEx.ExtendedRegEx'; this function extracts a POSIX-conformant list from it. + + * The major differences are, that: + + Only data from parenthesized sub-expressions ('RegEx.Alternatives') is captured. + + Only the /last/ repetition of a repeated sub-expression is returned. + . + + The data captured within each parenthesized sub-expression, is summarised as a single 'DataSpan.DataSpan'. + + POSIX specifies a 'Span.Span'-offset of @-1@, for sub-expressions which match zero times; cf sub-expressions which consume nothing, once. + . + @ + ("ace" Text.Regex.Posix.=~ "a(b)*c(d)?e") :: Text.Regex.Base.RegexLike.MatchArray + array (0,2) [(0,(0,3)),(1,(-1,0)),(2,(-1,0))] + + ("ace" Text.Regex.Posix.=~ "a(b*)c(d?)e") :: Text.Regex.Base.RegexLike.MatchArray + array (0,2) [(0,(0,3)),(1,(1,0)),(2,(2,0))] + @ + I consider this a poor convention, resulting from the focus of POSIX on C, which makes subsequent calculation from the list of 'DataSpan's difficult & error-prone. +-} +extractCaptureGroups + :: Bool -- ^ Whether to strictly comply with /POSIX/. + -> DataSpanTreeList a -- ^ The tree-structure from which to extract the capture-groups. + -> [DataSpan.DataSpan a] +extractCaptureGroups complyStrictlyWithPosix = extractCaptureGroups' 0 where + extractCaptureGroups' :: ConsumptionBounds.DataLength -> DataSpanTreeList a -> [DataSpan.DataSpan a] + extractCaptureGroups' _ [] = [] + extractCaptureGroups' offset (tree : treeList) = let +-- recurseHorizontallyFrom :: ConsumptionBounds.DataLength -> [DataSpan.DataSpan a] + recurseHorizontallyFrom = (`extractCaptureGroups'` treeList) + in case tree of + Tree.Leaf dataSpan -> recurseHorizontallyFrom $ DataSpan.after dataSpan + Tree.Node [] -> DataSpan.empty (if complyStrictlyWithPosix then -1 else offset) : recurseHorizontallyFrom offset -- POSIX specifies an Span-offset of -1, for sub-expressions which match 0 times; cf sub-expressions which consumes nothing, once. + Tree.Node treeLists -> joinedFlattenedTreeList : (extractCaptureGroups' offset lastMatch {-recurse vertically-} ++ recurseHorizontallyFrom (DataSpan.after joinedFlattenedTreeList)) where +-- lastMatch :: DataSpanTreeList a + lastMatch = last treeLists -- . + +-- joinedFlattenedTreeList :: DataSpan.DataSpan a + joinedFlattenedTreeList = DataSpan.join offset $ flattenTreeList offset lastMatch + diff --git a/src-lib/RegExDot/ExecutionOptions.hs b/src-lib/RegExDot/ExecutionOptions.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/ExecutionOptions.hs @@ -0,0 +1,102 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * Defines those options relating to the "RegExDot.RegEx"-runtime. + + * These tend to be options which don't affect the result, only the means by which it is achieved. + + * Whilst similar structures are present in other regex-implementations, there's no standardisation of the fields. +-} + +module RegExDot.ExecutionOptions( +-- * Types +-- ** Data-types + ExecutionOptions(..), +-- * Functions +-- ** Mutators + setVerbose +) where + +import qualified ToolShed.Defaultable +import qualified ToolShed.Options + +-- | The switches used to control execution of the /regex/-engine. +data ExecutionOptions = MkExecutionOptions { + abortTrialRepetitionsOnInherentFailure :: Bool, -- ^ If an /alternative/ can't match, irrespective of the subsequent /RegExDot.RegEx.Concatenation/, then avoid futile trial /RegExDot.Repeatable.Repetitions/. The converse of 'checkForUnconsumableData'. + abortTrialRepetitionsOnZeroConsumption :: Bool, -- ^ Check for zero data-consumption by the @n@-th /RegExDot.Repeatable.Repeatable RegExDot.RegEx.CaptureGroup/, before attempting @n+1@. + bypassInputDataForLiberalConsumer :: Bool, -- ^ Whether to bypass reading of the input data, if the mapping to specific /RegExDot.RegEx.RepeatablePattern/s isn't required, & the /RegExDot.RegEx.ExtendedRegEx/ can consume the required quantity of anything. + catchIncompatibleAnchors :: Bool, -- ^ Avoid futile trial solutions, involving repetitions of anchored alternatives, which must consume data. + checkExistenceOfInelasticTail :: Bool, -- ^ If the /RegExDot.RegEx.ExtendedRegEx/ ends in an inelastic (zero /Star-height/) tail, confirm its existence at the end of the /RegExDot.RegEx.InputData/. + checkForUnconsumableData :: Bool, -- ^ Check whether there's no possibility of consuming some of the input data. The converse of 'abortTrialRepetitionsOnInherentFailure'. + moderateGreed :: Bool, -- ^ Greedily consume data, only up to the limit beyond which, future requirements would be compromised. + permitReorderingOfAlternatives :: Bool, -- ^ Permit /RegExDot.RegEx.Alternatives/ to be re-ordered, in an attempt to more quickly locate a result. + preferAlternativesWhichFeedTheGreedy :: Bool, -- ^ Within the /RegExDot.RegEx.MatchedData/ from which each candidate /RegExDot.RegEx.Match/ amongst sequences of /RegExDot.RegEx.Alternatives/, is ultimately composed, prefer /RegExDot.ConsumptionBounds.ConsumptionBounds/ of /RegExDot.RegEx.InputData/, beyond /RegExDot.Repeatable.getFewest/, by /RegExDot.Repeatable.isGreedy RegExDot.RegEx.RepeatablePattern/s. + preferAlternativesWhichMimickUnrolling :: Bool, -- ^ Compare /RegExDot.ConsumptionBounds.ConsumptionBounds/ on successive /RegExDot.Repeatable.Repetitions/ of /RegExDot.RegEx.CaptureGroup/, between candidate /RegExDot.RegEx.Match/es, to mimic the behaviour of the unrolled /RegExDot.Repeatable.Repetitions/. + preferFewerRepeatedAlternatives :: Bool, -- ^ Prefer fewer /RegExDot.Repeatable.Repetitions/ of /RegExDot.RegEx.Alternatives/, to discourage the capture of null lists of /RegExDot.RegEx.InputData/. + requireMatchList :: Bool, -- ^ If merely interested in a 'Bool' result, rather than the optimal mapping of input data to /RegExDot.RegEx.RepeatablePattern/s, avoid unnecessary evaluation of the /RegEx.Match/. + unrollRepeatedSingletonAlternative :: Bool, -- ^ Check whether /RegExDot.RegEx.Alternatives/ consists of just a singleton /RegExDot.RegEx.ExtendedRegEx/, & has therefore been used merely as a capture-group. Though this doesn't affect the result, it vastly improves efficiency. + useFirstMatchAmongAlternatives :: Bool, -- ^ Rather than performing an exhaustive search for the optimal choice amongst /RegExDot.RegEx.Alternatives/, merely select the first that matches; conform to /Perl/ rather than /POSIX/. + validateMinConsumptionOfAlternatives :: Bool -- ^ When the number of repetitions of a /RegExDot.RegEx.CaptureGroup/ is precisely specified, check whether the resulting minimum data-requirement is available. +} deriving (Eq, Show) + +instance ToolShed.Defaultable.Defaultable ExecutionOptions where + defaultValue = setVerbose False $ ToolShed.Options.blankValue { + abortTrialRepetitionsOnInherentFailure = True, -- Regrettably, this slightly reduces performance for most non-pathological patterns. + catchIncompatibleAnchors = True, + checkExistenceOfInelasticTail = True, + checkForUnconsumableData = True, -- Expensive, particularly when (not requireMatchList), & only typically useful in failure-scenarios. + moderateGreed = True, -- Cost may exceed benefit. TODO: confirm. + preferAlternativesWhichFeedTheGreedy = True, + preferAlternativesWhichMimickUnrolling = True, + preferFewerRepeatedAlternatives = True, + unrollRepeatedSingletonAlternative = True, -- Affects only efficiency, not the result. + useFirstMatchAmongAlternatives = False, -- Perl-style matching may be faster, but may also yield a sub-optimal Match. + validateMinConsumptionOfAlternatives = False -- The cost outweighs the small infrequent dividend. + } + +instance ToolShed.Options.Options ExecutionOptions where + blankValue = MkExecutionOptions { + abortTrialRepetitionsOnInherentFailure = undefined, + abortTrialRepetitionsOnZeroConsumption = undefined, + bypassInputDataForLiberalConsumer = undefined, + catchIncompatibleAnchors = undefined, + checkExistenceOfInelasticTail = undefined, + checkForUnconsumableData = undefined, + moderateGreed = undefined, + permitReorderingOfAlternatives = undefined, + preferAlternativesWhichFeedTheGreedy = undefined, + preferAlternativesWhichMimickUnrolling = undefined, + preferFewerRepeatedAlternatives = undefined, + requireMatchList = undefined, + unrollRepeatedSingletonAlternative = undefined, + useFirstMatchAmongAlternatives = undefined, + validateMinConsumptionOfAlternatives = undefined + } + +-- | Sets those fields which depend crucially on whether the caller wants to retrieve any /RegExDot.RegEx.MatchList/ from the /RegExDot.RegEx.Result/, or just query whether there is one. +setVerbose :: Bool -> ExecutionOptions -> ExecutionOptions +setVerbose verbose e = e { + abortTrialRepetitionsOnZeroConsumption = verbose, -- The corresponding check, involves evaluation of a /RegExDot.RegEx.MatchList/, which is too expensive if the /RegExDot.RegEx.Matchlist/ isn't otherwise required. + bypassInputDataForLiberalConsumer = not verbose, -- Potentially bypasses reading of /RegExDot.RegEx.InputData/, which is inappropriate if the mapping into a /RegExDot.RegEx.Result/ is required. + permitReorderingOfAlternatives = not verbose, -- Doesn't help when 'requireMatchList', since an exhaustive search of /RegExDot.RegEx.Alternatives/, for the optimal solution, is performed. + requireMatchList = verbose +} + diff --git a/src-lib/RegExDot/InstanceInt.hs b/src-lib/RegExDot/InstanceInt.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/InstanceInt.hs @@ -0,0 +1,43 @@ +{-# OPTIONS_GHC -fno-warn-orphans #-} +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] Provides a default implementation of the type-classes, required for 'RegEx.ExtendedRegEx' Int. + + [@CAVEATS@] + + In contrast to the traditional regex, there's no reason to escape a 'Char', other than when referring to a /shortcut/. + Since no /shortcuts/ are defined in this minimal definition, it's always an error to escape a character. +-} + +module RegExDot.InstanceInt() where + +import qualified RegExDot.BracketExpressionMember as BracketExpressionMember +import qualified RegExDot.Meta as Meta +import qualified RegExDot.RegEx as RegEx + +instance BracketExpressionMember.ShortcutExpander Int where + findPredicate _ = Nothing + +instance Meta.ShortcutExpander Int where + expand c = error $ "RegExDot.Meta.ShortcutExpander.expand RegExDot.InstanceInt:\tunrecognised shortcut '" ++ show c ++ "'." -- A shortcut is the only legitimate reason to escape a character. + +instance RegEx.ShortcutExpander Int where + expand c = error $ "RegExDot.RegEx.ShortcutExpander.expand RegExDot.InstanceInt:\tunrecognised shortcut '" ++ show c ++ "'." -- A shortcut is the only legitimate reason to escape a character. + diff --git a/src-lib/RegExDot/Meta.hs b/src-lib/RegExDot/Meta.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/Meta.hs @@ -0,0 +1,144 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * Describes the set of polymorphic data, which can be matched. + + * Only permits a match against exactly one polymorphic datum, which distinguishes it from a /zero-width assertion/, like an /anchor/, /word-boundary/, or /look-ahead assertion/. + + * Designed to be used by a polymorphic /regex/-engine, to implement the traditional meta-characters; @. [] [^]@. + + * Permits /Perl-style shortcuts/ for commonly used 'Meta'-data, to be canned & assigned a single-'Char' mnemonic for subsequent reference; + the implementation of 'Read' looks for a back-slashed 'Char', for which it expects there to be a corresponding canned 'Meta'. + Since this class is polymorphic, it has no knowledge of what shortcuts might be appropriate for the chosen type-parameter, + so the expansion from the back-slashed 'Char' to corresponding 'Meta'-data, is performed through the 'expand' interface of the 'ShortcutExpander' class, which should be implemented elsewhere. +-} + +module RegExDot.Meta( +-- * Type-classes + ShortcutExpander(..), +-- * Types +-- ** Data-types + Meta(..), +-- * Constants + shortcutToken, + anyToken, + tokens, +-- * Functions +-- ** Predicates + isMatch +) where + +import qualified Control.Arrow +import qualified Control.DeepSeq +import qualified RegExDot.BracketExpression as BracketExpression +import qualified RegExDot.BracketExpressionMember as BracketExpressionMember +import qualified RegExDot.Consumer as Consumer +import qualified RegExDot.ConsumptionProfile as ConsumptionProfile +import qualified RegExDot.ShowablePredicate as ShowablePredicate +import qualified ToolShed.SelfValidate + +{- | + * The interface via which /Perl-style shortcut/s are expanded, in a manner appropriate to the chosen type-parameter. + + * Since the expansion of /Perl-style shortcut/s, is more restricted inside than outside a 'BracketExpression.BracketExpression', + the former is considered to be a superclass, providing a base from which to build alternative implementations. +-} +class BracketExpressionMember.ShortcutExpander m => ShortcutExpander m where + expand :: Char -> Meta m -- ^ Expand a /Perl-style shortcut/. + +-- | Declares a polymorphic data-type. +data Meta m = + Any -- ^ Any datum matches. Equivalent to @NoneOf []@, but more efficient. CAVEAT: independent of the type-parameter @a@. + | Literal m -- ^ The datum matches, if it's equal to the specified value. Equivalent to @AnyOf [BracketExpression.Literal x]@, but more efficient. + | AnyOf (BracketExpression.BracketExpression m) -- ^ The datum matches, if 'BracketExpression.containsMatch'. + | NoneOf (BracketExpression.BracketExpression m) -- ^ The datum matches, if @not BracketExpression.containsMatch@. + | Predicate (ShowablePredicate.ShowablePredicate m) -- ^ The datum matches if 'ShowablePredicate.ShowablePredicate'. + deriving ( + Eq +-- Read, -- Specialised below. +-- Show -- Specialised below. + ) + +instance ToolShed.SelfValidate.SelfValidator (Meta m) where + getErrors _ = [] + +instance Show m => Show (Meta m) where + showsPrec _ Any = showChar anyToken + showsPrec _ (Literal m) = shows m + showsPrec _ (AnyOf bracketExpression) = shows bracketExpression + showsPrec _ (NoneOf bracketExpression) = showChar x . showChar BracketExpression.negationToken . showString xs where (x : xs) = show $ AnyOf bracketExpression + showsPrec _ (Predicate showablePredicate) = shows showablePredicate + +instance (ShortcutExpander m, Read m) => Read (Meta m) where + readsPrec _ [] = [] -- No parse. + readsPrec _ (' ' : s) = reads s -- Consume white-space. + readsPrec _ ('\t' : s) = reads s -- Consume white-space. + readsPrec _ ('.' : s) = [(Any, s)] + readsPrec _ ('[' : '^' : noneOf) = Control.Arrow.first NoneOf `map` reads (fst BracketExpression.delimiterTokens : noneOf) {-Reconstruct without negation, & recurse-} + readsPrec _ anyOf@('[' : _) = Control.Arrow.first AnyOf `map` reads anyOf {-singleton-} + readsPrec _ ('\\' : c : s) = [(expand c, s)] + readsPrec _ literal = Control.Arrow.first Literal `map` reads literal {-singleton-} + +instance Consumer.Consumer (Meta m) where + consumptionProfile meta = let + hasSpecificDataRequirement :: Bool + hasSpecificDataRequirement = case meta of + Any -> False + NoneOf [] -> False + _ -> True + in ConsumptionProfile.MkConsumptionProfile { + ConsumptionProfile.consumptionBounds = (1, Just 1), + ConsumptionProfile.hasSpecificRequirement = hasSpecificDataRequirement, + ConsumptionProfile.canConsumeAnything = not hasSpecificDataRequirement + } + + starHeight _ = 0 + +instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (Meta m) where + rnf Any = () + rnf (Literal m) = Control.DeepSeq.rnf m + rnf (AnyOf bracketExpression) = Control.DeepSeq.rnf bracketExpression + rnf (NoneOf bracketExpression) = Control.DeepSeq.rnf bracketExpression + rnf (Predicate showablePredicate) = Control.DeepSeq.rnf showablePredicate + +-- | True if the specified datum matches. +isMatch :: Eq m + => m -- ^ The input datum. + -> Meta m -- ^ The meta-entity against which the input datum is to be matched. + -> Bool -- ^ The result of the match-operation. +isMatch _ Any = True +isMatch datum (Literal literal) = datum == literal +isMatch datum (AnyOf bracketExpression) = datum `BracketExpression.containsMatch` bracketExpression +isMatch datum (NoneOf bracketExpression) = not $ datum `isMatch` AnyOf bracketExpression -- This implementation leverages future enhancements to 'AnyOf'. +isMatch datum (Predicate showablePredicate) = ShowablePredicate.predicate showablePredicate datum + +-- | The token used to precede a /Perl-style shortcut/, when in the 'String'-form. +shortcutToken :: Char +shortcutToken = '\\' + +-- | The token used to denote 'Any', when in the 'String'-form. +anyToken :: Char +anyToken = '.' + +-- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. +tokens :: String +tokens = [shortcutToken, anyToken] ++ BracketExpression.tokens + diff --git a/src-lib/RegExDot/RegEx.hs b/src-lib/RegExDot/RegEx.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/RegEx.hs @@ -0,0 +1,1198 @@ +{- + Copyright (C) 2010-2015 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * This implementation of extended /regex/es, generalises the familiar concept of pattern-matching of character-strings, to matching lists composed from an arbitrary data-type. + The polymorphic data, from which the input data-list is composed, need only support @Eq@. + + * Because of the unknown stringified form of the underlying polymorphic data, + the /regex/ must be described by a comparatively verbose bracketed & comma-separated list, rather than the traditional /String/ containing Meta-characters. + Each element of this 'Concatenation' is a 'RepeatablePattern', which describes a permissible match against 'InputData'. + + * 'RepeatablePattern' can take one of two forms. + In the simplest case, it matches just a single item of the underlying polymorphic type, perhaps literally, though looser specifications also exist: + @.@ matches any input datum; @[x, y, z]@ matches any of @x@, @y@, or @z@; @[^x, y, z]@ matches anything but @x@, @y@, or @z@. + To support POSIX /ERE/s, 'RepeatablePattern' can also be a list 'Alternatives', each of which is recursively defined as an 'ExtendedRegEx', to form a tree-structure. + + * Each 'Pattern', can optionally be /quantified/ by either a traditional /greedy/, or a /Perl/-style /non-greedy/, suffix, e.g.; @[*, +, ?, {n, m}, {n,}, {n}, *?, +?, ??, {n, m}?, {n,}?]@. + + * For convenience, common specifications can be canned & assigned a single 'Char' mnemonic, for subsequent reference. + Since 'ExtendedRegEx' is polymorphic, the set of abbreviations appropriate in the context of the unspecified base-type, + must be implemented externally through the 'Meta.ShortcutExpander' interface. + This permits the use, when the type-parameter is 'Char', of /Perl-style shortcuts/ @[\\d\\D\\s\\S\\w\\W]@. + + * The algorithm, is the classic /back-tracking/ one, rather than either a /DFA/ or /NFA/. + This permits construction of 'Result' via which one can discover the deep mapping of 'InputData' into 'ExtendedRegEx', + & provides the flexibility to add the features now expected by modern /regex/-engines. + Since the type-parameter is unknown, & may represent a large object, the exponential space-complexity of creating a /DFA/ may present additional problems. + The exponential time-complexity of the /back-tracking/ algorithm is partially tamed by targeting obvious inefficiencies with specific optimisations. + + * Char-based regexen, traditionally overload the delimiters of a set of 'Alternatives' (parentheses), as a request for data-capture. + Here, in contrast, all 'RepeatablePattern's capture data, & repeated sub-expressions capture a list of data, + rather than arbitrarily recording just the last () item. + + [@REFERENCES@] + + * + + * + + * + + * + + * + + * + + * + + [@CAVEATS@] + + * Because of the definition of mutually recursive data-types, it is difficult to split this annoyingly large module, & preserve compatibility across compilers, + but it may be possible to break this cyclic dependency, by defining an interface to which one of the data-types defined here conforms. + + * Doesn't implement /Back-references/, making the definition of the 'ExtendedRegEx' context-free. + + * There's no integration with the type-classes defined in "Text.Regex.Base.RegexLike", which assumes 'Char'-based 'InputData'; + though this could be added to a specialised instance. + + * When 'Alternatives' are defined, 'Result' becomes a tree-like structure. + Unless the alternative is a singleton, the specific alternative selected in the solution is typically unknown, & therefore the /structure/ of the branch of this tree is also unknown. + This lack of clarity is compounded when the 'Alternatives' are 'Repeatable.Repeatable', since a different one may be selected on each successive repetition. + Consequently, the user can't navigate this portion of the structure in a statically defined manner, to acquire the captured data. + Despite this, & in contrast to other /regex/-engines, access to the whole data-structure is available, since it doesn't seem advantage to hide it. + The user can then either use 'extractDataFromMatch' for that element of 'Result', thus aggregating the data from sections of unknown structure, or 'show' it, as an aid to debugging. + + [@TODO@] + + * Test parallel-operation, on a 3 or more processor machine. + If 'rnf' is less effective than 'rwhnf', + then the 'Control.DeepSeq.NFData' context can be removed, + reducing the requirements imposed on the type-parameter 'a'. + + * Try 'Data.List.Stream' (stream-fusion), a faster drop-in replacement for 'Data.List'; possibly integrated in GHC-6.12. + + * 'ExecutionOptions.bypassInputDataForLiberalConsumer' is too restrictive. + More generally, we can test whether the set of different 'a' in 'InputData', is a subset of those common to all remaining terms in the 'ExtendedRegEx'. + Using this rule, we can infer @"aaa ..." =~ MkExtendedRegEx [a,a+,a?,[ab]{2,3}]@, given compatible 'ConsumptionProfile.consumptionBounds'. + + * Nested repetitions, where nothing has been added to the expression, result in repeated trials of the same expression, + e.g.; @"(x{i,}){j,}"@ results in the same expansion for @(i, j) in [(2, 3), (3, 2), (6, 1), (1, 6)]@. + The resulting 'MatchList' may be different, but if the first such trial fails, so will all the remainder. + + * Should cope with empty sets of 'Alternatives' & zero repetitions, neither of which can ever match, but the wider pattern can, e.g. @(()|x{0}|y)@. + + * By removing 'RepeatablePattern' from 'Match', it can be isolated in a new module. + This would result in a significant loss of discoverability. + + * Expand repeated 'Anchor.Bow' with @fewest - 1@ null matches followed by recursive 'findMatch'-call with @repetitions = 1@. +-} + +module RegExDot.RegEx( +-- * Type-classes + ShortcutExpander(..), +-- * Types +-- ** Type-synonyms +-- BasicRegEx, + Concatenation, + ExternalMatch, + InputData, + MatchedData, + MatchList, +-- MetaDataList, + RepeatablePattern, +-- Transformation, +-- ** Data-types + Alternatives(..), + Match, + ExtendedRegEx(..), + Pattern(..), + Result, +-- * Constants + alternativeExtendedRegExSeparatorToken, +-- anyDatum, + captureGroupDelimiters, + tokens, +-- * Functions +-- accumulateDistinctInputData, +-- accumulateDistinctMetaDataFrom, + dock, +-- drift, +-- findMatch, + captureGroup, +-- mkNullMatchFromConcatenation, +-- mkNullMatchFromExtendedRegEx, +-- mkNullMatchFromRepeatablePattern, +-- safeReciprocal, +-- shiftMatchedData, +-- shiftMatch, + shiftMatchList, + showsMaybeAnchor, + simply, + transformExtendedRegEx, +-- ** Accessors (Deconstructors) + deconstructAlternatives, +-- getInputData, +-- ** Operators + (+~), + (=~), + (/~), + (.*), + (.*?), +-- (/+) +-- ** Predicates +-- hasBowAnchor, +-- hasSternAnchor, + isDefined, + isCaptureGroup, + isSingletonAlternatives, +-- isUnconsumableByAnyOf, +-- ** Query +-- getDistinctMetaDataFromAlternatives, +-- getDistinctMetaDataFromConcatenation, +-- getDistinctMetaDataFromExtendedRegEx, +-- getDistinctMetaDataFromPattern, +-- getDistinctMetaDataFromRepeatablePattern, + externalMatchLength, + extractDataFromMatch, + extractDataFromMatch', + extractDataFromMatchList +) where + +import Control.Arrow((&&&)) +import qualified Control.Arrow +import qualified Control.DeepSeq +import qualified Control.Parallel.Strategies +import qualified Data.Char +import qualified Data.Foldable +import qualified Data.List +import qualified Data.Maybe +import qualified Data.Ord +import qualified RegExDot.Anchor as Anchor +import qualified RegExDot.Consumer as Consumer +import qualified RegExDot.ConsumptionBounds as ConsumptionBounds +import qualified RegExDot.ConsumptionProfile as ConsumptionProfile +import qualified RegExDot.ExecutionOptions as ExecutionOptions +import qualified RegExDot.Meta as Meta +import qualified RegExDot.RegExOpts as RegExOpts +import qualified RegExDot.Repeatable as Repeatable +import RegExDot.Repeatable((^#->)) +import qualified RegExDot.Tree as Tree +import qualified ToolShed.Data.List +import qualified ToolShed.Data.List.Splits +import qualified ToolShed.SelfValidate + +infix 4 +~, =~, /~ -- Same as (==) & (/=). + +-- | The type of a /regex/, in which there's no provision for either 'Alternatives' or 'Anchor.Anchor's. +type BasicRegEx m = [Repeatable.Repeatable (Meta.Meta m)] + +{- | + * Defines the method required to expand a mnemonic into an 'ExtendedRegEx'. + + * CAVEAT: this interface must be declared locally, since it references 'ExtendedRegEx', & 'ExtendedRegEx' references it. +-} +class ShortcutExpander m where + expand :: Char -> ExtendedRegEx m -- ^ Expand a single-'Char' mnemonic into the corresponding 'ExtendedRegEx'. + +{- | + * Represents the /alternation/ feature of 'ExtendedRegEx's. + + * One could amalgamate this with 'Pattern', since it seems to exist merely as a peg to hang /instance/-declarations from. +-} +newtype Alternatives m = MkAlternatives [ExtendedRegEx m] deriving Eq + +-- | Accessor, to expose the guts. +deconstructAlternatives :: Alternatives m -> [ExtendedRegEx m] +deconstructAlternatives (MkAlternatives extendedRegExList) = extendedRegExList + +-- | Similar to 'fmap', but operates on ['ExtendedRegEx'], rather than just @a@. +transformAlternatives + :: ([ExtendedRegEx m] -> [ExtendedRegEx m]) -- ^ The transformation function to apply to the list of data behind the constructor. + -> Alternatives m + -> Alternatives m +transformAlternatives f = MkAlternatives . f . deconstructAlternatives + +instance ( + ShortcutExpander m, + Meta.ShortcutExpander m, + Eq m, + Read m + ) => Read (Alternatives m) where + readsPrec _ (' ' : s) = reads s -- Consume white-space. + readsPrec _ ('\t' : s) = reads s -- Consume white-space. + readsPrec _ s = case reads s of + [(extendedRegEx, s1)] -> case dropWhile Data.Char.isSpace s1 of + ('|' : s2) -> Control.Arrow.first (transformAlternatives (extendedRegEx :)) `map` reads s2 {-singleton-} + _ -> [(MkAlternatives [extendedRegEx], s1)] + _ -> [] -- No parse. + +instance Show m => Show (Alternatives m) where + showsPrec _ = foldl (.) (showString "") . Data.List.intersperse (showChar alternativeExtendedRegExSeparatorToken) . map shows . deconstructAlternatives -- Replace the default list-format, with 'egrep'-syntax. + +instance Consumer.Consumer (Alternatives m) where + consumptionProfile = Consumer.aggregateConsumptionProfilesFromAlternatives . deconstructAlternatives + starHeight = Consumer.starHeight . deconstructAlternatives -- Must evaluate all Alternatives to determine best. + +instance ToolShed.SelfValidate.SelfValidator (Alternatives m) where + getErrors = ToolShed.SelfValidate.getErrors . deconstructAlternatives + +instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (Alternatives m) where + rnf = Control.DeepSeq.rnf . deconstructAlternatives + +-- | 'Alternatives' can be employed as a simple /capture-group/ as well as a switch, under which circumstances there's no choice amongst multiple 'Alternatives'. +isSingletonAlternatives :: Alternatives m -> Bool +isSingletonAlternatives = (== 1) . length . deconstructAlternatives + +-- | A set of 'Meta.Meta' which can be consumed. +type MetaDataList m = [Meta.Meta m] + +{- | + * 'True' if there's no possibility, that the specified input datum, can be consumed by any of the specified 'MetaDataList'. + + * Because data may not be consumed, perhaps because of unsuitable 'ConsumptionBounds.ConsumptionBounds', the converse doesn't hold, + i.e. @not isUnconsumableByAnyOf@ /doesn't/ imply that the input datum can be consumed, just that it hasn't been proven that it can't. +-} +isUnconsumableByAnyOf :: Eq m + => m -- ^ The input datum. + -> MetaDataList m -- ^ The list of meta-data against any of which a match can occur. + -> Bool +isUnconsumableByAnyOf i = not . any (Meta.isMatch i) + +-- | The set of distinct 'Meta.Meta', in the specified 'Alternatives'. +getDistinctMetaDataFromAlternatives :: Eq m => Alternatives m -> MetaDataList m +getDistinctMetaDataFromAlternatives = foldr (Data.List.union . getDistinctMetaDataFromExtendedRegEx) [] . deconstructAlternatives + +-- | Defines either a simple 'Meta.Meta', which can match exactly one datum, or a set of 'Alternatives', each of which is recursively defined above, as an 'ExtendedRegEx'. +data Pattern m = + Require (Meta.Meta m) -- ^ Describes a requirement for a simple scalar datum of the polymorphic type. + | CaptureGroup (Alternatives m) -- ^ A sub-expression containing a selection of recursively defined alternatives, thus forming a tree-structure. + deriving Eq + +instance ( + Eq m, + Meta.ShortcutExpander m, + Read m, + ShortcutExpander m + ) => Read (Pattern m) where + readsPrec _ (' ' : s) = reads s -- Consume white-space. + readsPrec _ ('\t' : s) = reads s -- Consume white-space. + readsPrec _ ('(' : s) = case {-Alternatives.-} reads s of + [(alternatives, s1)] -> case dropWhile Data.Char.isSpace s1 of + (')' : s2) -> [(CaptureGroup alternatives, s2)] + _ -> [] -- No parse. + _ -> [] -- No parse. + readsPrec _ s = case reads s of + [pair] -> [Control.Arrow.first Require pair] + _ -> [] -- No parse. + +instance Show m => Show (Pattern m) where + showsPrec _ (Require meta) = shows meta + showsPrec _ (CaptureGroup alternatives) = showChar (fst captureGroupDelimiters) . shows alternatives . showChar (snd captureGroupDelimiters) + +instance Consumer.Consumer (Pattern m) where + consumptionProfile (Require meta) = Consumer.consumptionProfile meta + consumptionProfile (CaptureGroup alternatives) = Consumer.consumptionProfile alternatives + + starHeight (Require meta) = Consumer.starHeight meta + starHeight (CaptureGroup alternatives) = Consumer.starHeight alternatives + +instance ToolShed.SelfValidate.SelfValidator (Pattern m) where + getErrors (Require meta) = ToolShed.SelfValidate.getErrors meta + getErrors (CaptureGroup alternatives) = ToolShed.SelfValidate.getErrors alternatives + +instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (Pattern m) where + rnf (Require meta) = Control.DeepSeq.rnf meta + rnf (CaptureGroup alternatives) = Control.DeepSeq.rnf alternatives + +-- | Convenience-function to build a 'CaptureGroup' from a list of alternative 'ExtendedRegEx's. +captureGroup :: [ExtendedRegEx m] -> Pattern m +captureGroup = CaptureGroup . MkAlternatives + +-- | True if the 'Pattern' was constructed via 'CaptureGroup'. +isCaptureGroup :: Pattern m -> Bool +isCaptureGroup (CaptureGroup _) = True +isCaptureGroup _ = False + +-- | The set of distinct 'Meta.Meta', in the specified 'Pattern'. +getDistinctMetaDataFromPattern :: Eq m => Pattern m -> MetaDataList m +getDistinctMetaDataFromPattern pattern = case pattern of + Require meta -> [meta] + CaptureGroup alternatives -> getDistinctMetaDataFromAlternatives alternatives + +-- | Constant pattern, representing a lax 'Require'ment. +anyDatum :: Pattern a +anyDatum = Require Meta.Any + +-- | Make 'Pattern's, 'Repeatable.Repeatable'. +type RepeatablePattern m = Repeatable.Repeatable (Pattern m) + +-- | Construct a null 'Match' by assuming that the specified 'RepeatablePattern' consumes zero 'InputData'; of which the caller should ensure it's capable. +mkNullMatchFromRepeatablePattern + :: ConsumptionBounds.DataLength -- ^ The offset into the list of input data at which the zero-length match occurred. + -> RepeatablePattern m -- ^ The pattern that the input data matched. + -> Match m -- ^ The resulting match-structure. +mkNullMatchFromRepeatablePattern offset r = case Repeatable.base r of + Require _ -> Tree.Leaf (r, offset, []) + CaptureGroup _ -> Tree.Node $ Repeatable.getFewest r `replicate` [] + +-- | The set of distinct 'Meta.Meta', in the specified 'RepeatablePattern'. +getDistinctMetaDataFromRepeatablePattern :: Eq m => RepeatablePattern m -> MetaDataList m +getDistinctMetaDataFromRepeatablePattern = getDistinctMetaDataFromPattern . Repeatable.base + +{- | + * Represents a black hole, which will greedily consume all data. + + * CAVEAT: nullary, i.e. a constant. +-} +(.*) :: RepeatablePattern m +(.*) = Repeatable.zeroOrMore anyDatum + +-- | A /non-greedy/ version of '.*'. +(.*?) :: RepeatablePattern m +(.*?) = Repeatable.zeroOrMore' anyDatum + +-- | Convenience-function, to build a 'RepeatablePattern' from an unrepeated instance of the specified 'Meta.Meta'-datum. +simply :: Meta.Meta m -> RepeatablePattern m +simply = Repeatable.one . Require + +-- | Represents the /concatenation/ aspect of 'ExtendedRegEx's. +type Concatenation m = [RepeatablePattern m] + +-- | Construct a null 'Match' by assuming that the specified 'Concatenation' consumes zero 'InputData'. +mkNullMatchFromConcatenation + :: ConsumptionBounds.DataLength -- ^ The offset into the list of input data at which the zero-length match occurred. + -> Concatenation m -- ^ The list of repeatable patterns that the input data matched. + -> MatchList m -- ^ The resulting match-structure. +mkNullMatchFromConcatenation offset = map $ mkNullMatchFromRepeatablePattern offset + +-- | The set of distinct 'Meta.Meta', in the specified 'Concatenation'. +getDistinctMetaDataFromConcatenation :: Eq m => Concatenation m -> MetaDataList m +getDistinctMetaDataFromConcatenation = foldr (Data.List.union . getDistinctMetaDataFromRepeatablePattern) [] + +-- | The accumulating set of 'MetaDataList', resulting from each successive 'RepeatablePattern' in the specified 'Concatenation'. +accumulateDistinctMetaDataFrom :: Eq m => MetaDataList m -> Concatenation m -> [MetaDataList m] +accumulateDistinctMetaDataFrom = scanr (Data.List.union . getDistinctMetaDataFromRepeatablePattern) + +-- | Constructs an 'ExtendedRegEx', by surrounding a 'Concatenation' with optional 'Anchor.Anchor's. +data ExtendedRegEx m = MkExtendedRegEx { + bowAnchor :: Maybe Anchor.Anchor, -- ^ An option to anchor the /regex/ to the start of the 'InputData'. + concatenation :: Concatenation m, -- ^ The sequence of 'RepeatablePattern's defining the 'Require'ments that the 'InputData' must meet. + sternAnchor :: Maybe Anchor.Anchor -- ^ An option to anchor the /regex/ to the end of the 'InputData'. +} deriving Eq + +-- | The type of a function which transforms an 'ExtendedRegEx'. +type Transformation m = ExtendedRegEx m -> ExtendedRegEx m + +-- | Similar to 'fmap', but operates on 'Concatenation', rather than just @a@. +transformExtendedRegEx + :: (Concatenation m -> Concatenation m) -- ^ The function used to transform the data behind the constructor. + -> Transformation m +transformExtendedRegEx f extendedRegEx = extendedRegEx { concatenation = f $ concatenation extendedRegEx } + +instance ( + Eq m, + Meta.ShortcutExpander m, + Read m, + ShortcutExpander m + ) => Read (ExtendedRegEx m) where + readsPrec _ [] = [] + readsPrec _ (' ' : s) = reads s -- Consume white-space. + readsPrec _ ('\t' : s) = reads s -- Consume white-space. + readsPrec _ ('\\' : c : s) = [(expand c, s)] + readsPrec _ s = case singleton of + [(extendedRegEx, _)] + | ToolShed.SelfValidate.isValid extendedRegEx -> singleton + | otherwise -> error $ ToolShed.SelfValidate.getFirstError extendedRegEx -- Parsed OK, but invalid. + _ -> [] -- No parse. + where +-- singleton :: (Eq m, Meta.ShortcutExpander m, Read m, ShortcutExpander m) => [(ExtendedRegEx m, String)] + singleton = [ + ( + MkExtendedRegEx { + bowAnchor = maybeBowAnchor, + concatenation = concatenation', + sternAnchor = maybeSternAnchor + }, + remainder + ) | + (maybeBowAnchor, s1) <- readsMaybeAnchor s, + (concatenation', s2) <- reads s1, + (maybeSternAnchor, remainder) <- readsMaybeAnchor s2 + ] {-list-comprehension-} where + readsMaybeAnchor :: ReadS (Maybe Anchor.Anchor) + readsMaybeAnchor s' = return {-to List-monad-} $ case reads s' of + [pair] -> Control.Arrow.first Just pair + _ -> (Nothing, s') + +-- | Shows either the specified 'Anchor.Anchor', or a null string where 'Nothing' is specified. +showsMaybeAnchor :: Maybe Anchor.Anchor -> String -> String +showsMaybeAnchor maybeAnchor + | Data.Maybe.isJust maybeAnchor = shows $ Data.Maybe.fromJust maybeAnchor + | otherwise = id + +instance Show m => Show (ExtendedRegEx m) where + showsPrec _ MkExtendedRegEx { + bowAnchor = maybeBowAnchor, -- CAVEAT: this could be 'Nothing' or perversely an 'Anchor.Stern'. + concatenation = concatenation', + sternAnchor = maybeSternAnchor -- CAVEAT: this could be 'Nothing' or perversely an 'Anchor.Bow'. + } = showsMaybeAnchor maybeBowAnchor . shows concatenation' . showsMaybeAnchor maybeSternAnchor + +instance Consumer.Consumer (ExtendedRegEx m) where + consumptionProfile = Consumer.aggregateConsumptionProfilesFromConcatenation . concatenation + starHeight = Consumer.starHeight . concatenation + +instance ToolShed.SelfValidate.SelfValidator (ExtendedRegEx m) where + getErrors = ToolShed.SelfValidate.getErrors . concatenation + +instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (ExtendedRegEx m) where + rnf MkExtendedRegEx { + bowAnchor = maybeBowAnchor, + concatenation = concatenation', + sternAnchor = maybeSternAnchor + } = Control.DeepSeq.rnf (maybeBowAnchor, concatenation', maybeSternAnchor) + +-- | Drop 'Anchor.Anchor's at both bow & stern of the specified 'ExtendedRegEx'. +dock :: Transformation m +dock e = e { + bowAnchor = Just Anchor.Bow, + sternAnchor = Just Anchor.Stern +} + +-- | Amend the 'Concatenation' in the specified 'ExtendedRegEx', by prepending '.*' where there's a missing top-level 'Anchor.Bow' & appending '.*?' where there's a missing top-level 'Anchor.Stern'. +drift :: Transformation m +drift extendedRegEx = transformExtendedRegEx ( + ( + if hasBowAnchor extendedRegEx then id else {-no anchor-} ((.*?) :) + ) . ( + if hasSternAnchor extendedRegEx then id else {-no anchor-} (++ [(.*)]) + ) + ) extendedRegEx + +-- | True if the specified 'ExtendedRegEx' has an initial 'Anchor.Bow'. CAVEAT: though typically the alternative is no 'Anchor.Anchor', there may perversely be an initial 'Anchor.Stern', but neither qualify. +hasBowAnchor :: ExtendedRegEx m -> Bool +hasBowAnchor = (== Just Anchor.Bow) . bowAnchor + +-- | True if the specified 'ExtendedRegEx' has a terminal 'Anchor.Stern'. CAVEAT: though typically the alternative is no 'Anchor.Anchor', there may perversely be a terminal 'Anchor.Bow', but neither qualify. +hasSternAnchor :: ExtendedRegEx m -> Bool +hasSternAnchor = (== Just Anchor.Stern) . sternAnchor + +-- | True if there's at least one 'RepeatablePattern' in the 'Concatenation', i.e. that it's non-null. +isDefined :: ExtendedRegEx m -> Bool +isDefined = not . null . concatenation + +-- | Construct a null 'Match' by assuming that the specified 'ExtendedRegEx' consumes zero 'InputData'. +mkNullMatchFromExtendedRegEx :: ConsumptionBounds.DataLength -> ExtendedRegEx m -> MatchList m +mkNullMatchFromExtendedRegEx offset = mkNullMatchFromConcatenation offset . concatenation + +-- | The set of distinct (i.e. unique) 'Meta.Meta', in the specified 'ExtendedRegEx'. +getDistinctMetaDataFromExtendedRegEx :: Eq m => ExtendedRegEx m -> MetaDataList m +getDistinctMetaDataFromExtendedRegEx = getDistinctMetaDataFromConcatenation . concatenation + +{- | + * The input-data is just a list. + + * Whilst typically this list is also a 'String', & could therefore be more efficiently implemented using "Data.ByteString", + we can't assume that the polymorphic base-type is always 'Char'. +-} +type InputData m = [m] + +-- | The accumulating sets of distinct input data. +accumulateDistinctInputData :: Eq m => InputData m -> [InputData m] +accumulateDistinctInputData = scanr (Data.List.union . return) [] + +-- | Tag the 'InputData' with the 'RepeatablePattern' it matched (which unfortunately confines the definition to this (bloated) module), & the offset from the start of the data; +type MatchedData m = (RepeatablePattern m, ConsumptionBounds.DataLength, InputData m) + +-- | Shifts the offset of the specified 'MatchedData'. +shiftMatchedData + :: ConsumptionBounds.DataLength -- ^ The offset by which to shift the position into the input-data at which a match occurred. + -> MatchedData m -- ^ The match-structure whose offset is to be shifted. + -> MatchedData m +shiftMatchedData i (r, offset, d) = (r, offset + i, d) + +-- | Accessor. +getInputData :: MatchedData m -> InputData m +getInputData (_, _, inputData) = inputData + +-- | Describes the manner in which a 'RepeatablePattern' successfully consumed 'InputData'. +type Match m = Tree.Tree (MatchedData m) + +-- | Shifts the offsets of all the 'MatchedData' contained in the specified 'Match'. +shiftMatch + :: ConsumptionBounds.DataLength -- ^ The offset by which to shift the position into the input-data at which a match occurred. + -> Match m -- ^ The match-structure whose offset is to be shifted. + -> Match m +shiftMatch i = (shiftMatchedData i `fmap`) + +-- | Extract & concatenate, the 'InputData' from a 'Match'. +extractDataFromMatch :: Match m -> InputData m +extractDataFromMatch = Data.Foldable.foldMap getInputData -- Uses the List-monoid's associative binary operator (++), to concatenate the values returned by the specified function. + +-- | Extract & concatenate, the 'InputData' from a 'Match'; null if it didn't match any. +extractDataFromMatch' :: Maybe (Match m) -> InputData m +extractDataFromMatch' (Just match) = extractDataFromMatch match +extractDataFromMatch' _ = [] + +-- | Describes the manner in which a 'Concatenation' successfully consumed 'InputData'. +type MatchList m = [Match m] + +-- | Shifts the offsets of all the 'MatchedData' contained in the specified 'MatchList'. +shiftMatchList + :: ConsumptionBounds.DataLength -- ^ The offset by which to shift the position into the input-data at which a each listed match occurred. + -> MatchList m -- ^ The list of match-structures, each of whose offsets are to be shifted. + -> MatchList m +shiftMatchList i = map (shiftMatch i) + +-- | Extract & concatenate, the 'InputData', from the 'MatchList'. +extractDataFromMatchList :: MatchList m -> InputData m +extractDataFromMatchList = concatMap extractDataFromMatch -- CAVEAT: too fine-grain for effective data-parallelism. + +-- | At the top-level of an 'ExtendedRegEx', the lack of an 'Anchor.Anchor' allows the 'ExtendedRegEx' to drift away from the corresponding end of the input-data; this data-gap is captured here. +type ExternalMatch m = Maybe (Match m) + +-- | Captures the list of input-data consumed by the 'Concatenation', bracketed by any data-prefix or data-suffix. +type Result m = (ExternalMatch m, Maybe (MatchList m), ExternalMatch m) + +-- | Returns the length of data consumed by the specified 'ExternalMatch'. +externalMatchLength :: ExternalMatch m -> ConsumptionBounds.DataLength +externalMatchLength = length . extractDataFromMatch' + +{- | + * Similar to '+~', but exposes 'ExecutionOptions.ExecutionOptions', permitting greater control; & enough rope to hang yourself. + + * The parameter-order facilitates partial application, to acquire a matcher-function for a specific 'ExtendedRegEx', which can be applied repeatedly to different 'InputData'. + + * Performs some one-off @O(n)@ time-complexity preparation, then delegates the grunt-work to a private recursive function. + + * One could waste a lot of time trying to consume 'InputData' from the head of a high-'Consumer.starHeight' 'ExtendedRegEx', + only to find that there's a low-'Consumer.starHeight' tail that can't ever match the 'InputData'; checks this first before delegating. +-} +findMatch :: (Eq m, Control.DeepSeq.NFData m) + => RegExOpts.RegExOpts (ExtendedRegEx m) -- ^ The match-options parameterised by the regex against which to match the input data. + -> InputData m -- ^ The input data within which to locate a match. + -> Maybe (MatchList m) +findMatch regExOpts@RegExOpts.MkRegExOpts { + RegExOpts.executionOptions = executionOptions, + RegExOpts.regEx = extendedRegEx +} originalInputData + | let + extractLowStarHeightTail :: ExtendedRegEx m -> Concatenation m + extractLowStarHeightTail = {-#SCC "extractLowStarHeightTail" #-} fromConcatenation . reverse . concatenation where + fromConcatenation :: Concatenation m -> Concatenation m + fromConcatenation [] = [] + fromConcatenation (repeatablePattern : concatenation') + | Consumer.starHeight repeatablePattern == 0 = repeatablePattern : fromConcatenation {-recurse-} concatenation' + | Repeatable.getFewest repeatablePattern > 0 = case Repeatable.base repeatablePattern of + CaptureGroup alternatives + | isSingletonAlternatives alternatives -> extractLowStarHeightTail {-recurse-} . head $ deconstructAlternatives alternatives + | otherwise -> {-non-singleton-} [] + _ -> [] -- Zero 'Consumer.StarHeight' would have been detected by 'fromConcatenation'. + | otherwise = [] + +-- lowStarHeightTail :: Concatenation a + lowStarHeightTail = extractLowStarHeightTail extendedRegEx + in and [ + ExecutionOptions.checkExistenceOfInelasticTail executionOptions, + not $ null lowStarHeightTail, -- Prevent infinite recursion. + reverse originalInputData /~ RegExOpts.mkRegEx MkExtendedRegEx { + bowAnchor = Just Anchor.Bow, + concatenation = lowStarHeightTail, + sternAnchor = Nothing + } -- Check for mismatch with the corresponding tail of 'InputData'. + ] = Nothing + | otherwise = findMatchSlave originalConcatenation ( + Consumer.accumulateConsumptionProfiles originalConcatenation + ) ( + accumulateDistinctMetaDataFrom [] originalConcatenation + ) originalInputData originalInputDataLength ( + accumulateDistinctInputData originalInputData + ) + where +-- originalConcatenation :: Concatenation a + originalConcatenation = concatenation extendedRegEx + + originalInputDataLength :: ConsumptionBounds.DataLength + originalInputDataLength = length originalInputData +{- + This beast takes a lot of parameters. + Whilst the only strictly necessary parameters are the lists 'Concatenation' & 'InputData', many ancillary lists derived from them, are also passed by parameter. + These are proportionally reduced on recursion, to avoid the requirement to regenerate them each time they're required. +-} + findMatchSlave :: (Eq m, Control.DeepSeq.NFData m) + => Concatenation m -- The list of 'RepeatablePattern's from which the regex is constructed. + -> ConsumptionProfile.AccumulatedConsumptionProfiles -- The capacity-bounds for the consumption of 'InputData', that the 'Concatenation' extending right from any given 'RepeatablePattern', can consume. + -> [MetaDataList m] -- The set of distinct 'Meta.Meta'-data, in the 'Concatenation extending right from any given 'RepeatablePattern'. + -> InputData m -- The input data, which will be fed to the 'Concatenation'. + -> ConsumptionBounds.DataLength -- The length of the previously specified 'InputData'. + -> [InputData m] -- The set of distinct input data, extending right from any given point. + -> Maybe (MatchList m) + findMatchSlave [] _ _ [] _ _ = Just [] -- Simultaneous exhaustion of the 'Concatenation' of 'RepeatablePattern's & the 'InputData' => success. + findMatchSlave [] _ _ _ _ _ = Nothing + findMatchSlave concatenation'@(repeatablePatternHead : concatenationTail) (accumulatedConsumptionProfileHead : accumulatedConsumptionProfilesTail) (distinctMetaDataHead : distinctMetaDataTail) inputData inputDataLength distinctInputData@(distinctInputDataHead : _) + | not $ inputDataLength `ConsumptionProfile.withinConsumptionBounds` accumulatedConsumptionProfileHead = Nothing + | null inputData = Just $ mkNullMatchFromConcatenation inputDataOffset concatenation' -- Build a 'Match', of null 'InputData'. + | and [ + ExecutionOptions.bypassInputDataForLiberalConsumer executionOptions, -- We may be able to establish success, without evaluating 'inputData' any further than required to determine its length. + not $ ExecutionOptions.requireMatchList executionOptions, -- Otherwise the precise mapping of the 'inputData' to 'RepeatablePattern's must be determined. + not $ ConsumptionProfile.hasSpecificRequirement accumulatedConsumptionProfileHead -- Otherwise a match for the specific 'Meta'-data must be found. + ] = Just undefined {-shouldn't be evaluated according to 'ExecutionOptions.requireMatchList'-} + | and [ + ExecutionOptions.checkForUnconsumableData executionOptions, + not $ ConsumptionProfile.canConsumeAnything accumulatedConsumptionProfileHead, -- Otherwise the subsequent test will always fail. + (`isUnconsumableByAnyOf` distinctMetaDataHead) `any` distinctInputDataHead -- Occasionally failure is both inevitable & obvious. + ] = Nothing + | otherwise = {-#SCC "findMatchSlave" #-} let + tailConsumptionProfile :: ConsumptionProfile.ConsumptionProfile + tailConsumptionProfile@( + ConsumptionProfile.MkConsumptionProfile { + ConsumptionProfile.consumptionBounds = (minConsumptionConcatenationTail, maybeMaxConsumptionConcatenationTail) + } + ) = head accumulatedConsumptionProfilesTail -- Extract the aggregate consumption-profile, of the tail of the 'Concatenation'. + + maxDataAvailable :: ConsumptionBounds.DataLength + maxDataAvailable = inputDataLength - minConsumptionConcatenationTail -- The maximum data available to match 'repeatablePatternHead'. + in if maxDataAvailable < 0 + then Nothing + else {-inputData is sufficient for concatenationTail-} let + Repeatable.MkRepeatable { + Repeatable.base = base, + Repeatable.repetitionBounds = (fewest, most), + Repeatable.isGreedy = isGreedy + } = repeatablePatternHead -- Completely deconstruct the 'Repeatable' at the head of the 'Concatenation'. +{- + Find the maximum sequence of inputData, preceding the first of the mandatory requirements of 'concatenationTail'. + This enables one to more tightly constrain the maximum number of repetitions of a 'RepeatablePattern', + which consequently reduces the subsequent requirement to back-track. +-} +-- maximumDataBeforePegs :: Maybe (InputData m) + maximumDataBeforePegs = {-#SCC "maximumDataBeforePegs" #-} reverse (extractMinimumRequirementFromConcatenation concatenationTail) `maximumDataAfterPegs` reverse inputData where +{- + Extract those 'RepeatablePattern's which must match at least once. + Singleton 'Alternatives' are flattened by replication, & non-singleton 'Alternatives' removed, allowing a reduction of the list's base-type to just 'Repeatable.Repeatable' 'Meta.Meta'. + The resulting 'BasicRegEx', with /Kleene Star/s weeded-out & 'Alternatives' pruned, can be matched in O(n)-time, to verify minimal requirements. +-} + extractMinimumRequirementFromConcatenation :: Concatenation m -> BasicRegEx m + extractMinimumRequirementFromConcatenation = concatMap ( + \repeatablePattern -> case Repeatable.base repeatablePattern of + Require datum -> return {-to List-monad-} repeatablePattern { Repeatable.base = datum } + CaptureGroup a + | isSingletonAlternatives a -> concat . replicate (Repeatable.getFewest repeatablePattern) . extractMinimumRequirementFromConcatenation {-recurse-} . concatenation . head $ deconstructAlternatives a + | otherwise -> [] + ) . filter ((> 0) . Repeatable.getFewest) + + maximumDataAfterPegs :: Eq m => BasicRegEx m -> InputData m -> Maybe (InputData m) + maximumDataAfterPegs pegList@( + Repeatable.MkRepeatable { + Repeatable.base = metaPeg, + Repeatable.repetitionBounds = (fewestPeg, _) + } : pegListTail + ) inputData' + | null remainingInputData && length candidateMatchedInputData < fewestPeg = Nothing -- Insufficient data to ever match. + | fewestPeg == 1 || (`Meta.isMatch` metaPeg) `all` tail candidateMatchedInputData = pegListTail `maximumDataAfterPegs` remainingInputData -- Success => recurse. + | otherwise = pegList `maximumDataAfterPegs` tail candidateInputData -- Failed candidate => recurse. + where +-- candidateInputData, candidateMatchedInputData, remainingInputData :: InputData a + candidateInputData = dropWhile (not . (`Meta.isMatch` metaPeg)) inputData' + (candidateMatchedInputData, remainingInputData) = fewestPeg `splitAt` candidateInputData + maximumDataAfterPegs [] inputData' = Just inputData' -- All the pegs have been matched. +{- + Each 'repeatablePatternHead' can either be a 'Require' or a 'CaptureGroup'. + The former case needs further reduction & we can proceed to construct a 'Tree.Leaf' from any matching 'inputData'. + We attempt to reduce the latter case to an instance of the former, by expanding all combinations of repetitions of various 'Alternatives', constructing a new 'ExtendedRegEx' from each, & recursing. +-} + in case base of + Require meta + | null maybeMatchList -> Nothing + | otherwise -> head {-most suitable-} maybeMatchList + where +{- + Attempt to narrow the repetition-range, from the potentially unbounded specification, to that which is possible according to, + the 'inputDataLength', & the specified consumption-bounds of the remainder of the /regex/. + Once the repetition-range has been narrowed, iterate from the appropriate end of the range (depending on 'isGreedy'), extract the corresponding length from inputData, & check for a match. +-} +-- maybeMatchList :: [Maybe (MatchList m)] + maybeMatchList = {-#SCC "maybeMatchList" #-} dropWhile Data.Maybe.isNothing . map ( + \(consumedInputData, unconsumedInputData) -> let + consumedInputDataLength :: ConsumptionBounds.DataLength + consumedInputDataLength = length consumedInputData + in ( + Tree.Leaf (repeatablePatternHead, inputDataOffset, consumedInputData) : + ) `fmap` {-apply to Maybe Functor-} findMatchSlave concatenationTail accumulatedConsumptionProfilesTail distinctMetaDataTail unconsumedInputData ( + inputDataLength - consumedInputDataLength + ) ( + consumedInputDataLength `drop` distinctInputData + ) -- Recurse, to check whether the 'unconsumedInputData' tail also matches. + ) $ let + fewestData, mostData, mostData' :: Repeatable.Repetitions + fewestData = {-#SCC "fewestData" #-} case maybeMaxConsumptionConcatenationTail of + Just m -> (inputDataLength - m) {- :: ConsumptionBounds.DataLength-} `max` fewest {- :: Repeatable.Repetitions-} -- CAVEAT: conceptually different types. + _ -> fewest + +-- mostData = length . takeWhile (~= meta) $ (`take` inputData) maxData -- As slow as it is concise. + mostData = {-#SCC "mostData" #-} maxData - measureUnmatchableTail inputData maxData where + maxData :: ConsumptionBounds.DataLength + maxData = case most of + Just cap -> cap {- :: Repeatable.Repetitions-} `min` maxDataAvailable {- :: ConsumptionBounds.DataLength-} -- CAVEAT: conceptually different types. + _ -> maxDataAvailable + +-- measureUnmatchableTail :: InputData m -> ConsumptionBounds.DataLength -> ConsumptionBounds.DataLength + measureUnmatchableTail _ 0 = 0 -- Have matched all the data. + measureUnmatchableTail (a : as) unmatched + | a `Meta.isMatch` meta = measureUnmatchableTail as $ pred unmatched -- Recurse. + | otherwise = unmatched -- Return the unmatchable tail-length. + measureUnmatchableTail [] _ = error "RegExDot.RegEx.findMatch.findMatchSlave.maybeMatchList.mostData.measureUnmatchableTail:\tdata unexpectedly exhausted." +{- + 'most' has been extracted from 'repeatablePatternHead', + & reduced according to the length of 'inputData' which actually matches, to form 'mostData', + but if it still exceeds 'fewestData', the resulting repetition-range may trigger backtracking. + So, attempt to further reduce this ill-defined repetition-range, according to the competing requirements of 'concatenationTail'. +-} + mostData' + | and [ + ExecutionOptions.moderateGreed executionOptions, + isGreedy, -- Otherwise, since the search proceeds from 'fewest' to 'most', the optimal solution is located before backtracking. + mostData > fewestData -- Otherwise, there's no unbridled greed to moderate. + ] = case maximumDataBeforePegs of + Just maximumDataBeforePegs' -> mostData `min` length maximumDataBeforePegs' -- Cap upper bound. + _ -> negate 1 -- Guaranteed to be < 'fewestData'. + | otherwise {-no requirement for this optimisation-} = mostData + in ( + succ {-fence-post-} mostData' - fewestData -- CAVEAT: possibly <= 0. + ) `take` ( + if isGreedy + then ToolShed.Data.List.Splits.splitsLeftFrom mostData' + else ToolShed.Data.List.Splits.splitsRightFrom fewestData + ) inputData + + CaptureGroup alternatives + | and [ + not isGreedy, + fewest <= 0, -- Zero repetitions permissible. + Data.Maybe.isJust tailMatch + ] -> zeroRepetitions -- Zero repetitions is the optimal solution, rendering the choice of Alternative irrelevant. + | and [ + ExecutionOptions.catchIncompatibleAnchors executionOptions, + fewest > 1, -- Multiple repetitions. + minConsumptionAlternatives > 0, -- InputData required. + ($ extendedRegExAlternatives) `any` [(hasBowAnchor `all`), (hasSternAnchor `all`)] + ] -> Nothing -- 'extendedRegExFromAlternative' requires 'InputData', therefore 'Anchor.Bow' can only pass on the 1st repetition & 'Anchor.Stern' can only pass on the last. + | otherwise -> let +{- + We're about to try all repetitions of one Alternative before progressing to the next. + One could alternatively transpose this search-pattern, & investigate all 'Alternatives' permissible at n repetitions, before trying (n + 1); I've not pursued this option. + Either way, if the user requires data-capture, then for POSIX-compliance we must perform an exhaustive search of the O(Alternatives ^ Repetitions) permutations for the optimal solution; though the search may be narrowed as we proceed. +-} +-- matchPairList :: [(Match m, MatchList m)] + matchPairList = {-#SCC "matchPairList" #-} Data.Maybe.catMaybes . concat . Control.Parallel.Strategies.parMap Control.Parallel.Strategies.rseq {-Particularly effective when 'not ExecutionOptions.requireMatchList' -} ( +-- :: ExtendedRegEx m -> [Maybe (Match m, MatchList m)] + \extendedRegExFromAlternative@MkExtendedRegEx { + concatenation = concatenationFromAlternative + } -> let + components :: Int + components = length concatenationFromAlternative + + consumptionProfileConcatenationFromAlternative :: [ConsumptionProfile.ConsumptionProfile] + consumptionProfileConcatenationFromAlternative = map Consumer.consumptionProfile concatenationFromAlternative + + minConsumptionConcatenationFromAlternative :: ConsumptionBounds.DataLength + maybeMaxConsumptionConcatenationFromAlternative :: Maybe ConsumptionBounds.DataLength + (minConsumptionConcatenationFromAlternative, maybeMaxConsumptionConcatenationFromAlternative) = ConsumptionProfile.consumptionBounds $ ConsumptionProfile.aggregateFromConcatenation consumptionProfileConcatenationFromAlternative + in ( + if ExecutionOptions.abortTrialRepetitionsOnZeroConsumption executionOptions && Data.Maybe.isNothing most {-constrained only by 'inputDataLength'-} + then {-#SCC "abortTrialRepetitionsOnZeroConsumption" #-} ToolShed.Data.List.takeUntil $ \maybeMatchPair -> case maybeMatchPair of + Just (match, _) -> ($ Tree.pop match) `all` [ + (>= fewest) . length, -- Must achieve 'Repeatable.getFewest', regardless of data-consumption. + null . extractDataFromMatchList . last -- If the n-th repetition consumed nothing, so will the (n + 1)-th. + ] + _ -> False -- Failure with n repetitions, doesn't preclude success with n + 1. + else {-optimisation not required-} id + ) . ( +-- :: Maybe (Match m, MatchList m) -> Maybe (Match m, MatchList m) + \maybeMatchPairList -> if ExecutionOptions.abortTrialRepetitionsOnInherentFailure executionOptions +{- + We're iterating through the identified range of possible repetitions, of one possible 'concatenationFromAlternative'. + If failure can be proven, for an arbitrary number of repetitions, then we can bypass the futile remainder. + If the first trial succeeded, then clearly there's no such inherent problem with 'concatenationFromAlternative', & it's necessary to evaluate all subsequent trial-repetitions. + Otherwise, assuming there're subsequent trials to be bypassed, attempt to prove that failure is inevitable, by launching a pilot-study, + using a single repetition of 'concatenationFromAlternative' followed by a replacement infallible 'concatenationTail'. +-} + then {-#SCC "abortTrialRepetitionsOnInherentFailure" #-} case maybeMatchPairList of + (Nothing {-failed attempt-} : _ {-subsequent attempt worth bypassing-} : _) -> if ( + minConsumptionConcatenationFromAlternative == 0 -- Can delegate consumption of unconsumable data to the infallible tail => unprovable culpability. + ) || inputData =~ RegExOpts.mkRegEx MkExtendedRegEx { + bowAnchor = Just Anchor.Bow, + concatenation = concatenationFromAlternative ++ [anyDatum ^#-> minConsumptionConcatenationTail], -- Infallible tail. + sternAnchor = Just Anchor.Stern + } + then maybeMatchPairList + else [] -- Abandon further repetitions of 'Alternatives'. + _ {-Alternative is OK, or there's no subsequent attempt to bypass-} -> maybeMatchPairList + else {-optimisation not required-} maybeMatchPairList + ) . map ( +-- :: Repeatable.Repetitions -> Maybe (Match m, MatchList m) + \repetitions -> if hasSternAnchor extendedRegExFromAlternative + then if (repetitions == 1 || minConsumptionConcatenationFromAlternative == 0) && minConsumptionConcatenationTail == 0 + then ( + shiftMatch inputDataOffset . Tree.Node . ( + : replicate ( + pred repetitions -- Typically degenerates to zero, since anchored sub-expressions aren't normally repeatable. + ) ( + mkNullMatchFromExtendedRegEx 0 extendedRegExFromAlternative -- Other Alternatives may be available & suitable. + ) -- Repeated null match may, lacking other suitable Alternatives, correspond to this stern-anchored ExtendedRegEx, so data must be consumed before them. + ) &&& const (mkNullMatchFromConcatenation inputDataOffset concatenationTail) + ) `fmap` findMatch regExOpts { RegExOpts.regEx = extendedRegExFromAlternative } inputData -- Recurse, either consuming all inputData or failing. + else Nothing + else {-no Anchor.Stern-} let +{- + Expand the 'Repeatable CaptureGroup', so that after recursion it is ultimately reduced in complexity, to just a 'Concatenation' of 'Repeatable Require', & can be handled above as a POSIX BRE. + + There're two special cases; one repetition of n 'Alternatives' & n repetitions of one Alternative. + If there's only one repetition, we can reduce the implementation-logic substantially, to the point where the implementation's independent of whether there's a choice of 'Alternatives'. + If there only one Alternative, then the entire repetition-sequence can be immediately expanded, vastly reducing the required number of recursions, + otherwise we can only expand the first repetition, leaving the option of matching a different Alternative on subsequent repetitions. + + In all cases, we must subsequently re-compose any 'MatchList' resulting of the recursive call, into one corresponding to the structure of original ExtendedRegEx. + The expansion-process & corresponding subsequent collation, are associated in a tuple, in which the consumption-profile of the expansion is also contained. +-} +-- collater :: MatchList m -> (Match m, MatchList m) + consumptionProfileExpandedConcatenationPrefix :: [ConsumptionProfile.ConsumptionProfile] +-- expandedConcatenationPrefix :: Concatenation a + (collater, consumptionProfileExpandedConcatenationPrefix, expandedConcatenationPrefix) + | repetitions == 1 = ( + Control.Arrow.first ( + Tree.Node . return {-to List-monad-} + ) . splitAt components, -- Bisect the 'MatchList', into the part resulting from the Alternative, & that for 'concatenationTail'. + consumptionProfileConcatenationFromAlternative, + concatenationFromAlternative + ) -- Both tuples below, can degenerate to this simple case; it isn't a fundamentally different algorithm. + | and [ + ExecutionOptions.unrollRepeatedSingletonAlternative executionOptions, + isSingletonAlternative, + not $ hasBowAnchor extendedRegExFromAlternative + ] = ( + Control.Arrow.first ( + Tree.Node . ToolShed.Data.List.chunk components + ) . splitAt ( + components * repetitions + ), -- Bisect the 'MatchList', into the part resulting from (Alternative){n}, & that for 'concatenationTail'. + concat $ replicate repetitions consumptionProfileConcatenationFromAlternative, + concat $ replicate repetitions concatenationFromAlternative -- Expand all repetitions of the single alternative. + ) + | otherwise {-choice of Alternatives or has Anchor.Bow-} = let + remainingRepetitions :: Repeatable.Repetitions + remainingRepetitions = pred repetitions -- Expand just the first repetition of the set of 'Alternatives'. + +-- singletonRepeatable :: RepeatablePattern a + singletonRepeatable = Repeatable.toSingleton repeatablePatternHead + in ( + Control.Arrow.first ( + Tree.Node . uncurry (:) . Control.Arrow.second ( + concatMap Tree.pop -- Amalgamate the repeated singleton 'MatchList's. + ) . splitAt components -- Bisect first 'MatchList', into the part resulting from 'concatenationFromAlternative', & that for 'replicate remainingRepetitions'. + ) . splitAt ( + components + remainingRepetitions -- Bisect the 'MatchList', into the part resulting from (Alternatives){n}, & that for 'concatenationTail'. + ), + consumptionProfileConcatenationFromAlternative ++ replicate remainingRepetitions (Consumer.consumptionProfile singletonRepeatable), + concatenationFromAlternative {-expand 1st repetition-} ++ replicate remainingRepetitions {-potentially zero-} singletonRepeatable -- Enumerate all remaining repetitions of ANY Alternative; a different Alternative can match for each repetition. + ) + in {-#SCC "collater" #-} collater `fmap` {-apply to Maybe Functor-} findMatchSlave ( + expandedConcatenationPrefix ++ concatenationTail + ) ( + ConsumptionProfile.accumulateFrom tailConsumptionProfile {-initial value-} consumptionProfileExpandedConcatenationPrefix ++ tail accumulatedConsumptionProfilesTail + ) ( + accumulateDistinctMetaDataFrom (head distinctMetaDataTail) {-initial value-} expandedConcatenationPrefix ++ tail distinctMetaDataTail + ) inputData inputDataLength distinctInputData -- Recurse to get 'Maybe (MatchList a)'. + ) $ if Repeatable.isPrecise repeatablePatternHead -- These are frequently generated by the previous recursion, as a trial expansion of an Alternative. + then if ExecutionOptions.validateMinConsumptionOfAlternatives executionOptions && maxDataAvailable < fewest * minConsumptionAlternatives + then [] -- Failure is inevitable. + else [fewest] + else {-imprecise range-} let +{- + Attempt to narrow the repetition-range, from the potentially unbounded specification, to that which is possible according to (data available / consumption-requirement), + where there're bounds for both numerator & denominator. + Additionally, one must account for the possibility of the denominator being zero, & if so, select the optimal value if the numerator is also zero. + The range can be narrowed more aggressively, for the special case of a singleton 'Alternative', as used in a capture-group. +-} + fewestAlternatives, mostAlternatives, mostAlternatives' :: Repeatable.Repetitions + fewestAlternatives = {-#SCC "fewestAlternatives" #-} 1 {-zero is handled later-} `max` if any Data.Maybe.isNothing [maybeMaxConsumptionConcatenationTail, maybeMaxConsumptionAlternatives] + then fewest + else {-no unlimited capacities-} let + minDataAvailable, maxConsumptionAlternatives :: ConsumptionBounds.DataLength + minDataAvailable = inputDataLength - Data.Maybe.fromJust maybeMaxConsumptionConcatenationTail -- CAVEAT: can be negative. + maxConsumptionAlternatives = Data.Maybe.fromJust maybeMaxConsumptionAlternatives + in if isSingletonAlternative + then if maxConsumptionAlternatives == 0 -- Denominator. + then {-sole Alternative can't consume anything ?!-} if minDataAvailable <= 0 -- Numerator. + then {-tail may consume all data-} fewest -- Repeat to meet the minimum requirement. + else {-unconsumable data-} maxBound -- Failure is inevitable. + else {-non-zero => can divide-} max fewest $ minDataAvailable /+ maxConsumptionAlternatives + else {-choice of Alternatives-} let + minDataAvailable' :: ConsumptionBounds.DataLength + minDataAvailable' = minDataAvailable - Data.Maybe.fromJust maybeMaxConsumptionConcatenationFromAlternative -- CAVEAT: potentially negative. + in if maxConsumptionAlternatives == 0 -- Denominator. + then {-no Alternative can consume anything-} if minDataAvailable' <= 0 -- Numerator. + then {-zero unconsumable data-} fewest -- Repeat to meet the minimum requirement. + else {-unconsumable data-} maxBound -- Failure is inevitable. + else {-non-zero => can divide-} max fewest . succ {-account for expanded instance-} $ minDataAvailable' /+ maxConsumptionAlternatives + + mostAlternatives = {-#SCC "mostAlternatives" #-} case most of + Just cap -> cap `min` mostPermissibleRepetitions -- Create a ceiling above which the calculated number of permissible repetitions can't rise. + _ -> mostPermissibleRepetitions + where + mostPermissibleRepetitions :: Repeatable.Repetitions + mostPermissibleRepetitions + | isSingletonAlternative = if minConsumptionAlternatives == 0 -- Denominator. + then {-sole Alternative can consume zero data-} if maxDataAvailable == 0 -- Numerator. + then {-zero data available-} fewest -- Repeat to meet the minimum requirement. + else {-data available-} maxDataAvailable -- Either each repetition consumes something, or (n - 1) repetitions is a better solution. + else {-non-zero => can divide-} maxDataAvailable `div` minConsumptionAlternatives -- Divide & round down. + | otherwise {-choice of Alternatives-} = 1 {-account for instance expanded as 'concatenationFromAlternative'-} + let + maxDataAvailable' :: ConsumptionBounds.DataLength + maxDataAvailable' = maxDataAvailable - minConsumptionConcatenationFromAlternative -- CAVEAT: potentially negative. + in if minConsumptionAlternatives == 0 -- Denominator. + then {-@ least one Alternative can consume zero-} if maxDataAvailable' == 0 -- Numerator. + then {-zero data available $-} pred fewestAlternatives -- Repeat as required. CAVEAT: an annoying dependency, which prevents parallel-evaluation. + else {-data available-} if and [ + False, -- The cost outweighs the small infrequent dividend. + minConsumptionConcatenationFromAlternative == 0, + maxDataAvailable > fewest + ] -- If 'fewest' can be met, without unconsuming repetitions of 'concatenationFromAlternative', then stop short of permitting it. + then pred maxDataAvailable -- Any greater, & one repetition (possibly that currently expanded) must needlessly consume nothing. + else maxDataAvailable' -- Either all n repetitions consumes something, or (n - 1) is a better solution. + else {-non-zero => can divide-} maxDataAvailable' `div` minConsumptionAlternatives -- Divide & round down. +{- + 'most' has been extracted from 'repeatablePatternHead', + & reduced according to the number of times it can fit into the maximum available data, to form 'mostAlternatives', + but if it still exceeds 'fewestAlternatives', the resulting repetition-range may trigger backtracking. + So, attempt to further reduce this ill-defined repetition-range, according to the competing requirements of 'concatenationTail'. + CAVEAT: this pays-off infrequently, but not necessarily insignificantly. +-} + mostAlternatives' = {-#SCC "mostAlternatives'" #-} if and [ + ExecutionOptions.moderateGreed executionOptions, + mostAlternatives > fewestAlternatives, -- Otherwise, there's no unbridled greed to moderate. + minConsumptionAlternatives > 0 -- Otherwise, any number of repetitions can occur without necessarily triggering back-tracking. + ] + then case maximumDataBeforePegs of + Just maximumDataBeforePegs' -> mostAlternatives `min` ( + length maximumDataBeforePegs' `div` minConsumptionAlternatives + ) -- Cap upper bound. + _ -> negate 1 -- Guaranteed to be < 'fewestAlternatives'. + else {-no requirement for this optimisation-} mostAlternatives + in [fewestAlternatives .. mostAlternatives'] -- The permissible repetition-range of any Alternative, constrained by the amount of data available, after subtracting fixed consumption-requirements. + ) . ( + if ExecutionOptions.permitReorderingOfAlternatives executionOptions +{- + The order in which 'Alternatives' are evaluated, should make no difference to the final result; unless 'ExecutionOptions.useFirstMatchAmongAlternatives'. + This order can be manipulated in an attempt to more quickly locate any 'Just MatchList', though if 'ExecutionOptions.requireMatchList', an exhaustive search for the optimal solution must be performed anyway. + Tackling lower-complexity 'Alternatives' first makes sense, in the hope of finding any easy solution before becoming entrenched fighting some hideous monster. + Preferring 'Alternatives' which require more data also makes sense, regardless of 'Repeatable.isGreedy' which just a preference for a type of solution, because fewer repetitions are then permissible; + this is rather like trying to pack a box & preferring to try the biggest items first. +-} + then Data.List.sortBy $ Data.Ord.comparing ( + Consumer.starHeight &&& safeReciprocal . (fromIntegral :: ConsumptionBounds.DataLength -> Rational) . Consumer.getFewest + ) -- Firstly increasing complexity, then decreasing minimum data-capacity, otherwise stable. + else id + ) $ filter ( + \e -> ( + not (hasBowAnchor e) || inputDataOffset == 0 -- Necessary & sufficient. + ) && ( + not (hasSternAnchor e) || minConsumptionConcatenationTail == 0 -- Necessary but insufficient, since though minimum consumption is zero, maximum isn't necessarily. + ) + ) extendedRegExAlternatives + in if null matchPairList + then if fewest <= 0 + then zeroRepetitions -- Which might still fail, depending on 'tailMatch'. + else {-zero repetitions isn't permissible-} Nothing + else {-at least one Alternative matched-} {-#SCC "selectAlternative" #-} Just . uncurry (:) {-re-join head & tail-} $ ( +{- + At least one 'Match' has been found amongst the combinations of repetitions of various 'Alternatives', + & if the caller reduces the 'Just MatchList' to 'True', that'd be sufficient, but otherwise one must identify the optimal 'MatchList'. + The primary selection-criterion used is the POSIX one, based on total maximum (or minimum for non-'Repeatable.isGreedy' capture-groups) data-consumption. + /Perl/'s "first-past-the-post" policy is available as an alternative. +-} + if ExecutionOptions.useFirstMatchAmongAlternatives executionOptions || length matchPairList == 1 + then head + else snd {-remove prepended selection-criterion-} . Data.List.maximumBy {-select the best match-} ( + Data.Ord.comparing fst -- Compare using only the criterion, not the result from which it was derived. + ) . map ( + ( +{- + If the primary selection-criterion doesn't resolve the choice between candidate 'Match'es, I employ these ad-hoc criteria. + + * Fewer repetitions are preferred, which discourages the capture of null lists of 'InputData'. + + * Within the 'MatchedData', from which a candidate 'Match' is ultimately composed, consumption of 'InputData' beyond 'Repeatable.getFewest' by 'Repeatable.isGreedy' 'RepeatablePattern's, is preferred to non-greedy ones. + + * The consumption on each successive repetition is compared between candidate 'Match'es (which are now known to have used an equal number of repetitions); + this causes the data-consumption to flow towards earlier repetitions for greedy capture-groups, & towards later repetitions in non-greedy ones, + thus mimicking the behaviour of the unrolled repetition. + + * PS: more precise criteria are defined in . +-} + \match -> let + dataLengthCriterion :: InputData m -> Rational + dataLengthCriterion = ( + if isGreedy + then id + else safeReciprocal -- Prefer less data. + ) . fromIntegral . length + +-- matchLists :: [MatchList m] + matchLists = Tree.pop match + in ( + dataLengthCriterion $ extractDataFromMatch match, + if ExecutionOptions.preferFewerRepeatedAlternatives executionOptions + then safeReciprocal . toRational $ length matchLists + else 0, + if ExecutionOptions.preferAlternativesWhichFeedTheGreedy executionOptions + then let + excessConsumption :: MatchedData m -> ConsumptionBounds.DataLength + excessConsumption ( + Repeatable.MkRepeatable { + Repeatable.repetitionBounds = (f, _), + Repeatable.isGreedy = g + }, + _, + consumedInputData + ) + | g = l - f + | otherwise = f - l + where + l :: ConsumptionBounds.DataLength + l = length consumedInputData + in Data.Foldable.foldl' (\l -> (l +) . excessConsumption) 0 match + else 0, + if ExecutionOptions.preferAlternativesWhichMimickUnrolling executionOptions + then map (dataLengthCriterion . extractDataFromMatchList) matchLists + else [] + ) -- Create a tuple of selection-criteria, for simultaneous assessment. + ) . fst {-focus the choice on the Alternatives-} &&& id -- Prepend a selection-criterion to each result. + ) + ) matchPairList + where +-- tailMatch, zeroRepetitions :: Maybe (MatchList m) + tailMatch = findMatchSlave concatenationTail accumulatedConsumptionProfilesTail distinctMetaDataTail inputData inputDataLength distinctInputData -- Recurse. + zeroRepetitions = (Tree.Node [] :) `fmap` {-apply to Maybe Functor-} tailMatch -- Prepend a null 'MatchList'. + +-- extendedRegExAlternatives :: [ExtendedRegEx m] + extendedRegExAlternatives = Data.List.nub $ deconstructAlternatives alternatives + + isSingletonAlternative :: Bool + isSingletonAlternative = length extendedRegExAlternatives == 1 + + minConsumptionAlternatives :: ConsumptionBounds.DataLength + maybeMaxConsumptionAlternatives :: Maybe ConsumptionBounds.DataLength + (minConsumptionAlternatives, maybeMaxConsumptionAlternatives) = Consumer.getConsumptionBounds alternatives -- Independent of the choice of Alternative. + where + inputDataOffset :: ConsumptionBounds.DataLength + inputDataOffset = originalInputDataLength - inputDataLength + findMatchSlave _ _ _ _ _ _ = error "RegExDot.RegEx.findMatch.findMatchSlave:\tunexpected parameter-pattern." + +{- | + * Operator's name was chosen to suggest something more than '=~'. + + * CAVEAT: much more expensive then '=~': in /ghci/, 'Just' can be observed to be printed /long/ before the 'MatchList' from which 'Result' is constructed, + as the lazy algorithm finds the first solution, but not yet necessarily the optimal solution, amongst 'Alternatives'. +-} +(+~) :: (Eq m, Control.DeepSeq.NFData m) + => InputData m -- ^ The input data within which to locate a match. + -> RegExOpts.RegExOpts (ExtendedRegEx m) -- ^ The match-options parameterised by the regex against which to match the input data. + -> Result m +inputData +~ regExOpts = ( + if hasBowAnchor' then Nothing else fmap head maybeMatchList, -- Record the first 'Match', consumed in the absence of a top-level 'Anchor.Bow'. + fmap ( + ( + if hasBowAnchor' then id else tail -- Remove the first 'Match', since this wasn't consumed by the 'Concatenation'. + ) . ( + if hasSternAnchor' then id else init -- Remove the last 'Match', since this wasn't consumed by the 'Concatenation'. + ) + ) maybeMatchList, + if hasSternAnchor' then Nothing else fmap last maybeMatchList -- Record the last 'Match', consumed in the absence of a top-level 'Anchor.Stern'. + ) where +-- extendedRegEx :: ExtendedRegEx a + extendedRegEx = RegExOpts.regEx regExOpts + + hasBowAnchor', hasSternAnchor' :: Bool + hasBowAnchor' = hasBowAnchor extendedRegEx + hasSternAnchor' = hasSternAnchor extendedRegEx + +-- maybeMatchList :: Maybe (MatchList m) + maybeMatchList = fmap drift (RegExOpts.setVerbose True regExOpts) `findMatch` inputData + +{- | + * Pattern-match operator. + + * Identifier & parameter-order follow the lead of /Perl/'s pattern-match operator. + + * Considerably more efficient than '+~', since even though they are both implemented via 'findMatch', + the discovery of /any/ solution is sufficient to generate the return-value; + lazy-evaluation avoids the requirement to identify the irrelevant optimal solution. +-} +(=~) :: (Eq m, Control.DeepSeq.NFData m) + => InputData m -- ^ The input data within which to locate a match. + -> RegExOpts.RegExOpts (ExtendedRegEx m) -- ^ The match-options parameterised by the regex against which to match the input data. + -> Bool +inputData =~ regExOpts = Data.Maybe.isJust $ fmap drift (RegExOpts.setVerbose False regExOpts) `findMatch` inputData + +-- | Pattern-mismatch operator. +(/~) :: (Eq m, Control.DeepSeq.NFData m) + => InputData m -- ^ The input data within which to locate a match. + -> RegExOpts.RegExOpts (ExtendedRegEx m) -- ^ The match-options parameterised by the regex against which to match the input data. + -> Bool +(/~) inputData = not . (inputData =~) + +-- | The delimiters of 'Alternatives', when in the 'String'-form. +captureGroupDelimiters :: (Char, Char) +captureGroupDelimiters = ('(', ')') + +-- | The token used to separate alternative 'ExtendedRegEx's, when in the 'String'-form. +alternativeExtendedRegExSeparatorToken :: Char +alternativeExtendedRegExSeparatorToken = '|' + +-- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. +tokens :: String +tokens = [fst captureGroupDelimiters, snd captureGroupDelimiters, alternativeExtendedRegExSeparatorToken] + +{- | + * Invert the specified fraction, but return 'maxBound' if the result would otherwise be infinite. + + * CAVEAT: this is an awful concept, and therefore intended for internal use only. +-} +safeReciprocal :: (Eq f, Fractional f) => f -> f +safeReciprocal 0 = fromIntegral (maxBound :: Int) -- Handle divide-by-zero error. +safeReciprocal f = recip f + +infixl 7 /+ -- Same as (/). + +{- | + * Integral division, with any fractional remainder rounded-up. + + * A rather dubious requirement, so internal use only. +-} +-- (/+) :: Integral i => i -> i -> i +(/+) + :: Int -- ^ Numerator. + -> Int -- ^ Denominator. + -> Int -- 10% faster in unoptimised code, & more in optimised. +_ /+ 0 = error "RegExDot.RegEx.(/+):\tzero denominator => infinity" +-- numerator /+ denominator = ceiling ((fromIntegral numerator / fromIntegral denominator) :: Double) +numerator /+ denominator = uncurry (+) . Control.Arrow.second signum $ quotRem numerator denominator -- Slightly faster. diff --git a/src-lib/RegExDot/RegExOpts.hs b/src-lib/RegExDot/RegExOpts.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/RegExOpts.hs @@ -0,0 +1,65 @@ +{-# LANGUAGE MultiParamTypeClasses, TypeSynonymInstances, FlexibleInstances #-} +{-# OPTIONS_GHC -fno-warn-orphans #-} + +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * Provides a complete job-specification. + + * 'regEx' is polymorphic, allowing use of either /RegExDot.RegEx.ExtendedRegEx/ (which is also polymorphic) or /RegExChar.ExtendedRegExChar.ExtendedRegExChar/. +-} + +module RegExDot.RegExOpts( +-- * Types +-- ** Data-types + RegExOpts(..), +-- * Functions + setVerbose, +-- ** Constructors + mkRegEx +) where + +import qualified RegExDot.CompilationOptions as CompilationOptions +import qualified RegExDot.ExecutionOptions as ExecutionOptions +import qualified ToolShed.Defaultable + +-- | Aggregates both 'ExecutionOptions.ExecutionOptions' & 'CompilationOptions.CompilationOptions' with a polymorphic /regex/, to form a complete job-description. +data RegExOpts a = MkRegExOpts { + compilationOptions :: CompilationOptions.CompilationOptions, -- ^ Parameters governing the result, rather than the implementation. + executionOptions :: ExecutionOptions.ExecutionOptions, -- ^ Parameters governing the implementation, rather than the result. + regEx :: a -- ^ Polymorphic, to permit specialisation either of the type of /regex/ or the type of input-data. +} deriving Show + +instance Functor RegExOpts where + fmap f regExOpts = regExOpts { regEx = f $ regEx regExOpts } + +-- | Smart constructor using 'ToolShed.Defaultable.defaultValue'. +mkRegEx :: a -> RegExOpts a +mkRegEx r = MkRegExOpts { + compilationOptions = ToolShed.Defaultable.defaultValue, + executionOptions = ToolShed.Defaultable.defaultValue, + regEx = r +} + +-- | Mutator, which sets an appropriate 'ExecutionOptions.ExecutionOptions', for the depth to which the caller wants to probe the resulting match. +setVerbose :: Bool -> RegExOpts a -> RegExOpts a +setVerbose verbose regExOpts = regExOpts { executionOptions = ExecutionOptions.setVerbose verbose $ executionOptions regExOpts } + diff --git a/src-lib/RegExDot/Repeatable.hs b/src-lib/RegExDot/Repeatable.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/Repeatable.hs @@ -0,0 +1,400 @@ +{-# LANGUAGE CPP #-} +{- + Copyright (C) 2010-2015 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * A data-type, which represents the permissible range of instances, of the underlying polymorphic datum. + + * Designed for use in a polymorphic /regex/-engine, which specifies patterns composed of repeated /greedy/ & /non-greedy/ sequences of /Meta/-data; + + > * + ? {fewest, most} {fewest,} {fewest} + > *? +? ?? {fewest, most}? {fewest,}? + + * In the context of /regex/es, this concept is known as /Quantification/. + + * /regex/es evolved from the minimal ability to optionally qualify the datum with a suffix. + More exotic repetition-specifications could be composed by concatenating these atomic building-blocks. + Here, I've taken the contrary top-down view, & assumed that all data are qualified by a full 'RepetitionBounds', which in most cases will degenerate into a simpler form. + + * The type of entity which is being repeated, isn't the domain of this data-type; it's polymorphic. +-} + +module RegExDot.Repeatable( +-- * Types +-- ** Type-synonyms + Repetitions, + RepetitionBounds, +-- ** Data-types + Repeatable(..), +-- * Constants +-- nonGreedyToken, + oneOrMoreToken, + rangeDelimiters, + rangeSeparatorToken, + tokens, + zeroOrMoreToken, + zeroOrOneToken, +-- * Functions + one, + oneOrMore, + oneOrMore', + zeroOrMore, + zeroOrMore', + zeroOrOne, + zeroOrOne', +-- precisely, + repeatableParser, + showSuffix, +-- ** Accessors + getFewest, + getMost, +-- ** Mutators + focus, +-- setNonGreedy, + toSingleton, +-- ** Operators + (^#->#), + (^#->#?), + (^#->), + (^#->?), + (^#), +-- ** Predicates + isPrecise, + hasPreciseBounds +) where + +import Control.Arrow((***)) +import qualified Control.DeepSeq +import qualified Data.List +import qualified RegExDot.Consumer as Consumer +import qualified RegExDot.ConsumptionProfile as ConsumptionProfile +import qualified Text.ParserCombinators.Parsec as Parsec +import Text.ParserCombinators.Parsec(()) +import qualified ToolShed.Data.Pair +import qualified ToolShed.SelfValidate + +#if !MIN_VERSION_base(4,8,0) +import Control.Applicative((<$>), (<*>)) +#endif + +infix 6 ^#->#, ^#->#?, ^#->, ^#->?, ^# -- A notch tighter than "DSL"s binary operators. + +-- | A number of repetitions. +type Repetitions = Int + +-- | Defines the bounds of a range of permissible repetitions. +type RepetitionBounds = (Repetitions, Maybe Repetitions) + +-- | Creates a precise 'RepetitionBounds', i.e. both lower & upper bounds on the number of 'Repetitions' are equal to the same value. +precisely :: Repetitions -> RepetitionBounds +precisely i = (i, Just i) + +-- | Predicate which is 'True' if exactly one value is permissible, ie lower & upper bounds on the number of 'Repetitions' are identical. +hasPreciseBounds :: RepetitionBounds -> Bool +hasPreciseBounds (fewest, most) = Just fewest == most + +-- | Declares a polymorphic data-type, which augments the underlying 'base' datum, with the range of times it may be used. +data Repeatable a = MkRepeatable { + base :: a, -- ^ The underlying polymorphic datum. + repetitionBounds :: RepetitionBounds, -- ^ The bounds delimiting the range of permissible repetitions, of 'base'. + isGreedy :: Bool -- ^ Whether to demand as many matching instances of 'base' as possible; or as few (AKA /lazy quantification/). +} deriving Eq + +instance Functor Repeatable where + fmap f repeatable = repeatable { base = f $ base repeatable } + +-- | True if there's no choice in the number of repetitions; implemented via 'isPrecise'. +isPrecise :: Repeatable a -> Bool +isPrecise = hasPreciseBounds . repetitionBounds + +-- | Builds a parser for a specification of the number of permissible instances of the specified polymorphic parameter. +repeatableParser :: a -> Parsec.Parser (Repeatable a) +repeatableParser b = Parsec.option ( + one b -- The default; there's no concept of greediness here. + ) $ do + repeatable <- Parsec.choice [ + (Parsec.char oneOrMoreToken "Repeatable.oneOrMoreToken " ++ show oneOrMoreToken) >> return {-to ParsecT-monad-} (oneOrMore b), + (Parsec.char zeroOrOneToken "Repeatable.zeroOrOneToken " ++ show zeroOrOneToken) >> return {-to ParsecT-monad-} (zeroOrOne b), + (Parsec.char zeroOrMoreToken "Repeatable.zeroOrMoreToken " ++ show zeroOrMoreToken) >> return {-to ParsecT-monad-} (zeroOrMore b), + (b ^#->#) {-arbitrarily greedy for now-} <$> uncurry Parsec.between (ToolShed.Data.Pair.mirror Parsec.char rangeDelimiters) ( + do + fewest <- Parsec.spaces >> (read <$> Parsec.many1 Parsec.digit "Repetition-range minimum") + most <- Parsec.spaces >> Parsec.option ( + Just fewest -- The default. + ) ( + do + i <- ( + Parsec.char rangeSeparatorToken "Repeatable.rangeSeparatorToken " ++ show rangeSeparatorToken + ) >> Parsec.spaces >> Parsec.option Nothing {-default to open-ended range-} ( + Just . read <$> Parsec.many1 Parsec.digit "Repetition-range maximum" + ) + + Parsec.spaces >> return {-to ParsecT-monad-} i + ) + + return {-to ParsecT-monad-} (fewest, most) + ) "Repeatable.rangeDelimiters " ++ show rangeDelimiters + ] + + g <- Parsec.option True {-the default-} $ (Parsec.char nonGreedyToken "Repeatable.nonGreedyToken " ++ show nonGreedyToken) >> return {-to ParsecT-monad-} False + + return {-to ParsecT-monad-} repeatable { isGreedy = g } -- Correct prior assumption. + +instance Read a => Read (Repeatable a) where + readsPrec _ s = case reads s {-first, read the base-type-} of + [(base', s1)] -> (error . ("readsPrec Repeatable:\tparse-error; " ++) . show) `either` return $ Parsec.parse ((,) <$> repeatableParser base' <*> Parsec.getInput) "Repeatable" s1 + _ -> [] -- No parse. + +{- | + * A 'ShowS'-function for the suffix, denoting the permissible repetitions, of 'base'. + + * This function converts the internal, into the tradition /greedy/ & /non-greedy/ quantifiers of various specific varieties. +-} +showSuffix :: Repeatable a -> ShowS +showSuffix repeatable = let + showRange :: ShowS -> ShowS + showRange x = (\(begin, end) -> begin . x . end) $ ToolShed.Data.Pair.mirror showChar rangeDelimiters + in ( + case repetitionBounds repeatable of + (0, Nothing) -> showChar zeroOrMoreToken + (1, Nothing) -> showChar oneOrMoreToken + (fewest, Nothing) -> showRange $ shows fewest . showChar rangeSeparatorToken + (0, Just 1) -> showChar zeroOrOneToken + (1, Just 1) -> id -- CAVEAT: since there's no explicit repetition-operator, the non-greedy modifier can't be appended. + (fewest, Just most) -> showRange $ if fewest == most + then shows fewest -- Single-valued range. + else shows fewest . showChar rangeSeparatorToken . shows most + ) . if ($ repeatable) `any` [isGreedy, isPrecise] {-without a range of possibilities, non-greediness is irrelevant-} + then id + else showChar nonGreedyToken -- This can only be appended, if there a previous repetition-operator for it to modify. + +-- Replicate the syntax, for repetition, as used in a POSIX-standard /regex/. +instance Show a => Show (Repeatable a) where + showsPrec _ repeatable = shows (base repeatable) . showSuffix repeatable + +instance Consumer.Consumer a => Consumer.Consumer (Repeatable a) where + consumptionProfile MkRepeatable { + base = b, + repetitionBounds = (fewest, most) + } = baseConsumptionProfile { + ConsumptionProfile.consumptionBounds = (fewest *) *** ((*) <$> most <*>) $ ConsumptionProfile.consumptionBounds baseConsumptionProfile -- CAVEAT: special cases exist, where one or both halves of this calculation degenerate to a simpler form, but special treatment, in an attempt to improve performance, proved counterproductive. + } where + baseConsumptionProfile :: ConsumptionProfile.ConsumptionProfile + baseConsumptionProfile = Consumer.consumptionProfile b + + starHeight MkRepeatable { + base = b, + repetitionBounds = r + } = Consumer.starHeight b + if hasPreciseBounds r then 0 else 1 + +instance ToolShed.SelfValidate.SelfValidator a => ToolShed.SelfValidate.SelfValidator (Repeatable a) where + getErrors MkRepeatable { + base = b, + repetitionBounds = (fewest, most), + isGreedy = g + } + | not $ ToolShed.SelfValidate.isValid b = ToolShed.SelfValidate.getErrors b -- Delegate. + | otherwise = ToolShed.SelfValidate.extractErrors [ + (fewest < 0, "Negative fewest=" ++ show fewest ++ "."), + ( + case most of + Just m -> m < fewest + _ -> False, + "Invalid repetition-range; '" ++ show (fewest, most) ++ "'." + ), ( + not g && case most of + Just m -> fewest >= m -- There ought to be potential for non-greediness, where specified: the converse isn't true, since greediness isn't explicit, & may not have been wanted. + _ -> False, + "Invalid non-greedy repetition-range; '" ++ show (fewest, most) ++ "'." + ) + ] + +instance Control.DeepSeq.NFData a => Control.DeepSeq.NFData (Repeatable a) where + rnf MkRepeatable { + base = b, + repetitionBounds = r, + isGreedy = g + } = Control.DeepSeq.rnf (b, r, g) + +-- | Mutator. +setNonGreedy :: Repeatable a -> Repeatable a +setNonGreedy r = r { isGreedy = False } + +{- | + * Construct a greedy 'Repeatable', from a polymorphic datum, with the specified range of permissible instances. + + * The /#/s in the identifier represent the two bounds. + + * /a{f, m}/ +-} +(^#->#) + :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. + -> RepetitionBounds -- ^ The permissible repetition-bounds for the polymorphic data. + -> Repeatable a +b ^#-># bounds = MkRepeatable { + base = b, + repetitionBounds = bounds, + isGreedy = True +} + +{- | + * Construct a non-greedy version of '^#->#'. + + * /a{f, m}?/ +-} +(^#->#?) + :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. + -> RepetitionBounds -- ^ The permissible repetition-bounds for the polymorphic data. + -> Repeatable a +b ^#->#? bounds = setNonGreedy (b ^#-># bounds) + +{- | + * Construct a greedy 'Repeatable', tailored for data repeated at least the specified number of times. + + * The /#/ in the identifier represents the single bound. + + * /a{f,}/ +-} +(^#->) + :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. + -> Repetitions -- ^ The minimum permissible repetitions of the polymorphic data. + -> Repeatable a +b ^#-> fewest = b ^#-># (fewest, Nothing) + +{- | + * Construct a non-greedy version of '^#->'. + + * /a{f,}?/ +-} +(^#->?) + :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. + -> Repetitions -- ^ The minimum permissible repetitions of the polymorphic data. + -> Repeatable a +b ^#->? fewest = setNonGreedy (b ^#-> fewest) + +{- | + * Construct a 'Repeatable', tailored for data repeated a precise number of times. + + * The /#/ in the identifier represents the single bound. + + * /a{f}/ +-} +(^#) + :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. + -> Repetitions -- ^ The precise number of repetitions of the polymorphic data which is required. + -> Repeatable a +b ^# r = b ^#-># precisely r + +{- | + * Construct a 'Repeatable', tailored for unrepeated data. + + * A degenerate case of '^#'. +-} +one :: a -> Repeatable a +one = (^# 1) + +{- | + * Construct a greedy 'Repeatable', from a polymorphic datum, with 'fewest' == 0 & 'most' == 1. + + * A specific case of '^#->#'. +-} +zeroOrOne :: a -> Repeatable a +zeroOrOne = (^#-># (0, Just 1)) + +-- | Construct a non-greedy version of 'zeroOrOne'. +zeroOrOne' :: a -> Repeatable a +zeroOrOne' = setNonGreedy . zeroOrOne + +{- | + * Construct a greedy 'Repeatable', from a polymorphic datum, with 'fewest' == 0. + + * A specific case of '^#->'. +-} +zeroOrMore :: a -> Repeatable a +zeroOrMore = (^#-> 0) + +-- | Construct a non-greedy version of 'zeroOrMore'. +zeroOrMore' :: a -> Repeatable a +zeroOrMore' = setNonGreedy . zeroOrMore + +{- | + * Construct a greedy 'Repeatable', from a polymorphic datum, with lower 'RepetitionBounds' == one. + + * A specific case of '^#->'. +-} +oneOrMore :: a -> Repeatable a +oneOrMore = (^#-> 1) + +-- | Construct a non-greedy version of 'oneOrMore'. +oneOrMore' :: a -> Repeatable a +oneOrMore' = setNonGreedy . oneOrMore + +-- | Reduces a 'Repeatable', with a range of 'RepetitionBounds', to a precise number of repetitions. +focus :: Repeatable a -> Repetitions -> Repeatable a +focus r i = r { repetitionBounds = precisely i } + +{- | + * Reduces a 'Repeatable', with a range of 'RepetitionBounds', to a singleton. + + * A degenerate case of 'focus'. +-} +toSingleton :: Repeatable a -> Repeatable a +toSingleton = (`focus` 1) + +-- | Accessor. +getFewest :: Repeatable a -> Repetitions +getFewest MkRepeatable { repetitionBounds = (f, _) } = f + +-- | Accessor. +getMost :: Repeatable a -> Maybe Repetitions +getMost MkRepeatable { repetitionBounds = (_, m) } = m + +-- | The token used to denote /non-greedy/, when in the 'String'-form. +nonGreedyToken :: Char +nonGreedyToken = '?' + +{- | + * The token used to denote 'zeroOrMore', when in the 'String'-form. + + * AKA /Kleene Star/. +-} +zeroOrMoreToken :: Char +zeroOrMoreToken = '*' + +-- | The token used to denote 'zeroOrOne', when in the 'String'-form. +zeroOrOneToken :: Char +zeroOrOneToken = '?' + +-- | The token used to denote 'oneOrMore', when in the 'String'-form. +oneOrMoreToken :: Char +oneOrMoreToken = '+' + +-- | The delimiters of '^#->#', when in the 'String'-form. +rangeDelimiters :: (Char, Char) +rangeDelimiters = ('{', '}') + +-- | The token used to separate 'RepetitionBounds', when in the 'String'-form. +rangeSeparatorToken :: Char +rangeSeparatorToken = ',' + +-- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. +tokens :: String +tokens = Data.List.nub [nonGreedyToken, zeroOrMoreToken, zeroOrOneToken, oneOrMoreToken, fst rangeDelimiters, snd rangeDelimiters, rangeSeparatorToken] + diff --git a/src-lib/RegExDot/Result.hs b/src-lib/RegExDot/Result.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/Result.hs @@ -0,0 +1,60 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] Tools to manipulate a 'RegEx.Result'. + + [@CAVEAT@] The data-definition remains in "RegEx", since it references 'RegEx.Match'. +-} + +module RegExDot.Result( +-- * Functions +-- ** Accessors (Deconstructors) + getMatchList, + getPreMatch, + getPostMatch, +-- ** Predicates + isMatch, +-- ** Query + countMatches +) where + +import qualified Data.Maybe +import qualified RegExDot.RegEx as RegEx + +-- | Accessor. +getPreMatch :: RegEx.Result a -> RegEx.ExternalMatch a +getPreMatch (externalMatch, _, _) = externalMatch + +-- | Accessor. +getPostMatch :: RegEx.Result a -> RegEx.ExternalMatch a +getPostMatch (_, _, externalMatch) = externalMatch + +-- | Accessor. +getMatchList :: RegEx.Result a -> Maybe (RegEx.MatchList a) +getMatchList (_, maybeMatchList, _) = maybeMatchList + +-- | True if the 'RegEx.InputData' matched the 'RegEx.ExtendedRegEx'. +isMatch :: RegEx.Result a -> Bool +isMatch = Data.Maybe.isJust . getMatchList + +-- | Counts the number of top-level 'RegEx.Match'es. +countMatches :: RegEx.Result a -> Int +countMatches (_, Just matchList, _) = length matchList +countMatches _ = 0 + diff --git a/src-lib/RegExDot/ShowablePredicate.hs b/src-lib/RegExDot/ShowablePredicate.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/ShowablePredicate.hs @@ -0,0 +1,53 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] Used to implement /Perl-shortcut/s, /BracketExpression/-ranges & /POSIX Character-classes/. +-} + +module RegExDot.ShowablePredicate( +-- * Types + Predicate, +-- ** Data-types + ShowablePredicate( + MkShowablePredicate, + name, + predicate + ) +) where + +import qualified Control.DeepSeq + +-- | An arbitrary polymorphic predicate function. +type Predicate a = a -> Bool + +-- | Container for both the /predicate/, & the name used in the implementation of 'Show'. +data ShowablePredicate a = MkShowablePredicate { + name :: String, -- ^ The identifier of this predicate when in string-form. + predicate :: Predicate a -- ^ The function used to determine whether an input datum matches. +} + +instance Show (ShowablePredicate a) where + showsPrec _ = showString . name + +instance Eq (ShowablePredicate a) where + l == r = name l == name r -- Ignore 'predicate'. + +instance Control.DeepSeq.NFData (ShowablePredicate a) where + rnf = Control.DeepSeq.rnf . name -- Ignore 'predicate'. + diff --git a/src-lib/RegExDot/Span.hs b/src-lib/RegExDot/Span.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/Span.hs @@ -0,0 +1,64 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] Delimits a data-sequence, via its offset & length, within a wider list. +-} + +module RegExDot.Span( +-- * Types +-- ** Type-synonyms + Span, +-- * Functions + after, + empty, + join +) where + +import Control.Arrow((***)) +import qualified RegExDot.ConsumptionBounds as ConsumptionBounds + +{- | + * The offset & length of a specific data-sequence, within a wider list. + + * Similar to . +-} +type Span = (ConsumptionBounds.DataLength {-offset-}, ConsumptionBounds.DataLength) + +-- | An empty instance, located at the specified offset. +empty :: ConsumptionBounds.DataLength -> Span +empty offset = (offset, 0) + +-- | Returns the furthest extent, i.e. the offset just after the end. +after :: Span -> ConsumptionBounds.DataLength +after = uncurry (+) + +{- | + * Condenses a list into a single value. + + * Expects a gap-free, sorted list. +-} +join + :: ConsumptionBounds.DataLength -- ^ The offset into the list of input-data to use when a null list of spans is received. + -> [Span] + -> Span +join offset [] = empty offset -- The offset can't be deduced from a null list, so use the value provided. +join _ [singleton] = singleton -- Merely for efficiency. +join _ spanList = head *** sum $ unzip spanList + + diff --git a/src-lib/RegExDot/Tree.hs b/src-lib/RegExDot/Tree.hs new file mode 100644 --- /dev/null +++ b/src-lib/RegExDot/Tree.hs @@ -0,0 +1,80 @@ +{- + Copyright (C) 2010 Dr. Alistair Ward + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . +-} +{- | + [@AUTHOR@] Dr. Alistair Ward + + [@DESCRIPTION@] + + * Defines a specific type of tree-structure, which is composed from either a datum, or a list of Tree-lists; . + + * This more general tree-structure has the shape of a regular-expression match. + + * This contrasts with the typical binary-tree, which is either empty, or contains a (left Tree, datum, right Tree). +-} + +module RegExDot.Tree( +-- * Types +-- ** Type-synonyms +-- TreeList, +-- ** Data-types + Tree(..), +-- * Functions + pop +) where + +import qualified Control.Arrow +import qualified Control.DeepSeq +import qualified Data.Foldable +import qualified Data.List +import qualified Data.Monoid + +-- | A general purpose tree-type structure. +data Tree a = + Leaf a -- ^ The payload. + | Node [TreeList a] -- ^ Recurse. NB: a list of lists is required to contain the /MatchLists/ resulting from repeated /Alternatives/. + deriving Eq + +-- | A list of 'Tree's; significant only because it is the essence of the recursive nature of 'Tree'. +type TreeList a = [Tree a] + +instance Show a => Show (Tree a) where + showsPrec _ (Leaf a) = shows a + showsPrec _ (Node treeLists) = showList treeLists + +instance Read a => Read (Tree a) where + readsPrec _ node@('[' : _) = Control.Arrow.first Node `map` readList node {-singleton-} + readsPrec _ leaf = Control.Arrow.first Leaf `map` reads leaf {-singleton-} + +instance Control.DeepSeq.NFData a => Control.DeepSeq.NFData (Tree a) where + rnf (Leaf a) = Control.DeepSeq.rnf a + rnf (Node l) = Control.DeepSeq.rnf l + +instance Functor Tree where + fmap f (Leaf a) = Leaf $ f a + fmap f (Node treeLists) = Node $ map (fmap {-recurse-} f `map`) treeLists + +instance Data.Foldable.Foldable Tree where + foldMap f (Leaf a) = f a -- CAVEAT: 'f' should be Associative, as required by a Monoid. + foldMap f (Node treeLists) = Data.List.foldl' (Data.List.foldl' (\monoid -> (monoid `Data.Monoid.mappend`) . Data.Foldable.foldMap f)) Data.Monoid.mempty treeLists + +-- | Deconstruct the specified 'Node'; i.e. lop the apex from the 'Tree', leaving a flat top. +pop + :: Tree a -- ^ The tree from which to extract the list of trees hanging immediately beneath its apex. + -> [TreeList a] +pop (Node treeLists) = treeLists +pop _ = error "RegExDot.Tree.pop:\tunexpected Leaf" + diff --git a/src/RegExDot/Anchor.hs b/src/RegExDot/Anchor.hs deleted file mode 100644 --- a/src/RegExDot/Anchor.hs +++ /dev/null @@ -1,90 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * A type of /zero-width assertion/: - the 'Bow'-predicate is 'True' if no input-data has yet been consumed; the 'Stern'-predicate is 'True' if no input-data remains to be consumed. - - * Whilst at the top-level of the regex, the /absence/ of an anchor is equivalent to a non-capturing @.*?@, - tempting one to implement them using this property, - it doesn't apply to those anchors encountered at either end of a sub-expression. - So in all cases, an anchor constrains the permissible consumption of input-data, - but at the top level it also prevents the regex drifting away from either end of the input-data. --} - -module RegExDot.Anchor( --- * Types --- ** Data-types - Anchor(..), --- ** Type-synonyms - AnchorSpecification, --- * Constants - bowToken, - sternToken, - tokens, - unanchored -) where - -import qualified Control.DeepSeq - --- | Defines the types on /anchor/ by which a /regex/ can be moored to a part of the input-data. -data Anchor = - Bow -- ^ Matches only if no input data has yet been consumed. Can only exist at the start of the entire regex, or (in theory) the start of any /alternative/. - | Stern -- ^ Matches only if no input data remains to be consumed. Can only exist at the end of the entire regex, or (in theory) the end of any /alternative/. - deriving ( - Eq --- Read, -- See specialisation below. --- Show -- See specialisation below. - ) - -instance Show Anchor where - showsPrec _ Bow = showChar bowToken - showsPrec _ Stern = showChar sternToken - -instance Read Anchor where - readsPrec _ [] = [] -- No parse. - readsPrec _ (' ' : s) = reads s -- Consume white-space. - readsPrec _ ('\t' : s) = reads s -- Consume white-space. - readsPrec _ (c : s) = case c `lookup` [(bowToken, Bow), (sternToken, Stern)] of - Just anchor -> [(anchor, s)] - _ -> [] -- No parse. - -instance Control.DeepSeq.NFData Anchor where - rnf _ = () - --- | The conventional token used to denote a 'Bow'-anchor, when in 'String'-form. -bowToken :: Char -bowToken = '^' - --- | The conventional token used to denote a 'Stern'-anchor, when in 'String'-form. -sternToken :: Char -sternToken = '$' - --- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. -tokens :: String -tokens = [bowToken, sternToken] - --- | A specification for the anchors required around the expression -type AnchorSpecification = (Maybe Anchor, Maybe Anchor) - --- | The specification for an unanchored expression. -unanchored :: AnchorSpecification -unanchored = (Nothing, Nothing) - diff --git a/src/RegExDot/BracketExpression.hs b/src/RegExDot/BracketExpression.hs deleted file mode 100644 --- a/src/RegExDot/BracketExpression.hs +++ /dev/null @@ -1,63 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - This concept was traditionally known as a /Character-class/, - but to distinguish it from the more recently introduced & narrower /POSIX Character-class/, it has been renamed /Bracket-expression/. --} - -module RegExDot.BracketExpression( --- * Types --- ** Type-synonyms - BracketExpression, --- * Constants - delimiterTokens, - negationToken, - tokens, --- * Functions --- ** Predicates - containsMatch -) where - -import RegExDot.BracketExpressionMember((=~)) -import qualified RegExDot.BracketExpressionMember as BracketExpressionMember - --- | A /Bracket-expression/ consists of a list of 'BracketExpressionMember.Member's. -type BracketExpression m = [BracketExpressionMember.Member m] - --- | True if the specified datum matches any 'BracketExpressionMember.Member' of the 'BracketExpression'. -containsMatch :: Eq m - => m -- ^ The input datum. - -> BracketExpression m -- ^ The list of bracket-expression members within which to find a match. - -> Bool -containsMatch datum = any (datum =~) - --- | The delimiters of a /bracket-expression/, when in 'String'-form. -delimiterTokens :: (Char, Char) -delimiterTokens = ('[', ']') - --- | Used to denote a negated /bracket-expression/, when in 'String'-form. -negationToken :: Char -negationToken = '^' - --- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. -tokens :: String -tokens = [fst delimiterTokens, snd delimiterTokens, negationToken] - diff --git a/src/RegExDot/BracketExpressionMember.hs b/src/RegExDot/BracketExpressionMember.hs deleted file mode 100644 --- a/src/RegExDot/BracketExpressionMember.hs +++ /dev/null @@ -1,93 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - Permits /Perl-style shortcut/s to be canned & assigned a single-'Char' mnemonic for subsequent reference; - the implementation of 'Read' looks for a back-slashed 'Char', for which it expects there to be a corresponding canned 'ShowablePredicate.ShowablePredicate'. - - [@CAVEATS@] - - Since the underlying polymorphic data-type isn't required to implement neither 'Enum' nor 'Ord', the implementation of 'Read' can't cope with range-specifications. - Lacking this, Bracket-expression members must be enumerated exhaustively. --} - -module RegExDot.BracketExpressionMember( --- * Type-classes - ShortcutExpander(..), --- * Types --- ** Data-types - Member(..), --- * Functions --- ** Operators - (=~) -) where - -import qualified Control.Arrow -import qualified Control.DeepSeq -import qualified RegExDot.ShowablePredicate as ShowablePredicate - -infix 4 =~ -- Same as (==). - -{- | - * The interface via which /Perl-style shortcut/s are expanded (when they occur within a /bracket-expression/), in a manner appropriate to the chosen type-parameter. - - * The expansion of /Perl-style shortcut/s, is more restricted inside than outside, a /bracket-expression/, - & consequently are merely represented here by a 'ShowablePredicate.ShowablePredicate', rather than providing a more general form suitable also for those /Perl-style shortcuts/ found outside /bracket-expression/s. - - * This interface is implemented elsewhere, where the specific type-parameter & consequently the appropriate set of /Perl-style shortcut/s, are defined. --} -class ShortcutExpander a where - findPredicate :: Char -> Maybe (ShowablePredicate.ShowablePredicate a) -- ^ Attempt to find the appropriate 'ShowablePredicate.ShowablePredicate' to implement this /Perl-style shortcut/. - -{- | - * A /BracketExpression/ can contain either a literal, a range of literals given @(Enum a, Ord a)@, a /Perl-style shortcut/, or when 'Char' is the type-parameter, a /POSIX Character-class/. - - * This data-type reduces the representation of all these possibilities to either a predicate or a literal. --} -data Member m = - Predicate (ShowablePredicate.ShowablePredicate m) -- ^ This 'Member' is described using a /predicate/, which is run to determine whether the datum conforms & is a member of the "BracketExpression". - | Literal m -- ^ This 'Member' is defined literally, using an item of the polymorphic type. - deriving Eq - -instance Show m => Show (Member m) where - showsPrec _ (Predicate showablePredicate) = shows showablePredicate - showsPrec _ (Literal literal) = shows literal - -instance (ShortcutExpander m, Read m) => Read (Member m) where - readsPrec _ [] = [] -- No parse. - readsPrec _ (' ' : s) = reads s -- Consume white-space. - readsPrec _ ('\t' : s) = reads s -- Consume white-space. - readsPrec _ ('\\' : shortcut : s) = case findPredicate shortcut of - Just showablePredicate -> [(Predicate showablePredicate, s)] - _ -> error $ "readsPrec RegExDot.BracketExpressionMember.Member:\tfindPredicate failed for shortcut " ++ show shortcut - readsPrec _ literal = Control.Arrow.first Literal `map` reads literal - -instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (Member m) where - rnf (Predicate showablePredicate) = Control.DeepSeq.rnf showablePredicate - rnf (Literal literal) = Control.DeepSeq.rnf literal - --- | Match-operator. -(=~) :: Eq m - => m -- ^ The input datum. - -> Member m -- ^ The member of the bracket-expression against which the input-datum is to be matched. - -> Bool -- ^ The result of the match-operation. -datum =~ Predicate showablePredicate = ShowablePredicate.predicate showablePredicate datum -datum =~ Literal literal = datum == literal - diff --git a/src/RegExDot/CompilationOptions.hs b/src/RegExDot/CompilationOptions.hs deleted file mode 100644 --- a/src/RegExDot/CompilationOptions.hs +++ /dev/null @@ -1,48 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * Defines those options relating to compile-time. - - * These tend to be options which affect the result, rather than the means by which it is achieved. - - * Whilst similar structures are present in other regex-implementations, there's no standardisation of the fields. --} - -module RegExDot.CompilationOptions( --- * Types --- ** Data-types - CompilationOptions(..) -) where - -import qualified ToolShed.Defaultable -import qualified ToolShed.Options - --- | The switch(es) used to control compilation of the /regex/-engine. -data CompilationOptions = MkCompilationOptions { - complyStrictlyWithPosix :: Bool -- ^ Define the offset of captured data, corresponding to a sub-expression which matched zero times, as the artificial value @-1@ specified by POSIX. -} deriving (Eq, Show) - -instance ToolShed.Defaultable.Defaultable CompilationOptions where - defaultValue = ToolShed.Options.blankValue { complyStrictlyWithPosix = True } - -instance ToolShed.Options.Options CompilationOptions where - blankValue = MkCompilationOptions { complyStrictlyWithPosix = undefined } - diff --git a/src/RegExDot/Consumer.hs b/src/RegExDot/Consumer.hs deleted file mode 100644 --- a/src/RegExDot/Consumer.hs +++ /dev/null @@ -1,105 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * Define the requirements & capabilities of a data-type which is designed to consume data. - - * This module also facilitates investigation of group-behaviour, by defining operations on either concatenations or alternations of 'Consumer's. - - * This module is specifically for use in a /regex/-engine, - in which the meta-data from which the /regex/ is composed, are considered to /consume/ input data, - whilst concatenation & alternation of such elements, behave as a different types of groups of consumer. --} - -module RegExDot.Consumer( --- * Type-classes - Consumer(..), --- * Types --- ** Type-synonyms --- StarHeight, --- * Functions - accumulateConsumptionProfiles, --- accumulateConsumptionProfilesFrom, - aggregateConsumptionProfilesFromAlternatives, - aggregateConsumptionProfilesFromConcatenation, --- ** Query - getConsumptionBounds, - getFewest, - getHasSpecificRequirement -) where - -import RegExDot.ConsumptionProfile((<>), (|+|)) -import qualified Data.List -import qualified RegExDot.ConsumptionBounds as ConsumptionBounds -import qualified RegExDot.ConsumptionProfile as ConsumptionProfile - -{- | - * A measure of the complexity of a /regex/, which has some baring on either the time-complexity or the space-complexity of the solution. - - * --} -type StarHeight = Int - -{- | - * The interface, to which types, which have a data-requirement, may conform. - - * Regrettably, methods requiring reference to the type of the consumable, would require multi-parameter type-classes. - - * 'consumptionProfile' defines the quantity-range of data which may be consumed, & any specific data-requirements. - - * 'starHeight' measures the complexity of the 'Consumer'. --} -class Consumer c where - consumptionProfile :: c -> ConsumptionProfile.ConsumptionProfile - starHeight :: c -> StarHeight - -instance Consumer c => Consumer [c] where - consumptionProfile = undefined - starHeight = Data.List.foldl' (\l -> max l . starHeight) 0 - --- | Convenience-function, to query the 'ConsumptionBounds.ConsumptionBounds' of a 'Consumer'. -getConsumptionBounds :: Consumer c => c -> ConsumptionBounds.ConsumptionBounds -getConsumptionBounds = ConsumptionProfile.consumptionBounds . consumptionProfile - --- | Determine the minimum acceptable quantity of data. -getFewest :: Consumer c => c -> ConsumptionBounds.DataLength -getFewest = fst . getConsumptionBounds - --- | Determine whether the specified 'Consumer', has a specific requirement. -getHasSpecificRequirement :: Consumer c => c -> Bool -getHasSpecificRequirement = ConsumptionProfile.hasSpecificRequirement . consumptionProfile - --- | The aggregate of the specified alternation of 'ConsumptionProfile.ConsumptionProfile's. -aggregateConsumptionProfilesFromAlternatives :: Consumer c => [c] -> ConsumptionProfile.ConsumptionProfile -aggregateConsumptionProfilesFromAlternatives [] = error "RegExDot.Consumer.aggregateConsumptionProfilesFromAlternatives:\tnull list" -aggregateConsumptionProfilesFromAlternatives l = Data.List.foldl1' (<>) $ consumptionProfile `map` l - --- | Get the 'ConsumptionProfile.ConsumptionProfile's for the specified list of 'Consumer's, then find the net effect of concatenating them. -aggregateConsumptionProfilesFromConcatenation :: Consumer c => [c] -> ConsumptionProfile.ConsumptionProfile -aggregateConsumptionProfilesFromConcatenation = Data.List.foldl' (\acc -> (acc |+|) . consumptionProfile) ConsumptionProfile.zero {-initial value-} - --- | Get the 'ConsumptionProfile.ConsumptionProfile' for the specified list of 'Consumer's, then accumulate them, from the specified initial value. -accumulateConsumptionProfilesFrom :: Consumer c => ConsumptionProfile.ConsumptionProfile -> [c] -> ConsumptionProfile.AccumulatedConsumptionProfiles -accumulateConsumptionProfilesFrom = scanr ((|+|) . consumptionProfile) - --- | Get the 'ConsumptionProfile.ConsumptionProfile' for the specified list of 'Consumer's, then accumulate them. -accumulateConsumptionProfiles :: Consumer c => [c] -> ConsumptionProfile.AccumulatedConsumptionProfiles -accumulateConsumptionProfiles = {-init .-} accumulateConsumptionProfilesFrom ConsumptionProfile.zero -- It's useful to leave the initial value at the end of the list. - diff --git a/src/RegExDot/ConsumptionBounds.hs b/src/RegExDot/ConsumptionBounds.hs deleted file mode 100644 --- a/src/RegExDot/ConsumptionBounds.hs +++ /dev/null @@ -1,48 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] Describes the bounds of data-consumption. --} - -module RegExDot.ConsumptionBounds( --- * Types --- ** Type-synonyms - ConsumptionBounds, - DataLength, --- * Constants - zero, --- * Functions --- ** Predicates - isPrecise -) where - --- | A measure of the quantity of an unspecified type of data. -type DataLength = Int - --- | The minimum & maximum bounds of potential data-consumption. -type ConsumptionBounds = (DataLength, Maybe DataLength) - --- | Exactly zero consumption. -zero :: ConsumptionBounds -zero = (0, Just 0) - --- | Predicate which is 'True' if only a precise quantity of data is consumable; no more, no less. -isPrecise :: ConsumptionBounds -> Bool -isPrecise (fewest, most) = Just fewest == most - diff --git a/src/RegExDot/ConsumptionProfile.hs b/src/RegExDot/ConsumptionProfile.hs deleted file mode 100644 --- a/src/RegExDot/ConsumptionProfile.hs +++ /dev/null @@ -1,148 +0,0 @@ -{- - Copyright (C) 2010-2015 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - A 'RegExDot.ConsumptionProfile' is composed from both a capacity to consume, & an ability to discriminate based on the specific data. - The former is described by 'consumptionBounds' & the latter by both 'hasSpecificRequirement' & 'canConsumeAnything'. --} - -module RegExDot.ConsumptionProfile( --- * Types --- ** Type-synonyms - AccumulatedConsumptionProfiles, --- ** Data-types - ConsumptionProfile(..), --- * Constants - zero, --- * Functions - accumulateFrom, - aggregateFromConcatenation, --- ** Operators - (|+|), - (<>), --- ** Predicates --- isPrecise, - withinConsumptionBounds -) where - -import Prelude hiding ((<$>), (<*>)) -- The "Prelude" from 'base-4.8' exports this symbol. -import Control.Applicative((<$>), (<*>)) -import Control.Arrow((***)) -import qualified Data.List -import qualified Data.Maybe -import qualified RegExDot.ConsumptionBounds as ConsumptionBounds -import qualified ToolShed.SelfValidate - -infixr 5 |+| -- Same as (++). -infixr 2 <> -- Same as (||). - -{- | - * A 'Consumer' is considered to have a 'ConsumptionProfile' composed from both a capacity to consume, & an ability to discriminate. - - * Whilst 'hasSpecificRequirement' & 'canConsumeAnything' look like opposites, they can assume independent values; the instance can be both, but only be neither when empty. --} -data ConsumptionProfile = MkConsumptionProfile { - consumptionBounds :: ConsumptionBounds.ConsumptionBounds, -- ^ The permissible quantity of data, which can be consumed. - hasSpecificRequirement :: Bool, -- ^ Whether at least one specific input datum is required. - canConsumeAnything :: Bool -- ^ Whether at least one arbitrary input datum can be consumed. -} deriving (Eq, Read, Show) - -instance ToolShed.SelfValidate.SelfValidator ConsumptionProfile where - getErrors c@( - MkConsumptionProfile { - consumptionBounds = (fewest, most), - hasSpecificRequirement = hasSpecificRequirement', - canConsumeAnything = canConsumeAnything' - } - ) = ToolShed.SelfValidate.extractErrors [ - (fewest < 0, "Negative fewest=" ++ show fewest ++ "."), - ( - case most of - Just m -> m < fewest - _ -> False, - "Invalid 'most'; " ++ show c ++ "." - ), - (not $ or [c == zero, hasSpecificRequirement', canConsumeAnything'], "Invalid " ++ show c ++ ".") - ] - -{- --- | True if there's no choice in the quantity of data to consume. -isPrecise :: ConsumptionProfile -> Bool -isPrecise = ConsumptionBounds.isPrecise . consumptionBounds --} - --- | Predicate, which is 'True' if the specified data-length, falls within the specified 'ConsumptionBounds.ConsumptionBounds'. -withinConsumptionBounds - :: ConsumptionBounds.DataLength -- ^ The actual quantity of data consumed. - -> ConsumptionProfile -- ^ The bounds within which data-consumption is required to fall. - -> Bool -withinConsumptionBounds dataLength = uncurry (&&) . ((dataLength >=) *** Data.Maybe.maybe True (dataLength <=)) . consumptionBounds - --- | Both minimum & maximum set to zero; which can be used as the initial value when accumulating the sum of a list. -zero :: ConsumptionProfile -zero = MkConsumptionProfile { - consumptionBounds = ConsumptionBounds.zero, - hasSpecificRequirement = False, - canConsumeAnything = False -} - --- | The net effect of two concatenated 'ConsumptionProfile's. -(|+|) :: ConsumptionProfile -> ConsumptionProfile -> ConsumptionProfile -MkConsumptionProfile { - consumptionBounds = (lf, ls), - hasSpecificRequirement = lh, - canConsumeAnything = lc -} |+| MkConsumptionProfile { - consumptionBounds = (rf, rs), - hasSpecificRequirement = rh, - canConsumeAnything = rc -} = MkConsumptionProfile { - consumptionBounds = (lf + rf, (+) <$> ls <*> rs), -- The sum of those of the concatenation. - hasSpecificRequirement = lh || rh, -- The concatenation mandates consumption of at least one specific input datum, if either 'ConsumptionProfile' does. - canConsumeAnything = lc || rc -- The concatenation can consume at least one arbitrary input datum, if either 'ConsumptionProfile' can. -} - --- | The net effect of two alternative 'ConsumptionProfile's. -(<>) :: ConsumptionProfile -> ConsumptionProfile -> ConsumptionProfile -MkConsumptionProfile { - consumptionBounds = (lf, ls), - hasSpecificRequirement = lh, - canConsumeAnything = lc -} <> MkConsumptionProfile { - consumptionBounds = (rf, rs), - hasSpecificRequirement = rh, - canConsumeAnything = rc -} = MkConsumptionProfile { - consumptionBounds = (lf `min` rf, max <$> ls <*> rs), -- Stretched to envelope alternatives. - hasSpecificRequirement = lh && rh, -- The alternation mandates consumption of at least one specific input datum, if both 'ConsumptionProfile's do. - canConsumeAnything = lc || rc -- The alternation can consume at least one arbitrary input datum, if either 'ConsumptionProfile' can. -} - --- | The aggregate of the specified concatenation of 'ConsumptionProfile's. -aggregateFromConcatenation :: [ConsumptionProfile] -> ConsumptionProfile -aggregateFromConcatenation = Data.List.foldl' (|+|) zero {-initial value-} - --- | The 'ConsumptionProfile's of all the aggregated tails of a list of 'Consumer's. -type AccumulatedConsumptionProfiles = [ConsumptionProfile] - --- | Accumulate the specified list of 'ConsumptionProfile's, from the specified initial value. -accumulateFrom :: ConsumptionProfile -> [ConsumptionProfile] -> AccumulatedConsumptionProfiles -accumulateFrom = scanr (|+|) - diff --git a/src/RegExDot/DSL.hs b/src/RegExDot/DSL.hs deleted file mode 100644 --- a/src/RegExDot/DSL.hs +++ /dev/null @@ -1,117 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * Defines /binary/ operators, to form a /Domain-specific Language/, by which to compose 'RegEx.Concatenations' from 'RegEx.Pattern's. - - * Each operator quantifies the specified 'RegEx.Pattern', & prepends it to the specified 'RegEx.Concatenation'. - - [@TODO@] Could @[]@ & 'Meta' be instances of a type-class which has methods (-:) etc., - allowing the former to prepend 'RegEx.captureGroup' & the latter 'RegEx.Require' ? - --} - -module RegExDot.DSL( --- * Functions --- ** Operators - (-:), - (?:), - (??:), - (*:), - (*?:), - (+:), - (+?:), - ( #->#:), - ( #->#?:), - ( #->:), - ( #->?:), - ( #:), - (<~>) -) where - -import RegExDot.Repeatable((^#->#), (^#->#?), (^#->), (^#->?), (^#)) -import qualified RegExDot.Anchor as Anchor -import qualified RegExDot.RegEx as RegEx -import qualified RegExDot.Repeatable as Repeatable - -infixr 5 -:, ?:, ??:, *:, *?:, +:, +?:,#->#:, #->#?:, #->:, #->?:, #:, <~> -- Same as for ':', & lower than Repeatable's operators. - --- | Prepend an unrepeated 'RegEx.Pattern', to the specified 'RegEx.Concatenation'. -(-:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a -(-:) pattern = (Repeatable.one pattern :) - --- | Prepend an optional 'RegEx.Pattern', to the specified 'RegEx.Concatenation'. -(?:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a -(?:) pattern = (Repeatable.zeroOrOne pattern :) - --- | A /non-greedy/ version of '?:'. -(??:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a -(??:) pattern = (Repeatable.zeroOrOne' pattern :) - --- | Prepend a 'RegEx.Pattern', repeatable zero or more times, to the specified 'RegEx.Concatenation'. -(*:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a -(*:) pattern = (Repeatable.zeroOrMore pattern :) - --- | A /non-greedy/ version of '*:'. -(*?:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a -(*?:) pattern = (Repeatable.zeroOrMore' pattern :) - --- | Prepend a 'RegEx.Pattern', repeatable one or more times, to the specified 'RegEx.Concatenation'. -(+:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a -(+:) pattern = (Repeatable.oneOrMore pattern :) - --- | A /non-greedy/ version of '+:'. -(+?:) :: RegEx.Pattern a -> RegEx.Concatenation a -> RegEx.Concatenation a -(+?:) pattern = (Repeatable.oneOrMore' pattern :) - --- | Prepend a 'RegEx.Pattern', repeated a range of times, to the specified 'RegEx.Concatenation'. -( #->#:) :: (RegEx.Pattern a, Repeatable.RepetitionBounds) -> RegEx.Concatenation a -> RegEx.Concatenation a -( #->#:) (pattern, bounds) = (pattern ^#-># bounds :) - --- | A /non-greedy/ version of '#->#:'. -( #->#?:) :: (RegEx.Pattern a, Repeatable.RepetitionBounds) -> RegEx.Concatenation a -> RegEx.Concatenation a -( #->#?:) (pattern, bounds) = (pattern ^#->#? bounds :) - --- | Prepend a 'RegEx.Pattern', repeated at least a specified number of times, to the specified 'RegEx.Concatenation'. -( #->:) :: (RegEx.Pattern a, Repeatable.Repetitions) -> RegEx.Concatenation a -> RegEx.Concatenation a -( #->:) (pattern, fewest) = (pattern ^#-> fewest :) - --- | A /non-greedy/ version of '#->:'. -( #->?:) :: (RegEx.Pattern a, Repeatable.Repetitions) -> RegEx.Concatenation a -> RegEx.Concatenation a -( #->?:) (pattern, fewest) = (pattern ^#->? fewest :) - --- | Prepend a 'RegEx.Pattern', repeated a precise number of times, to the specified 'RegEx.Concatenation'. -( #:) :: (RegEx.Pattern a, Repeatable.Repetitions) -> RegEx.Concatenation a -> RegEx.Concatenation a -( #:) (pattern, r) = (pattern ^# r :) - -{- | - * Sandwiches a 'RegEx.Concatenation' between optional 'Anchor.Anchor's to construct a 'RegEx.ExtendedRegEx'. - - * Accounts for the unusual requirement to place an 'Anchor.Stern' at the start of the 'RegEx.Concatenation', or an 'Anchor.Bow' at the end. --} -(<~>) - :: Anchor.AnchorSpecification -- ^ A specification for the anchors required around the expression. - -> RegEx.Concatenation a -- ^ The bare list of repeatable patterns to anchor. - -> RegEx.ExtendedRegEx a -- ^ The completed regex. -(bow, stern) <~> concatenation = RegEx.MkExtendedRegEx { - RegEx.bowAnchor = bow, - RegEx.concatenation = concatenation, - RegEx.sternAnchor = stern -} diff --git a/src/RegExDot/DataSpan.hs b/src/RegExDot/DataSpan.hs deleted file mode 100644 --- a/src/RegExDot/DataSpan.hs +++ /dev/null @@ -1,61 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] Augments a 'Span.Span', with the specific data concerned. --} - -module RegExDot.DataSpan( --- * Types --- ** Type-synonyms - DataSpan, --- * Functions - after, - empty, - join -) where - -import Control.Arrow((***)) -import qualified RegExDot.ConsumptionBounds as ConsumptionBounds -import qualified RegExDot.Span as Span - -{- | - * Augment 'Span.Span' with a copy of the data to which it refers. - - * Similar to . --} -type DataSpan a = ([a], Span.Span) - --- | Constructs an empty instance, located at the specified offset. -empty :: ConsumptionBounds.DataLength -> DataSpan a -empty offset = ([], Span.empty offset) - --- | Returns the furthest extent, i.e. the offset just after the end. -after :: DataSpan a -> ConsumptionBounds.DataLength -after = Span.after . snd - -{- | - * Condenses a list into a single value. - - * Expects a gap-free, sorted list. --} -join - :: ConsumptionBounds.DataLength -- ^ The offset at which the concatenated match is considered to have occurred. - -> [DataSpan a] -- ^ The list to be concatenated. - -> DataSpan a -join offset = (concat *** Span.join offset) . unzip diff --git a/src/RegExDot/DataSpanTree.hs b/src/RegExDot/DataSpanTree.hs deleted file mode 100644 --- a/src/RegExDot/DataSpanTree.hs +++ /dev/null @@ -1,108 +0,0 @@ -{- - Copyright (C) 2010-2015 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] Permits transformation of 'RegEx.MatchList', to facilitate standardisation. --} - -module RegExDot.DataSpanTree( --- * Types --- ** Type-synonyms --- DataSpanTree, --- DataSpanTreeList, --- * Functions - extractCaptureGroups, - flattenTreeList, - toTreeList -) where - -import qualified Data.Foldable -import qualified RegExDot.ConsumptionBounds as ConsumptionBounds -import qualified RegExDot.DataSpan as DataSpan -import qualified RegExDot.RegEx as RegEx -import qualified RegExDot.Tree as Tree - --- | Defines a similar 'Tree.Tree' to 'RegEx.Match', but with different 'Tree.Leaf's. -type DataSpanTree a = Tree.Tree (DataSpan.DataSpan a) - --- | Defines a similar structure to 'RegEx.MatchList'. -type DataSpanTreeList a = [DataSpanTree a] - --- | Converts a 'RegEx.MatchList' into a 'DataSpanTreeList', by transforming the 'Tree.Leaf's. -toTreeList :: RegEx.MatchList a -> DataSpanTreeList a -toTreeList = map toTree where - toTree :: RegEx.Match a -> DataSpanTree a - toTree = fmap toDataSpan where - toDataSpan :: RegEx.MatchedData a -> DataSpan.DataSpan a - toDataSpan (_, inputDataOffset, inputData) = (inputData, (inputDataOffset, length inputData)) - --- | Condenses a 'DataSpanTreeList's into a list of 'DataSpan.DataSpan's, using 'DataSpan.join'. -flattenTreeList - :: ConsumptionBounds.DataLength -- ^ The offset into the input-data at which a match occurred. - -> DataSpanTreeList a -- ^ The tree to flatten. - -> [DataSpan.DataSpan a] -flattenTreeList _ [] = [] -flattenTreeList offset (tree : treeList) = flattenedTree : flattenTreeList (DataSpan.after flattenedTree) treeList where --- flattenedTree :: DataSpan.DataSpan a - flattenedTree = DataSpan.join offset $ Data.Foldable.toList tree - -{- | - * POSIX describes the contents of /capture-groups/, as summarised in . - - * 'RegEx.Result', is a complete description of the match between 'RegEx.InputData' & RegEx.ExtendedRegEx'; this function extracts a POSIX-conformant list from it. - - * The major differences are, that: - - Only data from parenthesized sub-expressions ('RegEx.Alternatives') is captured. - - Only the /last/ repetition of a repeated sub-expression is returned. - . - - The data captured within each parenthesized sub-expression, is summarised as a single 'DataSpan.DataSpan'. - - POSIX specifies a 'Span.Span'-offset of @-1@, for sub-expressions which match zero times; cf sub-expressions which consume nothing, once. - . - @ - ("ace" Text.Regex.Posix.=~ "a(b)*c(d)?e") :: Text.Regex.Base.RegexLike.MatchArray - array (0,2) [(0,(0,3)),(1,(-1,0)),(2,(-1,0))] - - ("ace" Text.Regex.Posix.=~ "a(b*)c(d?)e") :: Text.Regex.Base.RegexLike.MatchArray - array (0,2) [(0,(0,3)),(1,(1,0)),(2,(2,0))] - @ - I consider this a poor convention, resulting from the focus of POSIX on C, which makes subsequent calculation from the list of 'DataSpan's difficult & error-prone. --} -extractCaptureGroups - :: Bool -- ^ Whether to strictly comply with /POSIX/. - -> DataSpanTreeList a -- ^ The tree-structure from which to extract the capture-groups. - -> [DataSpan.DataSpan a] -extractCaptureGroups complyStrictlyWithPosix = extractCaptureGroups' 0 where - extractCaptureGroups' :: ConsumptionBounds.DataLength -> DataSpanTreeList a -> [DataSpan.DataSpan a] - extractCaptureGroups' _ [] = [] - extractCaptureGroups' offset (tree : treeList) = let --- recurseHorizontallyFrom :: ConsumptionBounds.DataLength -> [DataSpan.DataSpan a] - recurseHorizontallyFrom = (`extractCaptureGroups'` treeList) - in case tree of - Tree.Leaf dataSpan -> recurseHorizontallyFrom $ DataSpan.after dataSpan - Tree.Node [] -> DataSpan.empty (if complyStrictlyWithPosix then -1 else offset) : recurseHorizontallyFrom offset -- POSIX specifies an Span-offset of -1, for sub-expressions which match 0 times; cf sub-expressions which consumes nothing, once. - Tree.Node treeLists -> joinedFlattenedTreeList : (extractCaptureGroups' offset lastMatch {-recurse vertically-} ++ recurseHorizontallyFrom (DataSpan.after joinedFlattenedTreeList)) where --- lastMatch :: DataSpanTreeList a - lastMatch = last treeLists -- . - --- joinedFlattenedTreeList :: DataSpan.DataSpan a - joinedFlattenedTreeList = DataSpan.join offset $ flattenTreeList offset lastMatch - diff --git a/src/RegExDot/ExecutionOptions.hs b/src/RegExDot/ExecutionOptions.hs deleted file mode 100644 --- a/src/RegExDot/ExecutionOptions.hs +++ /dev/null @@ -1,102 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * Defines those options relating to the "RegExDot.RegEx"-runtime. - - * These tend to be options which don't affect the result, only the means by which it is achieved. - - * Whilst similar structures are present in other regex-implementations, there's no standardisation of the fields. --} - -module RegExDot.ExecutionOptions( --- * Types --- ** Data-types - ExecutionOptions(..), --- * Functions --- ** Mutators - setVerbose -) where - -import qualified ToolShed.Defaultable -import qualified ToolShed.Options - --- | The switches used to control execution of the /regex/-engine. -data ExecutionOptions = MkExecutionOptions { - abortTrialRepetitionsOnInherentFailure :: Bool, -- ^ If an /alternative/ can't match, irrespective of the subsequent /RegExDot.RegEx.Concatenation/, then avoid futile trial /RegExDot.Repeatable.Repetitions/. The converse of 'checkForUnconsumableData'. - abortTrialRepetitionsOnZeroConsumption :: Bool, -- ^ Check for zero data-consumption by the @n@-th /RegExDot.Repeatable.Repeatable RegExDot.RegEx.CaptureGroup/, before attempting @n+1@. - bypassInputDataForLiberalConsumer :: Bool, -- ^ Whether to bypass reading of the input data, if the mapping to specific /RegExDot.RegEx.RepeatablePattern/s isn't required, & the /RegExDot.RegEx.ExtendedRegEx/ can consume the required quantity of anything. - catchIncompatibleAnchors :: Bool, -- ^ Avoid futile trial solutions, involving repetitions of anchored alternatives, which must consume data. - checkExistenceOfInelasticTail :: Bool, -- ^ If the /RegExDot.RegEx.ExtendedRegEx/ ends in an inelastic (zero /Star-height/) tail, confirm its existence at the end of the /RegExDot.RegEx.InputData/. - checkForUnconsumableData :: Bool, -- ^ Check whether there's no possibility of consuming some of the input data. The converse of 'abortTrialRepetitionsOnInherentFailure'. - moderateGreed :: Bool, -- ^ Greedily consume data, only up to the limit beyond which, future requirements would be compromised. - permitReorderingOfAlternatives :: Bool, -- ^ Permit /RegExDot.RegEx.Alternatives/ to be re-ordered, in an attempt to more quickly locate a result. - preferAlternativesWhichFeedTheGreedy :: Bool, -- ^ Within the /RegExDot.RegEx.MatchedData/ from which each candidate /RegExDot.RegEx.Match/ amongst sequences of /RegExDot.RegEx.Alternatives/, is ultimately composed, prefer /RegExDot.ConsumptionBounds.ConsumptionBounds/ of /RegExDot.RegEx.InputData/, beyond /RegExDot.Repeatable.getFewest/, by /RegExDot.Repeatable.isGreedy RegExDot.RegEx.RepeatablePattern/s. - preferAlternativesWhichMimickUnrolling :: Bool, -- ^ Compare /RegExDot.ConsumptionBounds.ConsumptionBounds/ on successive /RegExDot.Repeatable.Repetitions/ of /RegExDot.RegEx.CaptureGroup/, between candidate /RegExDot.RegEx.Match/es, to mimic the behaviour of the unrolled /RegExDot.Repeatable.Repetitions/. - preferFewerRepeatedAlternatives :: Bool, -- ^ Prefer fewer /RegExDot.Repeatable.Repetitions/ of /RegExDot.RegEx.Alternatives/, to discourage the capture of null lists of /RegExDot.RegEx.InputData/. - requireMatchList :: Bool, -- ^ If merely interested in a 'Bool' result, rather than the optimal mapping of input data to /RegExDot.RegEx.RepeatablePattern/s, avoid unnecessary evaluation of the /RegEx.Match/. - unrollRepeatedSingletonAlternative :: Bool, -- ^ Check whether /RegExDot.RegEx.Alternatives/ consists of just a singleton /RegExDot.RegEx.ExtendedRegEx/, & has therefore been used merely as a capture-group. Though this doesn't affect the result, it vastly improves efficiency. - useFirstMatchAmongAlternatives :: Bool, -- ^ Rather than performing an exhaustive search for the optimal choice amongst /RegExDot.RegEx.Alternatives/, merely select the first that matches; conform to /Perl/ rather than /POSIX/. - validateMinConsumptionOfAlternatives :: Bool -- ^ When the number of repetitions of a /RegExDot.RegEx.CaptureGroup/ is precisely specified, check whether the resulting minimum data-requirement is available. -} deriving (Eq, Show) - -instance ToolShed.Defaultable.Defaultable ExecutionOptions where - defaultValue = setVerbose False $ ToolShed.Options.blankValue { - abortTrialRepetitionsOnInherentFailure = True, -- Regrettably, this slightly reduces performance for most non-pathological patterns. - catchIncompatibleAnchors = True, - checkExistenceOfInelasticTail = True, - checkForUnconsumableData = True, -- Expensive, particularly when (not requireMatchList), & only typically useful in failure-scenarios. - moderateGreed = True, -- Cost may exceed benefit. TODO: confirm. - preferAlternativesWhichFeedTheGreedy = True, - preferAlternativesWhichMimickUnrolling = True, - preferFewerRepeatedAlternatives = True, - unrollRepeatedSingletonAlternative = True, -- Affects only efficiency, not the result. - useFirstMatchAmongAlternatives = False, -- Perl-style matching may be faster, but may also yield a sub-optimal Match. - validateMinConsumptionOfAlternatives = False -- The cost outweighs the small infrequent dividend. - } - -instance ToolShed.Options.Options ExecutionOptions where - blankValue = MkExecutionOptions { - abortTrialRepetitionsOnInherentFailure = undefined, - abortTrialRepetitionsOnZeroConsumption = undefined, - bypassInputDataForLiberalConsumer = undefined, - catchIncompatibleAnchors = undefined, - checkExistenceOfInelasticTail = undefined, - checkForUnconsumableData = undefined, - moderateGreed = undefined, - permitReorderingOfAlternatives = undefined, - preferAlternativesWhichFeedTheGreedy = undefined, - preferAlternativesWhichMimickUnrolling = undefined, - preferFewerRepeatedAlternatives = undefined, - requireMatchList = undefined, - unrollRepeatedSingletonAlternative = undefined, - useFirstMatchAmongAlternatives = undefined, - validateMinConsumptionOfAlternatives = undefined - } - --- | Sets those fields which depend crucially on whether the caller wants to retrieve any /RegExDot.RegEx.MatchList/ from the /RegExDot.RegEx.Result/, or just query whether there is one. -setVerbose :: Bool -> ExecutionOptions -> ExecutionOptions -setVerbose verbose e = e { - abortTrialRepetitionsOnZeroConsumption = verbose, -- The corresponding check, involves evaluation of a /RegExDot.RegEx.MatchList/, which is too expensive if the /RegExDot.RegEx.Matchlist/ isn't otherwise required. - bypassInputDataForLiberalConsumer = not verbose, -- Potentially bypasses reading of /RegExDot.RegEx.InputData/, which is inappropriate if the mapping into a /RegExDot.RegEx.Result/ is required. - permitReorderingOfAlternatives = not verbose, -- Doesn't help when 'requireMatchList', since an exhaustive search of /RegExDot.RegEx.Alternatives/, for the optimal solution, is performed. - requireMatchList = verbose -} - diff --git a/src/RegExDot/InstanceInt.hs b/src/RegExDot/InstanceInt.hs deleted file mode 100644 --- a/src/RegExDot/InstanceInt.hs +++ /dev/null @@ -1,43 +0,0 @@ -{-# OPTIONS_GHC -fno-warn-orphans #-} -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] Provides a default implementation of the type-classes, required for 'RegEx.ExtendedRegEx' Int. - - [@CAVEATS@] - - In contrast to the traditional regex, there's no reason to escape a 'Char', other than when referring to a /shortcut/. - Since no /shortcuts/ are defined in this minimal definition, it's always an error to escape a character. --} - -module RegExDot.InstanceInt() where - -import qualified RegExDot.BracketExpressionMember as BracketExpressionMember -import qualified RegExDot.Meta as Meta -import qualified RegExDot.RegEx as RegEx - -instance BracketExpressionMember.ShortcutExpander Int where - findPredicate _ = Nothing - -instance Meta.ShortcutExpander Int where - expand c = error $ "RegExDot.Meta.ShortcutExpander.expand RegExDot.InstanceInt:\tunrecognised shortcut '" ++ show c ++ "'." -- A shortcut is the only legitimate reason to escape a character. - -instance RegEx.ShortcutExpander Int where - expand c = error $ "RegExDot.RegEx.ShortcutExpander.expand RegExDot.InstanceInt:\tunrecognised shortcut '" ++ show c ++ "'." -- A shortcut is the only legitimate reason to escape a character. - diff --git a/src/RegExDot/Meta.hs b/src/RegExDot/Meta.hs deleted file mode 100644 --- a/src/RegExDot/Meta.hs +++ /dev/null @@ -1,144 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * Describes the set of polymorphic data, which can be matched. - - * Only permits a match against exactly one polymorphic datum, which distinguishes it from a /zero-width assertion/, like an /anchor/, /word-boundary/, or /look-ahead assertion/. - - * Designed to be used by a polymorphic /regex/-engine, to implement the traditional meta-characters; @. [] [^]@. - - * Permits /Perl-style shortcuts/ for commonly used 'Meta'-data, to be canned & assigned a single-'Char' mnemonic for subsequent reference; - the implementation of 'Read' looks for a back-slashed 'Char', for which it expects there to be a corresponding canned 'Meta'. - Since this class is polymorphic, it has no knowledge of what shortcuts might be appropriate for the chosen type-parameter, - so the expansion from the back-slashed 'Char' to corresponding 'Meta'-data, is performed through the 'expand' interface of the 'ShortcutExpander' class, which should be implemented elsewhere. --} - -module RegExDot.Meta( --- * Type-classes - ShortcutExpander(..), --- * Types --- ** Data-types - Meta(..), --- * Constants - shortcutToken, - anyToken, - tokens, --- * Functions --- ** Predicates - isMatch -) where - -import qualified Control.Arrow -import qualified Control.DeepSeq -import qualified RegExDot.BracketExpression as BracketExpression -import qualified RegExDot.BracketExpressionMember as BracketExpressionMember -import qualified RegExDot.Consumer as Consumer -import qualified RegExDot.ConsumptionProfile as ConsumptionProfile -import qualified RegExDot.ShowablePredicate as ShowablePredicate -import qualified ToolShed.SelfValidate - -{- | - * The interface via which /Perl-style shortcut/s are expanded, in a manner appropriate to the chosen type-parameter. - - * Since the expansion of /Perl-style shortcut/s, is more restricted inside than outside a 'BracketExpression.BracketExpression', - the former is considered to be a superclass, providing a base from which to build alternative implementations. --} -class BracketExpressionMember.ShortcutExpander m => ShortcutExpander m where - expand :: Char -> Meta m -- ^ Expand a /Perl-style shortcut/. - --- | Declares a polymorphic data-type. -data Meta m = - Any -- ^ Any datum matches. Equivalent to @NoneOf []@, but more efficient. CAVEAT: independent of the type-parameter @a@. - | Literal m -- ^ The datum matches, if it's equal to the specified value. Equivalent to @AnyOf [BracketExpression.Literal x]@, but more efficient. - | AnyOf (BracketExpression.BracketExpression m) -- ^ The datum matches, if 'BracketExpression.containsMatch'. - | NoneOf (BracketExpression.BracketExpression m) -- ^ The datum matches, if @not BracketExpression.containsMatch@. - | Predicate (ShowablePredicate.ShowablePredicate m) -- ^ The datum matches if 'ShowablePredicate.ShowablePredicate'. - deriving ( - Eq --- Read, -- Specialised below. --- Show -- Specialised below. - ) - -instance ToolShed.SelfValidate.SelfValidator (Meta m) where - getErrors _ = [] - -instance Show m => Show (Meta m) where - showsPrec _ Any = showChar anyToken - showsPrec _ (Literal m) = shows m - showsPrec _ (AnyOf bracketExpression) = shows bracketExpression - showsPrec _ (NoneOf bracketExpression) = showChar x . showChar BracketExpression.negationToken . showString xs where (x : xs) = show $ AnyOf bracketExpression - showsPrec _ (Predicate showablePredicate) = shows showablePredicate - -instance (ShortcutExpander m, Read m) => Read (Meta m) where - readsPrec _ [] = [] -- No parse. - readsPrec _ (' ' : s) = reads s -- Consume white-space. - readsPrec _ ('\t' : s) = reads s -- Consume white-space. - readsPrec _ ('.' : s) = [(Any, s)] - readsPrec _ ('[' : '^' : noneOf) = Control.Arrow.first NoneOf `map` reads (fst BracketExpression.delimiterTokens : noneOf) {-Reconstruct without negation, & recurse-} - readsPrec _ anyOf@('[' : _) = Control.Arrow.first AnyOf `map` reads anyOf {-singleton-} - readsPrec _ ('\\' : c : s) = [(expand c, s)] - readsPrec _ literal = Control.Arrow.first Literal `map` reads literal {-singleton-} - -instance Consumer.Consumer (Meta m) where - consumptionProfile meta = let - hasSpecificDataRequirement :: Bool - hasSpecificDataRequirement = case meta of - Any -> False - NoneOf [] -> False - _ -> True - in ConsumptionProfile.MkConsumptionProfile { - ConsumptionProfile.consumptionBounds = (1, Just 1), - ConsumptionProfile.hasSpecificRequirement = hasSpecificDataRequirement, - ConsumptionProfile.canConsumeAnything = not hasSpecificDataRequirement - } - - starHeight _ = 0 - -instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (Meta m) where - rnf Any = () - rnf (Literal m) = Control.DeepSeq.rnf m - rnf (AnyOf bracketExpression) = Control.DeepSeq.rnf bracketExpression - rnf (NoneOf bracketExpression) = Control.DeepSeq.rnf bracketExpression - rnf (Predicate showablePredicate) = Control.DeepSeq.rnf showablePredicate - --- | True if the specified datum matches. -isMatch :: Eq m - => m -- ^ The input datum. - -> Meta m -- ^ The meta-entity against which the input datum is to be matched. - -> Bool -- ^ The result of the match-operation. -isMatch _ Any = True -isMatch datum (Literal literal) = datum == literal -isMatch datum (AnyOf bracketExpression) = datum `BracketExpression.containsMatch` bracketExpression -isMatch datum (NoneOf bracketExpression) = not $ datum `isMatch` AnyOf bracketExpression -- This implementation leverages future enhancements to 'AnyOf'. -isMatch datum (Predicate showablePredicate) = ShowablePredicate.predicate showablePredicate datum - --- | The token used to precede a /Perl-style shortcut/, when in the 'String'-form. -shortcutToken :: Char -shortcutToken = '\\' - --- | The token used to denote 'Any', when in the 'String'-form. -anyToken :: Char -anyToken = '.' - --- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. -tokens :: String -tokens = [shortcutToken, anyToken] ++ BracketExpression.tokens - diff --git a/src/RegExDot/RegEx.hs b/src/RegExDot/RegEx.hs deleted file mode 100644 --- a/src/RegExDot/RegEx.hs +++ /dev/null @@ -1,1198 +0,0 @@ -{- - Copyright (C) 2010-2015 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * This implementation of extended /regex/es, generalises the familiar concept of pattern-matching of character-strings, to matching lists composed from an arbitrary data-type. - The polymorphic data, from which the input data-list is composed, need only support @Eq@. - - * Because of the unknown stringified form of the underlying polymorphic data, - the /regex/ must be described by a comparatively verbose bracketed & comma-separated list, rather than the traditional /String/ containing Meta-characters. - Each element of this 'Concatenation' is a 'RepeatablePattern', which describes a permissible match against 'InputData'. - - * 'RepeatablePattern' can take one of two forms. - In the simplest case, it matches just a single item of the underlying polymorphic type, perhaps literally, though looser specifications also exist: - @.@ matches any input datum; @[x, y, z]@ matches any of @x@, @y@, or @z@; @[^x, y, z]@ matches anything but @x@, @y@, or @z@. - To support POSIX /ERE/s, 'RepeatablePattern' can also be a list 'Alternatives', each of which is recursively defined as an 'ExtendedRegEx', to form a tree-structure. - - * Each 'Pattern', can optionally be /quantified/ by either a traditional /greedy/, or a /Perl/-style /non-greedy/, suffix, e.g.; @[*, +, ?, {n, m}, {n,}, {n}, *?, +?, ??, {n, m}?, {n,}?]@. - - * For convenience, common specifications can be canned & assigned a single 'Char' mnemonic, for subsequent reference. - Since 'ExtendedRegEx' is polymorphic, the set of abbreviations appropriate in the context of the unspecified base-type, - must be implemented externally through the 'Meta.ShortcutExpander' interface. - This permits the use, when the type-parameter is 'Char', of /Perl-style shortcuts/ @[\\d\\D\\s\\S\\w\\W]@. - - * The algorithm, is the classic /back-tracking/ one, rather than either a /DFA/ or /NFA/. - This permits construction of 'Result' via which one can discover the deep mapping of 'InputData' into 'ExtendedRegEx', - & provides the flexibility to add the features now expected by modern /regex/-engines. - Since the type-parameter is unknown, & may represent a large object, the exponential space-complexity of creating a /DFA/ may present additional problems. - The exponential time-complexity of the /back-tracking/ algorithm is partially tamed by targeting obvious inefficiencies with specific optimisations. - - * Char-based regexen, traditionally overload the delimiters of a set of 'Alternatives' (parentheses), as a request for data-capture. - Here, in contrast, all 'RepeatablePattern's capture data, & repeated sub-expressions capture a list of data, - rather than arbitrarily recording just the last () item. - - [@REFERENCES@] - - * - - * - - * - - * - - * - - * - - * - - [@CAVEATS@] - - * Because of the definition of mutually recursive data-types, it is difficult to split this annoyingly large module, & preserve compatibility across compilers, - but it may be possible to break this cyclic dependency, by defining an interface to which one of the data-types defined here conforms. - - * Doesn't implement /Back-references/, making the definition of the 'ExtendedRegEx' context-free. - - * There's no integration with the type-classes defined in "Text.Regex.Base.RegexLike", which assumes 'Char'-based 'InputData'; - though this could be added to a specialised instance. - - * When 'Alternatives' are defined, 'Result' becomes a tree-like structure. - Unless the alternative is a singleton, the specific alternative selected in the solution is typically unknown, & therefore the /structure/ of the branch of this tree is also unknown. - This lack of clarity is compounded when the 'Alternatives' are 'Repeatable.Repeatable', since a different one may be selected on each successive repetition. - Consequently, the user can't navigate this portion of the structure in a statically defined manner, to acquire the captured data. - Despite this, & in contrast to other /regex/-engines, access to the whole data-structure is available, since it doesn't seem advantage to hide it. - The user can then either use 'extractDataFromMatch' for that element of 'Result', thus aggregating the data from sections of unknown structure, or 'show' it, as an aid to debugging. - - [@TODO@] - - * Test parallel-operation, on a 3 or more processor machine. - If 'rnf' is less effective than 'rwhnf', - then the 'Control.DeepSeq.NFData' context can be removed, - reducing the requirements imposed on the type-parameter 'a'. - - * Try 'Data.List.Stream' (stream-fusion), a faster drop-in replacement for 'Data.List'; possibly integrated in GHC-6.12. - - * 'ExecutionOptions.bypassInputDataForLiberalConsumer' is too restrictive. - More generally, we can test whether the set of different 'a' in 'InputData', is a subset of those common to all remaining terms in the 'ExtendedRegEx'. - Using this rule, we can infer @"aaa ..." =~ MkExtendedRegEx [a,a+,a?,[ab]{2,3}]@, given compatible 'ConsumptionProfile.consumptionBounds'. - - * Nested repetitions, where nothing has been added to the expression, result in repeated trials of the same expression, - e.g.; @"(x{i,}){j,}"@ results in the same expansion for @(i, j) in [(2, 3), (3, 2), (6, 1), (1, 6)]@. - The resulting 'MatchList' may be different, but if the first such trial fails, so will all the remainder. - - * Should cope with empty sets of 'Alternatives' & zero repetitions, neither of which can ever match, but the wider pattern can, e.g. @(()|x{0}|y)@. - - * By removing 'RepeatablePattern' from 'Match', it can be isolated in a new module. - This would result in a significant loss of discoverability. - - * Expand repeated 'Anchor.Bow' with @fewest - 1@ null matches followed by recursive 'findMatch'-call with @repetitions = 1@. --} - -module RegExDot.RegEx( --- * Type-classes - ShortcutExpander(..), --- * Types --- ** Type-synonyms --- BasicRegEx, - Concatenation, - ExternalMatch, - InputData, - MatchedData, - MatchList, --- MetaDataList, - RepeatablePattern, --- Transformation, --- ** Data-types - Alternatives(..), - Match, - ExtendedRegEx(..), - Pattern(..), - Result, --- * Constants - alternativeExtendedRegExSeparatorToken, --- anyDatum, - captureGroupDelimiters, - tokens, --- * Functions --- accumulateDistinctInputData, --- accumulateDistinctMetaDataFrom, - dock, --- drift, --- findMatch, - captureGroup, --- mkNullMatchFromConcatenation, --- mkNullMatchFromExtendedRegEx, --- mkNullMatchFromRepeatablePattern, --- safeReciprocal, --- shiftMatchedData, --- shiftMatch, - shiftMatchList, - showsMaybeAnchor, - simply, - transformExtendedRegEx, --- ** Accessors (Deconstructors) - deconstructAlternatives, --- getInputData, --- ** Operators - (+~), - (=~), - (/~), - (.*), - (.*?), --- (/+) --- ** Predicates --- hasBowAnchor, --- hasSternAnchor, - isDefined, - isCaptureGroup, - isSingletonAlternatives, --- isUnconsumableByAnyOf, --- ** Query --- getDistinctMetaDataFromAlternatives, --- getDistinctMetaDataFromConcatenation, --- getDistinctMetaDataFromExtendedRegEx, --- getDistinctMetaDataFromPattern, --- getDistinctMetaDataFromRepeatablePattern, - externalMatchLength, - extractDataFromMatch, - extractDataFromMatch', - extractDataFromMatchList -) where - -import Control.Arrow((&&&)) -import qualified Control.Arrow -import qualified Control.DeepSeq -import qualified Control.Parallel.Strategies -import qualified Data.Char -import qualified Data.Foldable -import qualified Data.List -import qualified Data.Maybe -import qualified Data.Ord -import qualified RegExDot.Anchor as Anchor -import qualified RegExDot.Consumer as Consumer -import qualified RegExDot.ConsumptionBounds as ConsumptionBounds -import qualified RegExDot.ConsumptionProfile as ConsumptionProfile -import qualified RegExDot.ExecutionOptions as ExecutionOptions -import qualified RegExDot.Meta as Meta -import qualified RegExDot.RegExOpts as RegExOpts -import qualified RegExDot.Repeatable as Repeatable -import RegExDot.Repeatable((^#->)) -import qualified RegExDot.Tree as Tree -import qualified ToolShed.Data.List -import qualified ToolShed.Data.List.Splits -import qualified ToolShed.SelfValidate - -infix 4 +~, =~, /~ -- Same as (==) & (/=). - --- | The type of a /regex/, in which there's no provision for either 'Alternatives' or 'Anchor.Anchor's. -type BasicRegEx m = [Repeatable.Repeatable (Meta.Meta m)] - -{- | - * Defines the method required to expand a mnemonic into an 'ExtendedRegEx'. - - * CAVEAT: this interface must be declared locally, since it references 'ExtendedRegEx', & 'ExtendedRegEx' references it. --} -class ShortcutExpander m where - expand :: Char -> ExtendedRegEx m -- ^ Expand a single-'Char' mnemonic into the corresponding 'ExtendedRegEx'. - -{- | - * Represents the /alternation/ feature of 'ExtendedRegEx's. - - * One could amalgamate this with 'Pattern', since it seems to exist merely as a peg to hang /instance/-declarations from. --} -newtype Alternatives m = MkAlternatives [ExtendedRegEx m] deriving Eq - --- | Accessor, to expose the guts. -deconstructAlternatives :: Alternatives m -> [ExtendedRegEx m] -deconstructAlternatives (MkAlternatives extendedRegExList) = extendedRegExList - --- | Similar to 'fmap', but operates on ['ExtendedRegEx'], rather than just @a@. -transformAlternatives - :: ([ExtendedRegEx m] -> [ExtendedRegEx m]) -- ^ The transformation function to apply to the list of data behind the constructor. - -> Alternatives m - -> Alternatives m -transformAlternatives f = MkAlternatives . f . deconstructAlternatives - -instance ( - ShortcutExpander m, - Meta.ShortcutExpander m, - Eq m, - Read m - ) => Read (Alternatives m) where - readsPrec _ (' ' : s) = reads s -- Consume white-space. - readsPrec _ ('\t' : s) = reads s -- Consume white-space. - readsPrec _ s = case reads s of - [(extendedRegEx, s1)] -> case dropWhile Data.Char.isSpace s1 of - ('|' : s2) -> Control.Arrow.first (transformAlternatives (extendedRegEx :)) `map` reads s2 {-singleton-} - _ -> [(MkAlternatives [extendedRegEx], s1)] - _ -> [] -- No parse. - -instance Show m => Show (Alternatives m) where - showsPrec _ = foldl (.) (showString "") . Data.List.intersperse (showChar alternativeExtendedRegExSeparatorToken) . map shows . deconstructAlternatives -- Replace the default list-format, with 'egrep'-syntax. - -instance Consumer.Consumer (Alternatives m) where - consumptionProfile = Consumer.aggregateConsumptionProfilesFromAlternatives . deconstructAlternatives - starHeight = Consumer.starHeight . deconstructAlternatives -- Must evaluate all Alternatives to determine best. - -instance ToolShed.SelfValidate.SelfValidator (Alternatives m) where - getErrors = ToolShed.SelfValidate.getErrors . deconstructAlternatives - -instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (Alternatives m) where - rnf = Control.DeepSeq.rnf . deconstructAlternatives - --- | 'Alternatives' can be employed as a simple /capture-group/ as well as a switch, under which circumstances there's no choice amongst multiple 'Alternatives'. -isSingletonAlternatives :: Alternatives m -> Bool -isSingletonAlternatives = (== 1) . length . deconstructAlternatives - --- | A set of 'Meta.Meta' which can be consumed. -type MetaDataList m = [Meta.Meta m] - -{- | - * 'True' if there's no possibility, that the specified input datum, can be consumed by any of the specified 'MetaDataList'. - - * Because data may not be consumed, perhaps because of unsuitable 'ConsumptionBounds.ConsumptionBounds', the converse doesn't hold, - i.e. @not isUnconsumableByAnyOf@ /doesn't/ imply that the input datum can be consumed, just that it hasn't been proven that it can't. --} -isUnconsumableByAnyOf :: Eq m - => m -- ^ The input datum. - -> MetaDataList m -- ^ The list of meta-data against any of which a match can occur. - -> Bool -isUnconsumableByAnyOf i = not . any (Meta.isMatch i) - --- | The set of distinct 'Meta.Meta', in the specified 'Alternatives'. -getDistinctMetaDataFromAlternatives :: Eq m => Alternatives m -> MetaDataList m -getDistinctMetaDataFromAlternatives = foldr (Data.List.union . getDistinctMetaDataFromExtendedRegEx) [] . deconstructAlternatives - --- | Defines either a simple 'Meta.Meta', which can match exactly one datum, or a set of 'Alternatives', each of which is recursively defined above, as an 'ExtendedRegEx'. -data Pattern m = - Require (Meta.Meta m) -- ^ Describes a requirement for a simple scalar datum of the polymorphic type. - | CaptureGroup (Alternatives m) -- ^ A sub-expression containing a selection of recursively defined alternatives, thus forming a tree-structure. - deriving Eq - -instance ( - Eq m, - Meta.ShortcutExpander m, - Read m, - ShortcutExpander m - ) => Read (Pattern m) where - readsPrec _ (' ' : s) = reads s -- Consume white-space. - readsPrec _ ('\t' : s) = reads s -- Consume white-space. - readsPrec _ ('(' : s) = case {-Alternatives.-} reads s of - [(alternatives, s1)] -> case dropWhile Data.Char.isSpace s1 of - (')' : s2) -> [(CaptureGroup alternatives, s2)] - _ -> [] -- No parse. - _ -> [] -- No parse. - readsPrec _ s = case reads s of - [pair] -> [Control.Arrow.first Require pair] - _ -> [] -- No parse. - -instance Show m => Show (Pattern m) where - showsPrec _ (Require meta) = shows meta - showsPrec _ (CaptureGroup alternatives) = showChar (fst captureGroupDelimiters) . shows alternatives . showChar (snd captureGroupDelimiters) - -instance Consumer.Consumer (Pattern m) where - consumptionProfile (Require meta) = Consumer.consumptionProfile meta - consumptionProfile (CaptureGroup alternatives) = Consumer.consumptionProfile alternatives - - starHeight (Require meta) = Consumer.starHeight meta - starHeight (CaptureGroup alternatives) = Consumer.starHeight alternatives - -instance ToolShed.SelfValidate.SelfValidator (Pattern m) where - getErrors (Require meta) = ToolShed.SelfValidate.getErrors meta - getErrors (CaptureGroup alternatives) = ToolShed.SelfValidate.getErrors alternatives - -instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (Pattern m) where - rnf (Require meta) = Control.DeepSeq.rnf meta - rnf (CaptureGroup alternatives) = Control.DeepSeq.rnf alternatives - --- | Convenience-function to build a 'CaptureGroup' from a list of alternative 'ExtendedRegEx's. -captureGroup :: [ExtendedRegEx m] -> Pattern m -captureGroup = CaptureGroup . MkAlternatives - --- | True if the 'Pattern' was constructed via 'CaptureGroup'. -isCaptureGroup :: Pattern m -> Bool -isCaptureGroup (CaptureGroup _) = True -isCaptureGroup _ = False - --- | The set of distinct 'Meta.Meta', in the specified 'Pattern'. -getDistinctMetaDataFromPattern :: Eq m => Pattern m -> MetaDataList m -getDistinctMetaDataFromPattern pattern = case pattern of - Require meta -> [meta] - CaptureGroup alternatives -> getDistinctMetaDataFromAlternatives alternatives - --- | Constant pattern, representing a lax 'Require'ment. -anyDatum :: Pattern a -anyDatum = Require Meta.Any - --- | Make 'Pattern's, 'Repeatable.Repeatable'. -type RepeatablePattern m = Repeatable.Repeatable (Pattern m) - --- | Construct a null 'Match' by assuming that the specified 'RepeatablePattern' consumes zero 'InputData'; of which the caller should ensure it's capable. -mkNullMatchFromRepeatablePattern - :: ConsumptionBounds.DataLength -- ^ The offset into the list of input data at which the zero-length match occurred. - -> RepeatablePattern m -- ^ The pattern that the input data matched. - -> Match m -- ^ The resulting match-structure. -mkNullMatchFromRepeatablePattern offset r = case Repeatable.base r of - Require _ -> Tree.Leaf (r, offset, []) - CaptureGroup _ -> Tree.Node $ Repeatable.getFewest r `replicate` [] - --- | The set of distinct 'Meta.Meta', in the specified 'RepeatablePattern'. -getDistinctMetaDataFromRepeatablePattern :: Eq m => RepeatablePattern m -> MetaDataList m -getDistinctMetaDataFromRepeatablePattern = getDistinctMetaDataFromPattern . Repeatable.base - -{- | - * Represents a black hole, which will greedily consume all data. - - * CAVEAT: nullary, i.e. a constant. --} -(.*) :: RepeatablePattern m -(.*) = Repeatable.zeroOrMore anyDatum - --- | A /non-greedy/ version of '.*'. -(.*?) :: RepeatablePattern m -(.*?) = Repeatable.zeroOrMore' anyDatum - --- | Convenience-function, to build a 'RepeatablePattern' from an unrepeated instance of the specified 'Meta.Meta'-datum. -simply :: Meta.Meta m -> RepeatablePattern m -simply = Repeatable.one . Require - --- | Represents the /concatenation/ aspect of 'ExtendedRegEx's. -type Concatenation m = [RepeatablePattern m] - --- | Construct a null 'Match' by assuming that the specified 'Concatenation' consumes zero 'InputData'. -mkNullMatchFromConcatenation - :: ConsumptionBounds.DataLength -- ^ The offset into the list of input data at which the zero-length match occurred. - -> Concatenation m -- ^ The list of repeatable patterns that the input data matched. - -> MatchList m -- ^ The resulting match-structure. -mkNullMatchFromConcatenation offset = map $ mkNullMatchFromRepeatablePattern offset - --- | The set of distinct 'Meta.Meta', in the specified 'Concatenation'. -getDistinctMetaDataFromConcatenation :: Eq m => Concatenation m -> MetaDataList m -getDistinctMetaDataFromConcatenation = foldr (Data.List.union . getDistinctMetaDataFromRepeatablePattern) [] - --- | The accumulating set of 'MetaDataList', resulting from each successive 'RepeatablePattern' in the specified 'Concatenation'. -accumulateDistinctMetaDataFrom :: Eq m => MetaDataList m -> Concatenation m -> [MetaDataList m] -accumulateDistinctMetaDataFrom = scanr (Data.List.union . getDistinctMetaDataFromRepeatablePattern) - --- | Constructs an 'ExtendedRegEx', by surrounding a 'Concatenation' with optional 'Anchor.Anchor's. -data ExtendedRegEx m = MkExtendedRegEx { - bowAnchor :: Maybe Anchor.Anchor, -- ^ An option to anchor the /regex/ to the start of the 'InputData'. - concatenation :: Concatenation m, -- ^ The sequence of 'RepeatablePattern's defining the 'Require'ments that the 'InputData' must meet. - sternAnchor :: Maybe Anchor.Anchor -- ^ An option to anchor the /regex/ to the end of the 'InputData'. -} deriving Eq - --- | The type of a function which transforms an 'ExtendedRegEx'. -type Transformation m = ExtendedRegEx m -> ExtendedRegEx m - --- | Similar to 'fmap', but operates on 'Concatenation', rather than just @a@. -transformExtendedRegEx - :: (Concatenation m -> Concatenation m) -- ^ The function used to transform the data behind the constructor. - -> Transformation m -transformExtendedRegEx f extendedRegEx = extendedRegEx { concatenation = f $ concatenation extendedRegEx } - -instance ( - Eq m, - Meta.ShortcutExpander m, - Read m, - ShortcutExpander m - ) => Read (ExtendedRegEx m) where - readsPrec _ [] = [] - readsPrec _ (' ' : s) = reads s -- Consume white-space. - readsPrec _ ('\t' : s) = reads s -- Consume white-space. - readsPrec _ ('\\' : c : s) = [(expand c, s)] - readsPrec _ s = case singleton of - [(extendedRegEx, _)] - | ToolShed.SelfValidate.isValid extendedRegEx -> singleton - | otherwise -> error $ ToolShed.SelfValidate.getFirstError extendedRegEx -- Parsed OK, but invalid. - _ -> [] -- No parse. - where --- singleton :: (Eq m, Meta.ShortcutExpander m, Read m, ShortcutExpander m) => [(ExtendedRegEx m, String)] - singleton = [ - ( - MkExtendedRegEx { - bowAnchor = maybeBowAnchor, - concatenation = concatenation', - sternAnchor = maybeSternAnchor - }, - remainder - ) | - (maybeBowAnchor, s1) <- readsMaybeAnchor s, - (concatenation', s2) <- reads s1, - (maybeSternAnchor, remainder) <- readsMaybeAnchor s2 - ] {-list-comprehension-} where - readsMaybeAnchor :: ReadS (Maybe Anchor.Anchor) - readsMaybeAnchor s' = return {-to List-monad-} $ case reads s' of - [pair] -> Control.Arrow.first Just pair - _ -> (Nothing, s') - --- | Shows either the specified 'Anchor.Anchor', or a null string where 'Nothing' is specified. -showsMaybeAnchor :: Maybe Anchor.Anchor -> String -> String -showsMaybeAnchor maybeAnchor - | Data.Maybe.isJust maybeAnchor = shows $ Data.Maybe.fromJust maybeAnchor - | otherwise = id - -instance Show m => Show (ExtendedRegEx m) where - showsPrec _ MkExtendedRegEx { - bowAnchor = maybeBowAnchor, -- CAVEAT: this could be 'Nothing' or perversely an 'Anchor.Stern'. - concatenation = concatenation', - sternAnchor = maybeSternAnchor -- CAVEAT: this could be 'Nothing' or perversely an 'Anchor.Bow'. - } = showsMaybeAnchor maybeBowAnchor . shows concatenation' . showsMaybeAnchor maybeSternAnchor - -instance Consumer.Consumer (ExtendedRegEx m) where - consumptionProfile = Consumer.aggregateConsumptionProfilesFromConcatenation . concatenation - starHeight = Consumer.starHeight . concatenation - -instance ToolShed.SelfValidate.SelfValidator (ExtendedRegEx m) where - getErrors = ToolShed.SelfValidate.getErrors . concatenation - -instance Control.DeepSeq.NFData m => Control.DeepSeq.NFData (ExtendedRegEx m) where - rnf MkExtendedRegEx { - bowAnchor = maybeBowAnchor, - concatenation = concatenation', - sternAnchor = maybeSternAnchor - } = Control.DeepSeq.rnf (maybeBowAnchor, concatenation', maybeSternAnchor) - --- | Drop 'Anchor.Anchor's at both bow & stern of the specified 'ExtendedRegEx'. -dock :: Transformation m -dock e = e { - bowAnchor = Just Anchor.Bow, - sternAnchor = Just Anchor.Stern -} - --- | Amend the 'Concatenation' in the specified 'ExtendedRegEx', by prepending '.*' where there's a missing top-level 'Anchor.Bow' & appending '.*?' where there's a missing top-level 'Anchor.Stern'. -drift :: Transformation m -drift extendedRegEx = transformExtendedRegEx ( - ( - if hasBowAnchor extendedRegEx then id else {-no anchor-} ((.*?) :) - ) . ( - if hasSternAnchor extendedRegEx then id else {-no anchor-} (++ [(.*)]) - ) - ) extendedRegEx - --- | True if the specified 'ExtendedRegEx' has an initial 'Anchor.Bow'. CAVEAT: though typically the alternative is no 'Anchor.Anchor', there may perversely be an initial 'Anchor.Stern', but neither qualify. -hasBowAnchor :: ExtendedRegEx m -> Bool -hasBowAnchor = (== Just Anchor.Bow) . bowAnchor - --- | True if the specified 'ExtendedRegEx' has a terminal 'Anchor.Stern'. CAVEAT: though typically the alternative is no 'Anchor.Anchor', there may perversely be a terminal 'Anchor.Bow', but neither qualify. -hasSternAnchor :: ExtendedRegEx m -> Bool -hasSternAnchor = (== Just Anchor.Stern) . sternAnchor - --- | True if there's at least one 'RepeatablePattern' in the 'Concatenation', i.e. that it's non-null. -isDefined :: ExtendedRegEx m -> Bool -isDefined = not . null . concatenation - --- | Construct a null 'Match' by assuming that the specified 'ExtendedRegEx' consumes zero 'InputData'. -mkNullMatchFromExtendedRegEx :: ConsumptionBounds.DataLength -> ExtendedRegEx m -> MatchList m -mkNullMatchFromExtendedRegEx offset = mkNullMatchFromConcatenation offset . concatenation - --- | The set of distinct (i.e. unique) 'Meta.Meta', in the specified 'ExtendedRegEx'. -getDistinctMetaDataFromExtendedRegEx :: Eq m => ExtendedRegEx m -> MetaDataList m -getDistinctMetaDataFromExtendedRegEx = getDistinctMetaDataFromConcatenation . concatenation - -{- | - * The input-data is just a list. - - * Whilst typically this list is also a 'String', & could therefore be more efficiently implemented using "Data.ByteString", - we can't assume that the polymorphic base-type is always 'Char'. --} -type InputData m = [m] - --- | The accumulating sets of distinct input data. -accumulateDistinctInputData :: Eq m => InputData m -> [InputData m] -accumulateDistinctInputData = scanr (Data.List.union . return) [] - --- | Tag the 'InputData' with the 'RepeatablePattern' it matched (which unfortunately confines the definition to this (bloated) module), & the offset from the start of the data; -type MatchedData m = (RepeatablePattern m, ConsumptionBounds.DataLength, InputData m) - --- | Shifts the offset of the specified 'MatchedData'. -shiftMatchedData - :: ConsumptionBounds.DataLength -- ^ The offset by which to shift the position into the input-data at which a match occurred. - -> MatchedData m -- ^ The match-structure whose offset is to be shifted. - -> MatchedData m -shiftMatchedData i (r, offset, d) = (r, offset + i, d) - --- | Accessor. -getInputData :: MatchedData m -> InputData m -getInputData (_, _, inputData) = inputData - --- | Describes the manner in which a 'RepeatablePattern' successfully consumed 'InputData'. -type Match m = Tree.Tree (MatchedData m) - --- | Shifts the offsets of all the 'MatchedData' contained in the specified 'Match'. -shiftMatch - :: ConsumptionBounds.DataLength -- ^ The offset by which to shift the position into the input-data at which a match occurred. - -> Match m -- ^ The match-structure whose offset is to be shifted. - -> Match m -shiftMatch i = (shiftMatchedData i `fmap`) - --- | Extract & concatenate, the 'InputData' from a 'Match'. -extractDataFromMatch :: Match m -> InputData m -extractDataFromMatch = Data.Foldable.foldMap getInputData -- Uses the List-monoid's associative binary operator (++), to concatenate the values returned by the specified function. - --- | Extract & concatenate, the 'InputData' from a 'Match'; null if it didn't match any. -extractDataFromMatch' :: Maybe (Match m) -> InputData m -extractDataFromMatch' (Just match) = extractDataFromMatch match -extractDataFromMatch' _ = [] - --- | Describes the manner in which a 'Concatenation' successfully consumed 'InputData'. -type MatchList m = [Match m] - --- | Shifts the offsets of all the 'MatchedData' contained in the specified 'MatchList'. -shiftMatchList - :: ConsumptionBounds.DataLength -- ^ The offset by which to shift the position into the input-data at which a each listed match occurred. - -> MatchList m -- ^ The list of match-structures, each of whose offsets are to be shifted. - -> MatchList m -shiftMatchList i = map (shiftMatch i) - --- | Extract & concatenate, the 'InputData', from the 'MatchList'. -extractDataFromMatchList :: MatchList m -> InputData m -extractDataFromMatchList = concatMap extractDataFromMatch -- CAVEAT: too fine-grain for effective data-parallelism. - --- | At the top-level of an 'ExtendedRegEx', the lack of an 'Anchor.Anchor' allows the 'ExtendedRegEx' to drift away from the corresponding end of the input-data; this data-gap is captured here. -type ExternalMatch m = Maybe (Match m) - --- | Captures the list of input-data consumed by the 'Concatenation', bracketed by any data-prefix or data-suffix. -type Result m = (ExternalMatch m, Maybe (MatchList m), ExternalMatch m) - --- | Returns the length of data consumed by the specified 'ExternalMatch'. -externalMatchLength :: ExternalMatch m -> ConsumptionBounds.DataLength -externalMatchLength = length . extractDataFromMatch' - -{- | - * Similar to '+~', but exposes 'ExecutionOptions.ExecutionOptions', permitting greater control; & enough rope to hang yourself. - - * The parameter-order facilitates partial application, to acquire a matcher-function for a specific 'ExtendedRegEx', which can be applied repeatedly to different 'InputData'. - - * Performs some one-off @O(n)@ time-complexity preparation, then delegates the grunt-work to a private recursive function. - - * One could waste a lot of time trying to consume 'InputData' from the head of a high-'Consumer.starHeight' 'ExtendedRegEx', - only to find that there's a low-'Consumer.starHeight' tail that can't ever match the 'InputData'; checks this first before delegating. --} -findMatch :: (Eq m, Control.DeepSeq.NFData m) - => RegExOpts.RegExOpts (ExtendedRegEx m) -- ^ The match-options parameterised by the regex against which to match the input data. - -> InputData m -- ^ The input data within which to locate a match. - -> Maybe (MatchList m) -findMatch regExOpts@RegExOpts.MkRegExOpts { - RegExOpts.executionOptions = executionOptions, - RegExOpts.regEx = extendedRegEx -} originalInputData - | let - extractLowStarHeightTail :: ExtendedRegEx m -> Concatenation m - extractLowStarHeightTail = {-#SCC "extractLowStarHeightTail" #-} fromConcatenation . reverse . concatenation where - fromConcatenation :: Concatenation m -> Concatenation m - fromConcatenation [] = [] - fromConcatenation (repeatablePattern : concatenation') - | Consumer.starHeight repeatablePattern == 0 = repeatablePattern : fromConcatenation {-recurse-} concatenation' - | Repeatable.getFewest repeatablePattern > 0 = case Repeatable.base repeatablePattern of - CaptureGroup alternatives - | isSingletonAlternatives alternatives -> extractLowStarHeightTail {-recurse-} . head $ deconstructAlternatives alternatives - | otherwise -> {-non-singleton-} [] - _ -> [] -- Zero 'Consumer.StarHeight' would have been detected by 'fromConcatenation'. - | otherwise = [] - --- lowStarHeightTail :: Concatenation a - lowStarHeightTail = extractLowStarHeightTail extendedRegEx - in and [ - ExecutionOptions.checkExistenceOfInelasticTail executionOptions, - not $ null lowStarHeightTail, -- Prevent infinite recursion. - reverse originalInputData /~ RegExOpts.mkRegEx MkExtendedRegEx { - bowAnchor = Just Anchor.Bow, - concatenation = lowStarHeightTail, - sternAnchor = Nothing - } -- Check for mismatch with the corresponding tail of 'InputData'. - ] = Nothing - | otherwise = findMatchSlave originalConcatenation ( - Consumer.accumulateConsumptionProfiles originalConcatenation - ) ( - accumulateDistinctMetaDataFrom [] originalConcatenation - ) originalInputData originalInputDataLength ( - accumulateDistinctInputData originalInputData - ) - where --- originalConcatenation :: Concatenation a - originalConcatenation = concatenation extendedRegEx - - originalInputDataLength :: ConsumptionBounds.DataLength - originalInputDataLength = length originalInputData -{- - This beast takes a lot of parameters. - Whilst the only strictly necessary parameters are the lists 'Concatenation' & 'InputData', many ancillary lists derived from them, are also passed by parameter. - These are proportionally reduced on recursion, to avoid the requirement to regenerate them each time they're required. --} - findMatchSlave :: (Eq m, Control.DeepSeq.NFData m) - => Concatenation m -- The list of 'RepeatablePattern's from which the regex is constructed. - -> ConsumptionProfile.AccumulatedConsumptionProfiles -- The capacity-bounds for the consumption of 'InputData', that the 'Concatenation' extending right from any given 'RepeatablePattern', can consume. - -> [MetaDataList m] -- The set of distinct 'Meta.Meta'-data, in the 'Concatenation extending right from any given 'RepeatablePattern'. - -> InputData m -- The input data, which will be fed to the 'Concatenation'. - -> ConsumptionBounds.DataLength -- The length of the previously specified 'InputData'. - -> [InputData m] -- The set of distinct input data, extending right from any given point. - -> Maybe (MatchList m) - findMatchSlave [] _ _ [] _ _ = Just [] -- Simultaneous exhaustion of the 'Concatenation' of 'RepeatablePattern's & the 'InputData' => success. - findMatchSlave [] _ _ _ _ _ = Nothing - findMatchSlave concatenation'@(repeatablePatternHead : concatenationTail) (accumulatedConsumptionProfileHead : accumulatedConsumptionProfilesTail) (distinctMetaDataHead : distinctMetaDataTail) inputData inputDataLength distinctInputData@(distinctInputDataHead : _) - | not $ inputDataLength `ConsumptionProfile.withinConsumptionBounds` accumulatedConsumptionProfileHead = Nothing - | null inputData = Just $ mkNullMatchFromConcatenation inputDataOffset concatenation' -- Build a 'Match', of null 'InputData'. - | and [ - ExecutionOptions.bypassInputDataForLiberalConsumer executionOptions, -- We may be able to establish success, without evaluating 'inputData' any further than required to determine its length. - not $ ExecutionOptions.requireMatchList executionOptions, -- Otherwise the precise mapping of the 'inputData' to 'RepeatablePattern's must be determined. - not $ ConsumptionProfile.hasSpecificRequirement accumulatedConsumptionProfileHead -- Otherwise a match for the specific 'Meta'-data must be found. - ] = Just undefined {-shouldn't be evaluated according to 'ExecutionOptions.requireMatchList'-} - | and [ - ExecutionOptions.checkForUnconsumableData executionOptions, - not $ ConsumptionProfile.canConsumeAnything accumulatedConsumptionProfileHead, -- Otherwise the subsequent test will always fail. - (`isUnconsumableByAnyOf` distinctMetaDataHead) `any` distinctInputDataHead -- Occasionally failure is both inevitable & obvious. - ] = Nothing - | otherwise = {-#SCC "findMatchSlave" #-} let - tailConsumptionProfile :: ConsumptionProfile.ConsumptionProfile - tailConsumptionProfile@( - ConsumptionProfile.MkConsumptionProfile { - ConsumptionProfile.consumptionBounds = (minConsumptionConcatenationTail, maybeMaxConsumptionConcatenationTail) - } - ) = head accumulatedConsumptionProfilesTail -- Extract the aggregate consumption-profile, of the tail of the 'Concatenation'. - - maxDataAvailable :: ConsumptionBounds.DataLength - maxDataAvailable = inputDataLength - minConsumptionConcatenationTail -- The maximum data available to match 'repeatablePatternHead'. - in if maxDataAvailable < 0 - then Nothing - else {-inputData is sufficient for concatenationTail-} let - Repeatable.MkRepeatable { - Repeatable.base = base, - Repeatable.repetitionBounds = (fewest, most), - Repeatable.isGreedy = isGreedy - } = repeatablePatternHead -- Completely deconstruct the 'Repeatable' at the head of the 'Concatenation'. -{- - Find the maximum sequence of inputData, preceding the first of the mandatory requirements of 'concatenationTail'. - This enables one to more tightly constrain the maximum number of repetitions of a 'RepeatablePattern', - which consequently reduces the subsequent requirement to back-track. --} --- maximumDataBeforePegs :: Maybe (InputData m) - maximumDataBeforePegs = {-#SCC "maximumDataBeforePegs" #-} reverse (extractMinimumRequirementFromConcatenation concatenationTail) `maximumDataAfterPegs` reverse inputData where -{- - Extract those 'RepeatablePattern's which must match at least once. - Singleton 'Alternatives' are flattened by replication, & non-singleton 'Alternatives' removed, allowing a reduction of the list's base-type to just 'Repeatable.Repeatable' 'Meta.Meta'. - The resulting 'BasicRegEx', with /Kleene Star/s weeded-out & 'Alternatives' pruned, can be matched in O(n)-time, to verify minimal requirements. --} - extractMinimumRequirementFromConcatenation :: Concatenation m -> BasicRegEx m - extractMinimumRequirementFromConcatenation = concatMap ( - \repeatablePattern -> case Repeatable.base repeatablePattern of - Require datum -> return {-to List-monad-} repeatablePattern { Repeatable.base = datum } - CaptureGroup a - | isSingletonAlternatives a -> concat . replicate (Repeatable.getFewest repeatablePattern) . extractMinimumRequirementFromConcatenation {-recurse-} . concatenation . head $ deconstructAlternatives a - | otherwise -> [] - ) . filter ((> 0) . Repeatable.getFewest) - - maximumDataAfterPegs :: Eq m => BasicRegEx m -> InputData m -> Maybe (InputData m) - maximumDataAfterPegs pegList@( - Repeatable.MkRepeatable { - Repeatable.base = metaPeg, - Repeatable.repetitionBounds = (fewestPeg, _) - } : pegListTail - ) inputData' - | null remainingInputData && length candidateMatchedInputData < fewestPeg = Nothing -- Insufficient data to ever match. - | fewestPeg == 1 || (`Meta.isMatch` metaPeg) `all` tail candidateMatchedInputData = pegListTail `maximumDataAfterPegs` remainingInputData -- Success => recurse. - | otherwise = pegList `maximumDataAfterPegs` tail candidateInputData -- Failed candidate => recurse. - where --- candidateInputData, candidateMatchedInputData, remainingInputData :: InputData a - candidateInputData = dropWhile (not . (`Meta.isMatch` metaPeg)) inputData' - (candidateMatchedInputData, remainingInputData) = fewestPeg `splitAt` candidateInputData - maximumDataAfterPegs [] inputData' = Just inputData' -- All the pegs have been matched. -{- - Each 'repeatablePatternHead' can either be a 'Require' or a 'CaptureGroup'. - The former case needs further reduction & we can proceed to construct a 'Tree.Leaf' from any matching 'inputData'. - We attempt to reduce the latter case to an instance of the former, by expanding all combinations of repetitions of various 'Alternatives', constructing a new 'ExtendedRegEx' from each, & recursing. --} - in case base of - Require meta - | null maybeMatchList -> Nothing - | otherwise -> head {-most suitable-} maybeMatchList - where -{- - Attempt to narrow the repetition-range, from the potentially unbounded specification, to that which is possible according to, - the 'inputDataLength', & the specified consumption-bounds of the remainder of the /regex/. - Once the repetition-range has been narrowed, iterate from the appropriate end of the range (depending on 'isGreedy'), extract the corresponding length from inputData, & check for a match. --} --- maybeMatchList :: [Maybe (MatchList m)] - maybeMatchList = {-#SCC "maybeMatchList" #-} dropWhile Data.Maybe.isNothing . map ( - \(consumedInputData, unconsumedInputData) -> let - consumedInputDataLength :: ConsumptionBounds.DataLength - consumedInputDataLength = length consumedInputData - in ( - Tree.Leaf (repeatablePatternHead, inputDataOffset, consumedInputData) : - ) `fmap` {-apply to Maybe Functor-} findMatchSlave concatenationTail accumulatedConsumptionProfilesTail distinctMetaDataTail unconsumedInputData ( - inputDataLength - consumedInputDataLength - ) ( - consumedInputDataLength `drop` distinctInputData - ) -- Recurse, to check whether the 'unconsumedInputData' tail also matches. - ) $ let - fewestData, mostData, mostData' :: Repeatable.Repetitions - fewestData = {-#SCC "fewestData" #-} case maybeMaxConsumptionConcatenationTail of - Just m -> (inputDataLength - m) {- :: ConsumptionBounds.DataLength-} `max` fewest {- :: Repeatable.Repetitions-} -- CAVEAT: conceptually different types. - _ -> fewest - --- mostData = length . takeWhile (~= meta) $ (`take` inputData) maxData -- As slow as it is concise. - mostData = {-#SCC "mostData" #-} maxData - measureUnmatchableTail inputData maxData where - maxData :: ConsumptionBounds.DataLength - maxData = case most of - Just cap -> cap {- :: Repeatable.Repetitions-} `min` maxDataAvailable {- :: ConsumptionBounds.DataLength-} -- CAVEAT: conceptually different types. - _ -> maxDataAvailable - --- measureUnmatchableTail :: InputData m -> ConsumptionBounds.DataLength -> ConsumptionBounds.DataLength - measureUnmatchableTail _ 0 = 0 -- Have matched all the data. - measureUnmatchableTail (a : as) unmatched - | a `Meta.isMatch` meta = measureUnmatchableTail as $ pred unmatched -- Recurse. - | otherwise = unmatched -- Return the unmatchable tail-length. - measureUnmatchableTail [] _ = error "RegExDot.RegEx.findMatch.findMatchSlave.maybeMatchList.mostData.measureUnmatchableTail:\tdata unexpectedly exhausted." -{- - 'most' has been extracted from 'repeatablePatternHead', - & reduced according to the length of 'inputData' which actually matches, to form 'mostData', - but if it still exceeds 'fewestData', the resulting repetition-range may trigger backtracking. - So, attempt to further reduce this ill-defined repetition-range, according to the competing requirements of 'concatenationTail'. --} - mostData' - | and [ - ExecutionOptions.moderateGreed executionOptions, - isGreedy, -- Otherwise, since the search proceeds from 'fewest' to 'most', the optimal solution is located before backtracking. - mostData > fewestData -- Otherwise, there's no unbridled greed to moderate. - ] = case maximumDataBeforePegs of - Just maximumDataBeforePegs' -> mostData `min` length maximumDataBeforePegs' -- Cap upper bound. - _ -> negate 1 -- Guaranteed to be < 'fewestData'. - | otherwise {-no requirement for this optimisation-} = mostData - in ( - succ {-fence-post-} mostData' - fewestData -- CAVEAT: possibly <= 0. - ) `take` ( - if isGreedy - then ToolShed.Data.List.Splits.splitsLeftFrom mostData' - else ToolShed.Data.List.Splits.splitsRightFrom fewestData - ) inputData - - CaptureGroup alternatives - | and [ - not isGreedy, - fewest <= 0, -- Zero repetitions permissible. - Data.Maybe.isJust tailMatch - ] -> zeroRepetitions -- Zero repetitions is the optimal solution, rendering the choice of Alternative irrelevant. - | and [ - ExecutionOptions.catchIncompatibleAnchors executionOptions, - fewest > 1, -- Multiple repetitions. - minConsumptionAlternatives > 0, -- InputData required. - ($ extendedRegExAlternatives) `any` [(hasBowAnchor `all`), (hasSternAnchor `all`)] - ] -> Nothing -- 'extendedRegExFromAlternative' requires 'InputData', therefore 'Anchor.Bow' can only pass on the 1st repetition & 'Anchor.Stern' can only pass on the last. - | otherwise -> let -{- - We're about to try all repetitions of one Alternative before progressing to the next. - One could alternatively transpose this search-pattern, & investigate all 'Alternatives' permissible at n repetitions, before trying (n + 1); I've not pursued this option. - Either way, if the user requires data-capture, then for POSIX-compliance we must perform an exhaustive search of the O(Alternatives ^ Repetitions) permutations for the optimal solution; though the search may be narrowed as we proceed. --} --- matchPairList :: [(Match m, MatchList m)] - matchPairList = {-#SCC "matchPairList" #-} Data.Maybe.catMaybes . concat . Control.Parallel.Strategies.parMap Control.Parallel.Strategies.rseq {-Particularly effective when 'not ExecutionOptions.requireMatchList' -} ( --- :: ExtendedRegEx m -> [Maybe (Match m, MatchList m)] - \extendedRegExFromAlternative@MkExtendedRegEx { - concatenation = concatenationFromAlternative - } -> let - components :: Int - components = length concatenationFromAlternative - - consumptionProfileConcatenationFromAlternative :: [ConsumptionProfile.ConsumptionProfile] - consumptionProfileConcatenationFromAlternative = map Consumer.consumptionProfile concatenationFromAlternative - - minConsumptionConcatenationFromAlternative :: ConsumptionBounds.DataLength - maybeMaxConsumptionConcatenationFromAlternative :: Maybe ConsumptionBounds.DataLength - (minConsumptionConcatenationFromAlternative, maybeMaxConsumptionConcatenationFromAlternative) = ConsumptionProfile.consumptionBounds $ ConsumptionProfile.aggregateFromConcatenation consumptionProfileConcatenationFromAlternative - in ( - if ExecutionOptions.abortTrialRepetitionsOnZeroConsumption executionOptions && Data.Maybe.isNothing most {-constrained only by 'inputDataLength'-} - then {-#SCC "abortTrialRepetitionsOnZeroConsumption" #-} ToolShed.Data.List.takeUntil $ \maybeMatchPair -> case maybeMatchPair of - Just (match, _) -> ($ Tree.pop match) `all` [ - (>= fewest) . length, -- Must achieve 'Repeatable.getFewest', regardless of data-consumption. - null . extractDataFromMatchList . last -- If the n-th repetition consumed nothing, so will the (n + 1)-th. - ] - _ -> False -- Failure with n repetitions, doesn't preclude success with n + 1. - else {-optimisation not required-} id - ) . ( --- :: Maybe (Match m, MatchList m) -> Maybe (Match m, MatchList m) - \maybeMatchPairList -> if ExecutionOptions.abortTrialRepetitionsOnInherentFailure executionOptions -{- - We're iterating through the identified range of possible repetitions, of one possible 'concatenationFromAlternative'. - If failure can be proven, for an arbitrary number of repetitions, then we can bypass the futile remainder. - If the first trial succeeded, then clearly there's no such inherent problem with 'concatenationFromAlternative', & it's necessary to evaluate all subsequent trial-repetitions. - Otherwise, assuming there're subsequent trials to be bypassed, attempt to prove that failure is inevitable, by launching a pilot-study, - using a single repetition of 'concatenationFromAlternative' followed by a replacement infallible 'concatenationTail'. --} - then {-#SCC "abortTrialRepetitionsOnInherentFailure" #-} case maybeMatchPairList of - (Nothing {-failed attempt-} : _ {-subsequent attempt worth bypassing-} : _) -> if ( - minConsumptionConcatenationFromAlternative == 0 -- Can delegate consumption of unconsumable data to the infallible tail => unprovable culpability. - ) || inputData =~ RegExOpts.mkRegEx MkExtendedRegEx { - bowAnchor = Just Anchor.Bow, - concatenation = concatenationFromAlternative ++ [anyDatum ^#-> minConsumptionConcatenationTail], -- Infallible tail. - sternAnchor = Just Anchor.Stern - } - then maybeMatchPairList - else [] -- Abandon further repetitions of 'Alternatives'. - _ {-Alternative is OK, or there's no subsequent attempt to bypass-} -> maybeMatchPairList - else {-optimisation not required-} maybeMatchPairList - ) . map ( --- :: Repeatable.Repetitions -> Maybe (Match m, MatchList m) - \repetitions -> if hasSternAnchor extendedRegExFromAlternative - then if (repetitions == 1 || minConsumptionConcatenationFromAlternative == 0) && minConsumptionConcatenationTail == 0 - then ( - shiftMatch inputDataOffset . Tree.Node . ( - : replicate ( - pred repetitions -- Typically degenerates to zero, since anchored sub-expressions aren't normally repeatable. - ) ( - mkNullMatchFromExtendedRegEx 0 extendedRegExFromAlternative -- Other Alternatives may be available & suitable. - ) -- Repeated null match may, lacking other suitable Alternatives, correspond to this stern-anchored ExtendedRegEx, so data must be consumed before them. - ) &&& const (mkNullMatchFromConcatenation inputDataOffset concatenationTail) - ) `fmap` findMatch regExOpts { RegExOpts.regEx = extendedRegExFromAlternative } inputData -- Recurse, either consuming all inputData or failing. - else Nothing - else {-no Anchor.Stern-} let -{- - Expand the 'Repeatable CaptureGroup', so that after recursion it is ultimately reduced in complexity, to just a 'Concatenation' of 'Repeatable Require', & can be handled above as a POSIX BRE. - - There're two special cases; one repetition of n 'Alternatives' & n repetitions of one Alternative. - If there's only one repetition, we can reduce the implementation-logic substantially, to the point where the implementation's independent of whether there's a choice of 'Alternatives'. - If there only one Alternative, then the entire repetition-sequence can be immediately expanded, vastly reducing the required number of recursions, - otherwise we can only expand the first repetition, leaving the option of matching a different Alternative on subsequent repetitions. - - In all cases, we must subsequently re-compose any 'MatchList' resulting of the recursive call, into one corresponding to the structure of original ExtendedRegEx. - The expansion-process & corresponding subsequent collation, are associated in a tuple, in which the consumption-profile of the expansion is also contained. --} --- collater :: MatchList m -> (Match m, MatchList m) - consumptionProfileExpandedConcatenationPrefix :: [ConsumptionProfile.ConsumptionProfile] --- expandedConcatenationPrefix :: Concatenation a - (collater, consumptionProfileExpandedConcatenationPrefix, expandedConcatenationPrefix) - | repetitions == 1 = ( - Control.Arrow.first ( - Tree.Node . return {-to List-monad-} - ) . splitAt components, -- Bisect the 'MatchList', into the part resulting from the Alternative, & that for 'concatenationTail'. - consumptionProfileConcatenationFromAlternative, - concatenationFromAlternative - ) -- Both tuples below, can degenerate to this simple case; it isn't a fundamentally different algorithm. - | and [ - ExecutionOptions.unrollRepeatedSingletonAlternative executionOptions, - isSingletonAlternative, - not $ hasBowAnchor extendedRegExFromAlternative - ] = ( - Control.Arrow.first ( - Tree.Node . ToolShed.Data.List.chunk components - ) . splitAt ( - components * repetitions - ), -- Bisect the 'MatchList', into the part resulting from (Alternative){n}, & that for 'concatenationTail'. - concat $ replicate repetitions consumptionProfileConcatenationFromAlternative, - concat $ replicate repetitions concatenationFromAlternative -- Expand all repetitions of the single alternative. - ) - | otherwise {-choice of Alternatives or has Anchor.Bow-} = let - remainingRepetitions :: Repeatable.Repetitions - remainingRepetitions = pred repetitions -- Expand just the first repetition of the set of 'Alternatives'. - --- singletonRepeatable :: RepeatablePattern a - singletonRepeatable = Repeatable.toSingleton repeatablePatternHead - in ( - Control.Arrow.first ( - Tree.Node . uncurry (:) . Control.Arrow.second ( - concatMap Tree.pop -- Amalgamate the repeated singleton 'MatchList's. - ) . splitAt components -- Bisect first 'MatchList', into the part resulting from 'concatenationFromAlternative', & that for 'replicate remainingRepetitions'. - ) . splitAt ( - components + remainingRepetitions -- Bisect the 'MatchList', into the part resulting from (Alternatives){n}, & that for 'concatenationTail'. - ), - consumptionProfileConcatenationFromAlternative ++ replicate remainingRepetitions (Consumer.consumptionProfile singletonRepeatable), - concatenationFromAlternative {-expand 1st repetition-} ++ replicate remainingRepetitions {-potentially zero-} singletonRepeatable -- Enumerate all remaining repetitions of ANY Alternative; a different Alternative can match for each repetition. - ) - in {-#SCC "collater" #-} collater `fmap` {-apply to Maybe Functor-} findMatchSlave ( - expandedConcatenationPrefix ++ concatenationTail - ) ( - ConsumptionProfile.accumulateFrom tailConsumptionProfile {-initial value-} consumptionProfileExpandedConcatenationPrefix ++ tail accumulatedConsumptionProfilesTail - ) ( - accumulateDistinctMetaDataFrom (head distinctMetaDataTail) {-initial value-} expandedConcatenationPrefix ++ tail distinctMetaDataTail - ) inputData inputDataLength distinctInputData -- Recurse to get 'Maybe (MatchList a)'. - ) $ if Repeatable.isPrecise repeatablePatternHead -- These are frequently generated by the previous recursion, as a trial expansion of an Alternative. - then if ExecutionOptions.validateMinConsumptionOfAlternatives executionOptions && maxDataAvailable < fewest * minConsumptionAlternatives - then [] -- Failure is inevitable. - else [fewest] - else {-imprecise range-} let -{- - Attempt to narrow the repetition-range, from the potentially unbounded specification, to that which is possible according to (data available / consumption-requirement), - where there're bounds for both numerator & denominator. - Additionally, one must account for the possibility of the denominator being zero, & if so, select the optimal value if the numerator is also zero. - The range can be narrowed more aggressively, for the special case of a singleton 'Alternative', as used in a capture-group. --} - fewestAlternatives, mostAlternatives, mostAlternatives' :: Repeatable.Repetitions - fewestAlternatives = {-#SCC "fewestAlternatives" #-} 1 {-zero is handled later-} `max` if any Data.Maybe.isNothing [maybeMaxConsumptionConcatenationTail, maybeMaxConsumptionAlternatives] - then fewest - else {-no unlimited capacities-} let - minDataAvailable, maxConsumptionAlternatives :: ConsumptionBounds.DataLength - minDataAvailable = inputDataLength - Data.Maybe.fromJust maybeMaxConsumptionConcatenationTail -- CAVEAT: can be negative. - maxConsumptionAlternatives = Data.Maybe.fromJust maybeMaxConsumptionAlternatives - in if isSingletonAlternative - then if maxConsumptionAlternatives == 0 -- Denominator. - then {-sole Alternative can't consume anything ?!-} if minDataAvailable <= 0 -- Numerator. - then {-tail may consume all data-} fewest -- Repeat to meet the minimum requirement. - else {-unconsumable data-} maxBound -- Failure is inevitable. - else {-non-zero => can divide-} max fewest $ minDataAvailable /+ maxConsumptionAlternatives - else {-choice of Alternatives-} let - minDataAvailable' :: ConsumptionBounds.DataLength - minDataAvailable' = minDataAvailable - Data.Maybe.fromJust maybeMaxConsumptionConcatenationFromAlternative -- CAVEAT: potentially negative. - in if maxConsumptionAlternatives == 0 -- Denominator. - then {-no Alternative can consume anything-} if minDataAvailable' <= 0 -- Numerator. - then {-zero unconsumable data-} fewest -- Repeat to meet the minimum requirement. - else {-unconsumable data-} maxBound -- Failure is inevitable. - else {-non-zero => can divide-} max fewest . succ {-account for expanded instance-} $ minDataAvailable' /+ maxConsumptionAlternatives - - mostAlternatives = {-#SCC "mostAlternatives" #-} case most of - Just cap -> cap `min` mostPermissibleRepetitions -- Create a ceiling above which the calculated number of permissible repetitions can't rise. - _ -> mostPermissibleRepetitions - where - mostPermissibleRepetitions :: Repeatable.Repetitions - mostPermissibleRepetitions - | isSingletonAlternative = if minConsumptionAlternatives == 0 -- Denominator. - then {-sole Alternative can consume zero data-} if maxDataAvailable == 0 -- Numerator. - then {-zero data available-} fewest -- Repeat to meet the minimum requirement. - else {-data available-} maxDataAvailable -- Either each repetition consumes something, or (n - 1) repetitions is a better solution. - else {-non-zero => can divide-} maxDataAvailable `div` minConsumptionAlternatives -- Divide & round down. - | otherwise {-choice of Alternatives-} = 1 {-account for instance expanded as 'concatenationFromAlternative'-} + let - maxDataAvailable' :: ConsumptionBounds.DataLength - maxDataAvailable' = maxDataAvailable - minConsumptionConcatenationFromAlternative -- CAVEAT: potentially negative. - in if minConsumptionAlternatives == 0 -- Denominator. - then {-@ least one Alternative can consume zero-} if maxDataAvailable' == 0 -- Numerator. - then {-zero data available $-} pred fewestAlternatives -- Repeat as required. CAVEAT: an annoying dependency, which prevents parallel-evaluation. - else {-data available-} if and [ - False, -- The cost outweighs the small infrequent dividend. - minConsumptionConcatenationFromAlternative == 0, - maxDataAvailable > fewest - ] -- If 'fewest' can be met, without unconsuming repetitions of 'concatenationFromAlternative', then stop short of permitting it. - then pred maxDataAvailable -- Any greater, & one repetition (possibly that currently expanded) must needlessly consume nothing. - else maxDataAvailable' -- Either all n repetitions consumes something, or (n - 1) is a better solution. - else {-non-zero => can divide-} maxDataAvailable' `div` minConsumptionAlternatives -- Divide & round down. -{- - 'most' has been extracted from 'repeatablePatternHead', - & reduced according to the number of times it can fit into the maximum available data, to form 'mostAlternatives', - but if it still exceeds 'fewestAlternatives', the resulting repetition-range may trigger backtracking. - So, attempt to further reduce this ill-defined repetition-range, according to the competing requirements of 'concatenationTail'. - CAVEAT: this pays-off infrequently, but not necessarily insignificantly. --} - mostAlternatives' = {-#SCC "mostAlternatives'" #-} if and [ - ExecutionOptions.moderateGreed executionOptions, - mostAlternatives > fewestAlternatives, -- Otherwise, there's no unbridled greed to moderate. - minConsumptionAlternatives > 0 -- Otherwise, any number of repetitions can occur without necessarily triggering back-tracking. - ] - then case maximumDataBeforePegs of - Just maximumDataBeforePegs' -> mostAlternatives `min` ( - length maximumDataBeforePegs' `div` minConsumptionAlternatives - ) -- Cap upper bound. - _ -> negate 1 -- Guaranteed to be < 'fewestAlternatives'. - else {-no requirement for this optimisation-} mostAlternatives - in [fewestAlternatives .. mostAlternatives'] -- The permissible repetition-range of any Alternative, constrained by the amount of data available, after subtracting fixed consumption-requirements. - ) . ( - if ExecutionOptions.permitReorderingOfAlternatives executionOptions -{- - The order in which 'Alternatives' are evaluated, should make no difference to the final result; unless 'ExecutionOptions.useFirstMatchAmongAlternatives'. - This order can be manipulated in an attempt to more quickly locate any 'Just MatchList', though if 'ExecutionOptions.requireMatchList', an exhaustive search for the optimal solution must be performed anyway. - Tackling lower-complexity 'Alternatives' first makes sense, in the hope of finding any easy solution before becoming entrenched fighting some hideous monster. - Preferring 'Alternatives' which require more data also makes sense, regardless of 'Repeatable.isGreedy' which just a preference for a type of solution, because fewer repetitions are then permissible; - this is rather like trying to pack a box & preferring to try the biggest items first. --} - then Data.List.sortBy $ Data.Ord.comparing ( - Consumer.starHeight &&& safeReciprocal . (fromIntegral :: ConsumptionBounds.DataLength -> Rational) . Consumer.getFewest - ) -- Firstly increasing complexity, then decreasing minimum data-capacity, otherwise stable. - else id - ) $ filter ( - \e -> ( - not (hasBowAnchor e) || inputDataOffset == 0 -- Necessary & sufficient. - ) && ( - not (hasSternAnchor e) || minConsumptionConcatenationTail == 0 -- Necessary but insufficient, since though minimum consumption is zero, maximum isn't necessarily. - ) - ) extendedRegExAlternatives - in if null matchPairList - then if fewest <= 0 - then zeroRepetitions -- Which might still fail, depending on 'tailMatch'. - else {-zero repetitions isn't permissible-} Nothing - else {-at least one Alternative matched-} {-#SCC "selectAlternative" #-} Just . uncurry (:) {-re-join head & tail-} $ ( -{- - At least one 'Match' has been found amongst the combinations of repetitions of various 'Alternatives', - & if the caller reduces the 'Just MatchList' to 'True', that'd be sufficient, but otherwise one must identify the optimal 'MatchList'. - The primary selection-criterion used is the POSIX one, based on total maximum (or minimum for non-'Repeatable.isGreedy' capture-groups) data-consumption. - /Perl/'s "first-past-the-post" policy is available as an alternative. --} - if ExecutionOptions.useFirstMatchAmongAlternatives executionOptions || length matchPairList == 1 - then head - else snd {-remove prepended selection-criterion-} . Data.List.maximumBy {-select the best match-} ( - Data.Ord.comparing fst -- Compare using only the criterion, not the result from which it was derived. - ) . map ( - ( -{- - If the primary selection-criterion doesn't resolve the choice between candidate 'Match'es, I employ these ad-hoc criteria. - - * Fewer repetitions are preferred, which discourages the capture of null lists of 'InputData'. - - * Within the 'MatchedData', from which a candidate 'Match' is ultimately composed, consumption of 'InputData' beyond 'Repeatable.getFewest' by 'Repeatable.isGreedy' 'RepeatablePattern's, is preferred to non-greedy ones. - - * The consumption on each successive repetition is compared between candidate 'Match'es (which are now known to have used an equal number of repetitions); - this causes the data-consumption to flow towards earlier repetitions for greedy capture-groups, & towards later repetitions in non-greedy ones, - thus mimicking the behaviour of the unrolled repetition. - - * PS: more precise criteria are defined in . --} - \match -> let - dataLengthCriterion :: InputData m -> Rational - dataLengthCriterion = ( - if isGreedy - then id - else safeReciprocal -- Prefer less data. - ) . fromIntegral . length - --- matchLists :: [MatchList m] - matchLists = Tree.pop match - in ( - dataLengthCriterion $ extractDataFromMatch match, - if ExecutionOptions.preferFewerRepeatedAlternatives executionOptions - then safeReciprocal . toRational $ length matchLists - else 0, - if ExecutionOptions.preferAlternativesWhichFeedTheGreedy executionOptions - then let - excessConsumption :: MatchedData m -> ConsumptionBounds.DataLength - excessConsumption ( - Repeatable.MkRepeatable { - Repeatable.repetitionBounds = (f, _), - Repeatable.isGreedy = g - }, - _, - consumedInputData - ) - | g = l - f - | otherwise = f - l - where - l :: ConsumptionBounds.DataLength - l = length consumedInputData - in Data.Foldable.foldl' (\l -> (l +) . excessConsumption) 0 match - else 0, - if ExecutionOptions.preferAlternativesWhichMimickUnrolling executionOptions - then map (dataLengthCriterion . extractDataFromMatchList) matchLists - else [] - ) -- Create a tuple of selection-criteria, for simultaneous assessment. - ) . fst {-focus the choice on the Alternatives-} &&& id -- Prepend a selection-criterion to each result. - ) - ) matchPairList - where --- tailMatch, zeroRepetitions :: Maybe (MatchList m) - tailMatch = findMatchSlave concatenationTail accumulatedConsumptionProfilesTail distinctMetaDataTail inputData inputDataLength distinctInputData -- Recurse. - zeroRepetitions = (Tree.Node [] :) `fmap` {-apply to Maybe Functor-} tailMatch -- Prepend a null 'MatchList'. - --- extendedRegExAlternatives :: [ExtendedRegEx m] - extendedRegExAlternatives = Data.List.nub $ deconstructAlternatives alternatives - - isSingletonAlternative :: Bool - isSingletonAlternative = length extendedRegExAlternatives == 1 - - minConsumptionAlternatives :: ConsumptionBounds.DataLength - maybeMaxConsumptionAlternatives :: Maybe ConsumptionBounds.DataLength - (minConsumptionAlternatives, maybeMaxConsumptionAlternatives) = Consumer.getConsumptionBounds alternatives -- Independent of the choice of Alternative. - where - inputDataOffset :: ConsumptionBounds.DataLength - inputDataOffset = originalInputDataLength - inputDataLength - findMatchSlave _ _ _ _ _ _ = error "RegExDot.RegEx.findMatch.findMatchSlave:\tunexpected parameter-pattern." - -{- | - * Operator's name was chosen to suggest something more than '=~'. - - * CAVEAT: much more expensive then '=~': in /ghci/, 'Just' can be observed to be printed /long/ before the 'MatchList' from which 'Result' is constructed, - as the lazy algorithm finds the first solution, but not yet necessarily the optimal solution, amongst 'Alternatives'. --} -(+~) :: (Eq m, Control.DeepSeq.NFData m) - => InputData m -- ^ The input data within which to locate a match. - -> RegExOpts.RegExOpts (ExtendedRegEx m) -- ^ The match-options parameterised by the regex against which to match the input data. - -> Result m -inputData +~ regExOpts = ( - if hasBowAnchor' then Nothing else fmap head maybeMatchList, -- Record the first 'Match', consumed in the absence of a top-level 'Anchor.Bow'. - fmap ( - ( - if hasBowAnchor' then id else tail -- Remove the first 'Match', since this wasn't consumed by the 'Concatenation'. - ) . ( - if hasSternAnchor' then id else init -- Remove the last 'Match', since this wasn't consumed by the 'Concatenation'. - ) - ) maybeMatchList, - if hasSternAnchor' then Nothing else fmap last maybeMatchList -- Record the last 'Match', consumed in the absence of a top-level 'Anchor.Stern'. - ) where --- extendedRegEx :: ExtendedRegEx a - extendedRegEx = RegExOpts.regEx regExOpts - - hasBowAnchor', hasSternAnchor' :: Bool - hasBowAnchor' = hasBowAnchor extendedRegEx - hasSternAnchor' = hasSternAnchor extendedRegEx - --- maybeMatchList :: Maybe (MatchList m) - maybeMatchList = fmap drift (RegExOpts.setVerbose True regExOpts) `findMatch` inputData - -{- | - * Pattern-match operator. - - * Identifier & parameter-order follow the lead of /Perl/'s pattern-match operator. - - * Considerably more efficient than '+~', since even though they are both implemented via 'findMatch', - the discovery of /any/ solution is sufficient to generate the return-value; - lazy-evaluation avoids the requirement to identify the irrelevant optimal solution. --} -(=~) :: (Eq m, Control.DeepSeq.NFData m) - => InputData m -- ^ The input data within which to locate a match. - -> RegExOpts.RegExOpts (ExtendedRegEx m) -- ^ The match-options parameterised by the regex against which to match the input data. - -> Bool -inputData =~ regExOpts = Data.Maybe.isJust $ fmap drift (RegExOpts.setVerbose False regExOpts) `findMatch` inputData - --- | Pattern-mismatch operator. -(/~) :: (Eq m, Control.DeepSeq.NFData m) - => InputData m -- ^ The input data within which to locate a match. - -> RegExOpts.RegExOpts (ExtendedRegEx m) -- ^ The match-options parameterised by the regex against which to match the input data. - -> Bool -(/~) inputData = not . (inputData =~) - --- | The delimiters of 'Alternatives', when in the 'String'-form. -captureGroupDelimiters :: (Char, Char) -captureGroupDelimiters = ('(', ')') - --- | The token used to separate alternative 'ExtendedRegEx's, when in the 'String'-form. -alternativeExtendedRegExSeparatorToken :: Char -alternativeExtendedRegExSeparatorToken = '|' - --- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. -tokens :: String -tokens = [fst captureGroupDelimiters, snd captureGroupDelimiters, alternativeExtendedRegExSeparatorToken] - -{- | - * Invert the specified fraction, but return 'maxBound' if the result would otherwise be infinite. - - * CAVEAT: this is an awful concept, and therefore intended for internal use only. --} -safeReciprocal :: (Eq f, Fractional f) => f -> f -safeReciprocal 0 = fromIntegral (maxBound :: Int) -- Handle divide-by-zero error. -safeReciprocal f = recip f - -infixl 7 /+ -- Same as (/). - -{- | - * Integral division, with any fractional remainder rounded-up. - - * A rather dubious requirement, so internal use only. --} --- (/+) :: Integral i => i -> i -> i -(/+) - :: Int -- ^ Numerator. - -> Int -- ^ Denominator. - -> Int -- 10% faster in unoptimised code, & more in optimised. -_ /+ 0 = error "RegExDot.RegEx.(/+):\tzero denominator => infinity" --- numerator /+ denominator = ceiling ((fromIntegral numerator / fromIntegral denominator) :: Double) -numerator /+ denominator = uncurry (+) . Control.Arrow.second signum $ quotRem numerator denominator -- Slightly faster. diff --git a/src/RegExDot/RegExOpts.hs b/src/RegExDot/RegExOpts.hs deleted file mode 100644 --- a/src/RegExDot/RegExOpts.hs +++ /dev/null @@ -1,65 +0,0 @@ -{-# LANGUAGE MultiParamTypeClasses, TypeSynonymInstances, FlexibleInstances #-} -{-# OPTIONS_GHC -fno-warn-orphans #-} - -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * Provides a complete job-specification. - - * 'regEx' is polymorphic, allowing use of either /RegExDot.RegEx.ExtendedRegEx/ (which is also polymorphic) or /RegExChar.ExtendedRegExChar.ExtendedRegExChar/. --} - -module RegExDot.RegExOpts( --- * Types --- ** Data-types - RegExOpts(..), --- * Functions - setVerbose, --- ** Constructors - mkRegEx -) where - -import qualified RegExDot.CompilationOptions as CompilationOptions -import qualified RegExDot.ExecutionOptions as ExecutionOptions -import qualified ToolShed.Defaultable - --- | Aggregates both 'ExecutionOptions.ExecutionOptions' & 'CompilationOptions.CompilationOptions' with a polymorphic /regex/, to form a complete job-description. -data RegExOpts a = MkRegExOpts { - compilationOptions :: CompilationOptions.CompilationOptions, -- ^ Parameters governing the result, rather than the implementation. - executionOptions :: ExecutionOptions.ExecutionOptions, -- ^ Parameters governing the implementation, rather than the result. - regEx :: a -- ^ Polymorphic, to permit specialisation either of the type of /regex/ or the type of input-data. -} deriving Show - -instance Functor RegExOpts where - fmap f regExOpts = regExOpts { regEx = f $ regEx regExOpts } - --- | Smart constructor using 'ToolShed.Defaultable.defaultValue'. -mkRegEx :: a -> RegExOpts a -mkRegEx r = MkRegExOpts { - compilationOptions = ToolShed.Defaultable.defaultValue, - executionOptions = ToolShed.Defaultable.defaultValue, - regEx = r -} - --- | Mutator, which sets an appropriate 'ExecutionOptions.ExecutionOptions', for the depth to which the caller wants to probe the resulting match. -setVerbose :: Bool -> RegExOpts a -> RegExOpts a -setVerbose verbose regExOpts = regExOpts { executionOptions = ExecutionOptions.setVerbose verbose $ executionOptions regExOpts } - diff --git a/src/RegExDot/Repeatable.hs b/src/RegExDot/Repeatable.hs deleted file mode 100644 --- a/src/RegExDot/Repeatable.hs +++ /dev/null @@ -1,397 +0,0 @@ -{- - Copyright (C) 2010-2015 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * A data-type, which represents the permissible range of instances, of the underlying polymorphic datum. - - * Designed for use in a polymorphic /regex/-engine, which specifies patterns composed of repeated /greedy/ & /non-greedy/ sequences of /Meta/-data; - - > * + ? {fewest, most} {fewest,} {fewest} - > *? +? ?? {fewest, most}? {fewest,}? - - * In the context of /regex/es, this concept is known as /Quantification/. - - * /regex/es evolved from the minimal ability to optionally qualify the datum with a suffix. - More exotic repetition-specifications could be composed by concatenating these atomic building-blocks. - Here, I've taken the contrary top-down view, & assumed that all data are qualified by a full 'RepetitionBounds', which in most cases will degenerate into a simpler form. - - * The type of entity which is being repeated, isn't the domain of this data-type; it's polymorphic. --} - -module RegExDot.Repeatable( --- * Types --- ** Type-synonyms - Repetitions, - RepetitionBounds, --- ** Data-types - Repeatable(..), --- * Constants --- nonGreedyToken, - oneOrMoreToken, - rangeDelimiters, - rangeSeparatorToken, - tokens, - zeroOrMoreToken, - zeroOrOneToken, --- * Functions - one, - oneOrMore, - oneOrMore', - zeroOrMore, - zeroOrMore', - zeroOrOne, - zeroOrOne', --- precisely, - repeatableParser, - showSuffix, --- ** Accessors - getFewest, - getMost, --- ** Mutators - focus, --- setNonGreedy, - toSingleton, --- ** Operators - (^#->#), - (^#->#?), - (^#->), - (^#->?), - (^#), --- ** Predicates - isPrecise, - hasPreciseBounds -) where - -import Prelude hiding ((<$>), (<*>)) -- The "Prelude" from 'base-4.8' exports this symbol. -import Control.Applicative((<$>), (<*>)) -import Control.Arrow((***)) -import qualified Control.DeepSeq -import qualified Data.List -import qualified RegExDot.Consumer as Consumer -import qualified RegExDot.ConsumptionProfile as ConsumptionProfile -import qualified Text.ParserCombinators.Parsec as Parsec -import Text.ParserCombinators.Parsec(()) -import qualified ToolShed.Data.Pair -import qualified ToolShed.SelfValidate - -infix 6 ^#->#, ^#->#?, ^#->, ^#->?, ^# -- A notch tighter than "DSL"s binary operators. - --- | A number of repetitions. -type Repetitions = Int - --- | Defines the bounds of a range of permissible repetitions. -type RepetitionBounds = (Repetitions, Maybe Repetitions) - --- | Creates a precise 'RepetitionBounds', i.e. both lower & upper bounds on the number of 'Repetitions' are equal to the same value. -precisely :: Repetitions -> RepetitionBounds -precisely i = (i, Just i) - --- | Predicate which is 'True' if exactly one value is permissible, ie lower & upper bounds on the number of 'Repetitions' are identical. -hasPreciseBounds :: RepetitionBounds -> Bool -hasPreciseBounds (fewest, most) = Just fewest == most - --- | Declares a polymorphic data-type, which augments the underlying 'base' datum, with the range of times it may be used. -data Repeatable a = MkRepeatable { - base :: a, -- ^ The underlying polymorphic datum. - repetitionBounds :: RepetitionBounds, -- ^ The bounds delimiting the range of permissible repetitions, of 'base'. - isGreedy :: Bool -- ^ Whether to demand as many matching instances of 'base' as possible; or as few (AKA /lazy quantification/). -} deriving Eq - -instance Functor Repeatable where - fmap f repeatable = repeatable { base = f $ base repeatable } - --- | True if there's no choice in the number of repetitions; implemented via 'isPrecise'. -isPrecise :: Repeatable a -> Bool -isPrecise = hasPreciseBounds . repetitionBounds - --- | Builds a parser for a specification of the number of permissible instances of the specified polymorphic parameter. -repeatableParser :: a -> Parsec.Parser (Repeatable a) -repeatableParser b = Parsec.option ( - one b -- The default; there's no concept of greediness here. - ) $ do - repeatable <- Parsec.choice [ - (Parsec.char oneOrMoreToken "Repeatable.oneOrMoreToken " ++ show oneOrMoreToken) >> return {-to ParsecT-monad-} (oneOrMore b), - (Parsec.char zeroOrOneToken "Repeatable.zeroOrOneToken " ++ show zeroOrOneToken) >> return {-to ParsecT-monad-} (zeroOrOne b), - (Parsec.char zeroOrMoreToken "Repeatable.zeroOrMoreToken " ++ show zeroOrMoreToken) >> return {-to ParsecT-monad-} (zeroOrMore b), - (b ^#->#) {-arbitrarily greedy for now-} <$> uncurry Parsec.between (ToolShed.Data.Pair.mirror Parsec.char rangeDelimiters) ( - do - fewest <- Parsec.spaces >> (read <$> Parsec.many1 Parsec.digit "Repetition-range minimum") - most <- Parsec.spaces >> Parsec.option ( - Just fewest -- The default. - ) ( - do - i <- ( - Parsec.char rangeSeparatorToken "Repeatable.rangeSeparatorToken " ++ show rangeSeparatorToken - ) >> Parsec.spaces >> Parsec.option Nothing {-default to open-ended range-} ( - Just . read <$> Parsec.many1 Parsec.digit "Repetition-range maximum" - ) - - Parsec.spaces >> return {-to ParsecT-monad-} i - ) - - return {-to ParsecT-monad-} (fewest, most) - ) "Repeatable.rangeDelimiters " ++ show rangeDelimiters - ] - - g <- Parsec.option True {-the default-} $ (Parsec.char nonGreedyToken "Repeatable.nonGreedyToken " ++ show nonGreedyToken) >> return {-to ParsecT-monad-} False - - return {-to ParsecT-monad-} repeatable { isGreedy = g } -- Correct prior assumption. - -instance Read a => Read (Repeatable a) where - readsPrec _ s = case reads s {-first, read the base-type-} of - [(base', s1)] -> (error . ("readsPrec Repeatable:\tparse-error; " ++) . show) `either` return $ Parsec.parse ((,) <$> repeatableParser base' <*> Parsec.getInput) "Repeatable" s1 - _ -> [] -- No parse. - -{- | - * A 'ShowS'-function for the suffix, denoting the permissible repetitions, of 'base'. - - * This function converts the internal, into the tradition /greedy/ & /non-greedy/ quantifiers of various specific varieties. --} -showSuffix :: Repeatable a -> ShowS -showSuffix repeatable = let - showRange :: ShowS -> ShowS - showRange x = (\(begin, end) -> begin . x . end) $ ToolShed.Data.Pair.mirror showChar rangeDelimiters - in ( - case repetitionBounds repeatable of - (0, Nothing) -> showChar zeroOrMoreToken - (1, Nothing) -> showChar oneOrMoreToken - (fewest, Nothing) -> showRange $ shows fewest . showChar rangeSeparatorToken - (0, Just 1) -> showChar zeroOrOneToken - (1, Just 1) -> id -- CAVEAT: since there's no explicit repetition-operator, the non-greedy modifier can't be appended. - (fewest, Just most) -> showRange $ if fewest == most - then shows fewest -- Single-valued range. - else shows fewest . showChar rangeSeparatorToken . shows most - ) . if ($ repeatable) `any` [isGreedy, isPrecise] {-without a range of possibilities, non-greediness is irrelevant-} - then id - else showChar nonGreedyToken -- This can only be appended, if there a previous repetition-operator for it to modify. - --- Replicate the syntax, for repetition, as used in a POSIX-standard /regex/. -instance Show a => Show (Repeatable a) where - showsPrec _ repeatable = shows (base repeatable) . showSuffix repeatable - -instance Consumer.Consumer a => Consumer.Consumer (Repeatable a) where - consumptionProfile MkRepeatable { - base = b, - repetitionBounds = (fewest, most) - } = baseConsumptionProfile { - ConsumptionProfile.consumptionBounds = (fewest *) *** ((*) <$> most <*>) $ ConsumptionProfile.consumptionBounds baseConsumptionProfile -- CAVEAT: special cases exist, where one or both halves of this calculation degenerate to a simpler form, but special treatment, in an attempt to improve performance, proved counterproductive. - } where - baseConsumptionProfile :: ConsumptionProfile.ConsumptionProfile - baseConsumptionProfile = Consumer.consumptionProfile b - - starHeight MkRepeatable { - base = b, - repetitionBounds = r - } = Consumer.starHeight b + if hasPreciseBounds r then 0 else 1 - -instance ToolShed.SelfValidate.SelfValidator a => ToolShed.SelfValidate.SelfValidator (Repeatable a) where - getErrors MkRepeatable { - base = b, - repetitionBounds = (fewest, most), - isGreedy = g - } - | not $ ToolShed.SelfValidate.isValid b = ToolShed.SelfValidate.getErrors b -- Delegate. - | otherwise = ToolShed.SelfValidate.extractErrors [ - (fewest < 0, "Negative fewest=" ++ show fewest ++ "."), - ( - case most of - Just m -> m < fewest - _ -> False, - "Invalid repetition-range; '" ++ show (fewest, most) ++ "'." - ), ( - not g && case most of - Just m -> fewest >= m -- There ought to be potential for non-greediness, where specified: the converse isn't true, since greediness isn't explicit, & may not have been wanted. - _ -> False, - "Invalid non-greedy repetition-range; '" ++ show (fewest, most) ++ "'." - ) - ] - -instance Control.DeepSeq.NFData a => Control.DeepSeq.NFData (Repeatable a) where - rnf MkRepeatable { - base = b, - repetitionBounds = r, - isGreedy = g - } = Control.DeepSeq.rnf (b, r, g) - --- | Mutator. -setNonGreedy :: Repeatable a -> Repeatable a -setNonGreedy r = r { isGreedy = False } - -{- | - * Construct a greedy 'Repeatable', from a polymorphic datum, with the specified range of permissible instances. - - * The /#/s in the identifier represent the two bounds. - - * /a{f, m}/ --} -(^#->#) - :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. - -> RepetitionBounds -- ^ The permissible repetition-bounds for the polymorphic data. - -> Repeatable a -b ^#-># bounds = MkRepeatable { - base = b, - repetitionBounds = bounds, - isGreedy = True -} - -{- | - * Construct a non-greedy version of '^#->#'. - - * /a{f, m}?/ --} -(^#->#?) - :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. - -> RepetitionBounds -- ^ The permissible repetition-bounds for the polymorphic data. - -> Repeatable a -b ^#->#? bounds = setNonGreedy (b ^#-># bounds) - -{- | - * Construct a greedy 'Repeatable', tailored for data repeated at least the specified number of times. - - * The /#/ in the identifier represents the single bound. - - * /a{f,}/ --} -(^#->) - :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. - -> Repetitions -- ^ The minimum permissible repetitions of the polymorphic data. - -> Repeatable a -b ^#-> fewest = b ^#-># (fewest, Nothing) - -{- | - * Construct a non-greedy version of '^#->'. - - * /a{f,}?/ --} -(^#->?) - :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. - -> Repetitions -- ^ The minimum permissible repetitions of the polymorphic data. - -> Repeatable a -b ^#->? fewest = setNonGreedy (b ^#-> fewest) - -{- | - * Construct a 'Repeatable', tailored for data repeated a precise number of times. - - * The /#/ in the identifier represents the single bound. - - * /a{f}/ --} -(^#) - :: a -- ^ The polymorphic payload from which to construct the 'Repeatable'. - -> Repetitions -- ^ The precise number of repetitions of the polymorphic data which is required. - -> Repeatable a -b ^# r = b ^#-># precisely r - -{- | - * Construct a 'Repeatable', tailored for unrepeated data. - - * A degenerate case of '^#'. --} -one :: a -> Repeatable a -one = (^# 1) - -{- | - * Construct a greedy 'Repeatable', from a polymorphic datum, with 'fewest' == 0 & 'most' == 1. - - * A specific case of '^#->#'. --} -zeroOrOne :: a -> Repeatable a -zeroOrOne = (^#-># (0, Just 1)) - --- | Construct a non-greedy version of 'zeroOrOne'. -zeroOrOne' :: a -> Repeatable a -zeroOrOne' = setNonGreedy . zeroOrOne - -{- | - * Construct a greedy 'Repeatable', from a polymorphic datum, with 'fewest' == 0. - - * A specific case of '^#->'. --} -zeroOrMore :: a -> Repeatable a -zeroOrMore = (^#-> 0) - --- | Construct a non-greedy version of 'zeroOrMore'. -zeroOrMore' :: a -> Repeatable a -zeroOrMore' = setNonGreedy . zeroOrMore - -{- | - * Construct a greedy 'Repeatable', from a polymorphic datum, with lower 'RepetitionBounds' == one. - - * A specific case of '^#->'. --} -oneOrMore :: a -> Repeatable a -oneOrMore = (^#-> 1) - --- | Construct a non-greedy version of 'oneOrMore'. -oneOrMore' :: a -> Repeatable a -oneOrMore' = setNonGreedy . oneOrMore - --- | Reduces a 'Repeatable', with a range of 'RepetitionBounds', to a precise number of repetitions. -focus :: Repeatable a -> Repetitions -> Repeatable a -focus r i = r { repetitionBounds = precisely i } - -{- | - * Reduces a 'Repeatable', with a range of 'RepetitionBounds', to a singleton. - - * A degenerate case of 'focus'. --} -toSingleton :: Repeatable a -> Repeatable a -toSingleton = (`focus` 1) - --- | Accessor. -getFewest :: Repeatable a -> Repetitions -getFewest MkRepeatable { repetitionBounds = (f, _) } = f - --- | Accessor. -getMost :: Repeatable a -> Maybe Repetitions -getMost MkRepeatable { repetitionBounds = (_, m) } = m - --- | The token used to denote /non-greedy/, when in the 'String'-form. -nonGreedyToken :: Char -nonGreedyToken = '?' - -{- | - * The token used to denote 'zeroOrMore', when in the 'String'-form. - - * AKA /Kleene Star/. --} -zeroOrMoreToken :: Char -zeroOrMoreToken = '*' - --- | The token used to denote 'zeroOrOne', when in the 'String'-form. -zeroOrOneToken :: Char -zeroOrOneToken = '?' - --- | The token used to denote 'oneOrMore', when in the 'String'-form. -oneOrMoreToken :: Char -oneOrMoreToken = '+' - --- | The delimiters of '^#->#', when in the 'String'-form. -rangeDelimiters :: (Char, Char) -rangeDelimiters = ('{', '}') - --- | The token used to separate 'RepetitionBounds', when in the 'String'-form. -rangeSeparatorToken :: Char -rangeSeparatorToken = ',' - --- | The set of 'Char' to which a specific meaning is attributed, when reading from 'String'. -tokens :: String -tokens = Data.List.nub [nonGreedyToken, zeroOrMoreToken, zeroOrOneToken, oneOrMoreToken, fst rangeDelimiters, snd rangeDelimiters, rangeSeparatorToken] - diff --git a/src/RegExDot/Result.hs b/src/RegExDot/Result.hs deleted file mode 100644 --- a/src/RegExDot/Result.hs +++ /dev/null @@ -1,60 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] Tools to manipulate a 'RegEx.Result'. - - [@CAVEAT@] The data-definition remains in "RegEx", since it references 'RegEx.Match'. --} - -module RegExDot.Result( --- * Functions --- ** Accessors (Deconstructors) - getMatchList, - getPreMatch, - getPostMatch, --- ** Predicates - isMatch, --- ** Query - countMatches -) where - -import qualified Data.Maybe -import qualified RegExDot.RegEx as RegEx - --- | Accessor. -getPreMatch :: RegEx.Result a -> RegEx.ExternalMatch a -getPreMatch (externalMatch, _, _) = externalMatch - --- | Accessor. -getPostMatch :: RegEx.Result a -> RegEx.ExternalMatch a -getPostMatch (_, _, externalMatch) = externalMatch - --- | Accessor. -getMatchList :: RegEx.Result a -> Maybe (RegEx.MatchList a) -getMatchList (_, maybeMatchList, _) = maybeMatchList - --- | True if the 'RegEx.InputData' matched the 'RegEx.ExtendedRegEx'. -isMatch :: RegEx.Result a -> Bool -isMatch = Data.Maybe.isJust . getMatchList - --- | Counts the number of top-level 'RegEx.Match'es. -countMatches :: RegEx.Result a -> Int -countMatches (_, Just matchList, _) = length matchList -countMatches _ = 0 - diff --git a/src/RegExDot/ShowablePredicate.hs b/src/RegExDot/ShowablePredicate.hs deleted file mode 100644 --- a/src/RegExDot/ShowablePredicate.hs +++ /dev/null @@ -1,53 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] Used to implement /Perl-shortcut/s, /BracketExpression/-ranges & /POSIX Character-classes/. --} - -module RegExDot.ShowablePredicate( --- * Types - Predicate, --- ** Data-types - ShowablePredicate( - MkShowablePredicate, - name, - predicate - ) -) where - -import qualified Control.DeepSeq - --- | An arbitrary polymorphic predicate function. -type Predicate a = a -> Bool - --- | Container for both the /predicate/, & the name used in the implementation of 'Show'. -data ShowablePredicate a = MkShowablePredicate { - name :: String, -- ^ The identifier of this predicate when in string-form. - predicate :: Predicate a -- ^ The function used to determine whether an input datum matches. -} - -instance Show (ShowablePredicate a) where - showsPrec _ = showString . name - -instance Eq (ShowablePredicate a) where - l == r = name l == name r -- Ignore 'predicate'. - -instance Control.DeepSeq.NFData (ShowablePredicate a) where - rnf = Control.DeepSeq.rnf . name -- Ignore 'predicate'. - diff --git a/src/RegExDot/Span.hs b/src/RegExDot/Span.hs deleted file mode 100644 --- a/src/RegExDot/Span.hs +++ /dev/null @@ -1,64 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] Delimits a data-sequence, via its offset & length, within a wider list. --} - -module RegExDot.Span( --- * Types --- ** Type-synonyms - Span, --- * Functions - after, - empty, - join -) where - -import Control.Arrow((***)) -import qualified RegExDot.ConsumptionBounds as ConsumptionBounds - -{- | - * The offset & length of a specific data-sequence, within a wider list. - - * Similar to . --} -type Span = (ConsumptionBounds.DataLength {-offset-}, ConsumptionBounds.DataLength) - --- | An empty instance, located at the specified offset. -empty :: ConsumptionBounds.DataLength -> Span -empty offset = (offset, 0) - --- | Returns the furthest extent, i.e. the offset just after the end. -after :: Span -> ConsumptionBounds.DataLength -after = uncurry (+) - -{- | - * Condenses a list into a single value. - - * Expects a gap-free, sorted list. --} -join - :: ConsumptionBounds.DataLength -- ^ The offset into the list of input-data to use when a null list of spans is received. - -> [Span] - -> Span -join offset [] = empty offset -- The offset can't be deduced from a null list, so use the value provided. -join _ [singleton] = singleton -- Merely for efficiency. -join _ spanList = head *** sum $ unzip spanList - - diff --git a/src/RegExDot/Tree.hs b/src/RegExDot/Tree.hs deleted file mode 100644 --- a/src/RegExDot/Tree.hs +++ /dev/null @@ -1,80 +0,0 @@ -{- - Copyright (C) 2010 Dr. Alistair Ward - - This program is free software: you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation, either version 3 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see . --} -{- | - [@AUTHOR@] Dr. Alistair Ward - - [@DESCRIPTION@] - - * Defines a specific type of tree-structure, which is composed from either a datum, or a list of Tree-lists; . - - * This more general tree-structure has the shape of a regular-expression match. - - * This contrasts with the typical binary-tree, which is either empty, or contains a (left Tree, datum, right Tree). --} - -module RegExDot.Tree( --- * Types --- ** Type-synonyms --- TreeList, --- ** Data-types - Tree(..), --- * Functions - pop -) where - -import qualified Control.Arrow -import qualified Control.DeepSeq -import qualified Data.Foldable -import qualified Data.List -import qualified Data.Monoid - --- | A general purpose tree-type structure. -data Tree a = - Leaf a -- ^ The payload. - | Node [TreeList a] -- ^ Recurse. NB: a list of lists is required to contain the /MatchLists/ resulting from repeated /Alternatives/. - deriving Eq - --- | A list of 'Tree's; significant only because it is the essence of the recursive nature of 'Tree'. -type TreeList a = [Tree a] - -instance Show a => Show (Tree a) where - showsPrec _ (Leaf a) = shows a - showsPrec _ (Node treeLists) = showList treeLists - -instance Read a => Read (Tree a) where - readsPrec _ node@('[' : _) = Control.Arrow.first Node `map` readList node {-singleton-} - readsPrec _ leaf = Control.Arrow.first Leaf `map` reads leaf {-singleton-} - -instance Control.DeepSeq.NFData a => Control.DeepSeq.NFData (Tree a) where - rnf (Leaf a) = Control.DeepSeq.rnf a - rnf (Node l) = Control.DeepSeq.rnf l - -instance Functor Tree where - fmap f (Leaf a) = Leaf $ f a - fmap f (Node treeLists) = Node $ map (fmap {-recurse-} f `map`) treeLists - -instance Data.Foldable.Foldable Tree where - foldMap f (Leaf a) = f a -- CAVEAT: 'f' should be Associative, as required by a Monoid. - foldMap f (Node treeLists) = Data.List.foldl' (Data.List.foldl' (\monoid -> (monoid `Data.Monoid.mappend`) . Data.Foldable.foldMap f)) Data.Monoid.mempty treeLists - --- | Deconstruct the specified 'Node'; i.e. lop the apex from the 'Tree', leaving a flat top. -pop - :: Tree a -- ^ The tree from which to extract the list of trees hanging immediately beneath its apex. - -> [TreeList a] -pop (Node treeLists) = treeLists -pop _ = error "RegExDot.Tree.pop:\tunexpected Leaf" -