diff --git a/bench/bench.hs b/bench/bench.hs --- a/bench/bench.hs +++ b/bench/bench.hs @@ -57,6 +57,21 @@ , bgroup "dangerouslyBig" $ benchToBoundedInteger dangerouslyBig , bgroup "64" $ benchToBoundedInteger 64 ] + + , bgroup "read" + [ benchRead "123456789.123456789" + , benchRead "12345678900000000000.12345678900000000000000000" + , benchRead "12345678900000000000.12345678900000000000000000e1234" + ] + + , bgroup "division" + [ bench (show n ++ " / " ++ show d) $ nf (uncurry (/)) t + | t@(n, d) <- + [ (0.4 , 20.0) + , (0.4e-100, 0.2e50) + ] :: [(Scientific, Scientific)] + ] + ] where pos :: Fractional a => a diff --git a/changelog b/changelog --- a/changelog +++ b/changelog @@ -1,3 +1,69 @@ +0.3.7.0 + + * Make division (/) on Scientifics slightly more efficient. + + * Fix the Show instance to surround negative numbers with parentheses when + necessary. + + * Add (Template Haskell) Lift Scientific instance + + * Mark modules as Safe or Trustworthy (Safe Haskell). + +0.3.6.2 + * Due to a regression introduced in 0.3.4.14 the RealFrac methods + and floatingOrInteger became vulnerable to a space blowup when + applied to scientifics with huge exponents. This has now been + fixed again. + +0.3.6.1 + * Fix build on GHC < 8. + +0.3.6.0 + * Make the methods of the Hashable, Eq and Ord instances safe to + use when applied to scientific numbers coming from untrusted + sources. Previously these methods first converted their arguments + to Rational before applying the operation. This is unsafe because + converting a Scientific to a Rational could fill up all space and + crash your program when the Scientific has a huge base10Exponent. + + Do note that the hash computation of the Hashable Scientific + instance has been changed because of this improvement! + + Thanks to Tom Sydney Kerckhove (@NorfairKing) for pushing me to + fix this. + + * fromRational :: Rational -> Scientific now throws an error + instead of diverging when applied to a repeating decimal. This + does mean it will consume space linear in the number of digits of + the resulting scientific. This makes "fromRational" and the other + Fractional methods "recip" and "/" a bit safer to use. + + * To get the old unsafe but more efficient behaviour the following + function was added: unsafeFromRational :: Rational -> Scientific. + + * Add alternatives for fromRationalRepetend: + + fromRationalRepetendLimited + :: Int -- ^ limit + -> Rational + -> Either (Scientific, Rational) + (Scientific, Maybe Int) + + and: + + fromRationalRepetendUnlimited + :: Rational -> (Scientific, Maybe Int) + + Thanks to Ian Jeffries (@seagreen) for the idea. + +0.3.5.3 + * Dropped upper version bounds of dependencies + because it's to much work to maintain. + +0.3.5.2 + * Remove unused ghc-prim dependency. + * Added unit tests for read and scientificP + 0.3.5.1 * Replace use of Vector from vector with Array from primitive. diff --git a/scientific.cabal b/scientific.cabal --- a/scientific.cabal +++ b/scientific.cabal @@ -1,8 +1,8 @@ -name: scientific -version: 0.3.5.1 -synopsis: Numbers represented using scientific notation +name: scientific +version: 0.3.8.1 +synopsis: Numbers represented using scientific notation description: - @Data.Scientific@ provides the number type 'Scientific'. Scientific numbers are + "Data.Scientific" provides the number type 'Scientific'. Scientific numbers are arbitrary precision and space efficient. They are represented using . The implementation uses a coefficient @c :: 'Integer'@ and a base-10 exponent @@ -25,95 +25,114 @@ @1e1000000000 :: 'Rational'@ will fill up all space and crash your program. Scientific works as expected: . - > > read "1e1000000000" :: Scientific - > 1.0e1000000000 + >>> read "1e1000000000" :: Scientific + 1.0e1000000000 . * Also, the space usage of converting scientific numbers with huge exponents to @'Integral's@ (like: 'Int') or @'RealFloat's@ (like: 'Double' or 'Float') will always be bounded by the target type. -homepage: https://github.com/basvandijk/scientific -bug-reports: https://github.com/basvandijk/scientific/issues -license: BSD3 -license-file: LICENSE -author: Bas van Dijk -maintainer: Bas van Dijk -category: Data -build-type: Simple -cabal-version: >=1.10 - -extra-source-files: - changelog +homepage: https://github.com/basvandijk/scientific +bug-reports: https://github.com/basvandijk/scientific/issues +license: BSD3 +license-file: LICENSE +author: Bas van Dijk +maintainer: Bas van Dijk +category: Data +build-type: Simple +cabal-version: >=1.10 +extra-source-files: changelog +tested-with: + GHC ==8.6.5 + || ==8.8.4 + || ==8.10.7 + || ==9.0.2 + || ==9.2.8 + || ==9.4.8 + || ==9.6.6 + || ==9.8.4 + || ==9.10.3 + || ==9.12.2 + || ==9.14.1 source-repository head type: git - location: git://github.com/basvandijk/scientific.git - -flag bytestring-builder - description: Depend on the bytestring-builder package for backwards compatibility. - default: False - manual: False + location: https://github.com/basvandijk/scientific.git flag integer-simple description: Use the integer-simple package instead of integer-gmp default: False library - exposed-modules: Data.ByteString.Builder.Scientific - Data.Scientific - Data.Text.Lazy.Builder.Scientific - other-modules: GHC.Integer.Compat - Utils - other-extensions: DeriveDataTypeable, BangPatterns - ghc-options: -Wall - build-depends: base >= 4.3 && < 4.11 - , ghc-prim - , integer-logarithms >= 1 && <1.1 - , deepseq >= 1.3 && < 1.5 - , text >= 0.8 && < 1.3 - , hashable >= 1.1.2 && < 1.3 - , primitive >= 0.1 && < 0.7 - , containers >= 0.1 && < 0.6 - , binary >= 0.4.1 && < 0.9 + exposed-modules: + Data.ByteString.Builder.Scientific + Data.Scientific + Data.Text.Lazy.Builder.Scientific - if flag(bytestring-builder) - build-depends: bytestring >= 0.9 && < 0.10.4 - , bytestring-builder >= 0.10.4 && < 0.11 - else - build-depends: bytestring >= 0.10.4 && < 0.11 + other-modules: + GHC.Integer.Compat + Utils - if flag(integer-simple) - build-depends: integer-simple + other-extensions: + BangPatterns + DeriveDataTypeable + Trustworthy + + ghc-options: -Wall + build-depends: + base >=4.12.0.0 && <4.23 + , binary >=0.8.6.0 && <0.9 + , bytestring >=0.10.8.2 && <0.13 + , containers >=0.6.0.1 && <0.9 + , deepseq >=1.4.4.0 && <1.6 + , hashable >=1.4.4.0 && <1.6 + , integer-logarithms >=1.0.3.1 && <1.1 + , primitive >=0.9.0.0 && <0.10 + , template-haskell >=2.14.0.0 && <2.25 + , text >=1.2.3.0 && <1.3 || >=2.0 && <2.2 + + if impl(ghc >=9.0) + build-depends: base >=4.15 + + if flag(integer-simple) + build-depends: invalid-cabal-flag-settings <0 + else + if flag(integer-simple) + build-depends: integer-simple + + else build-depends: integer-gmp - hs-source-dirs: src - default-language: Haskell2010 + if impl(ghc <8) + other-extensions: TemplateHaskell + if impl(ghc >=9.0) + -- these flags may abort compilation with GHC-8.10 + -- https://gitlab.haskell.org/ghc/ghc/-/merge_requests/3295 + ghc-options: -Winferred-safe-imports -Wmissing-safe-haskell-mode + + hs-source-dirs: src + default-language: Haskell2010 + test-suite test-scientific type: exitcode-stdio-1.0 hs-source-dirs: test main-is: test.hs default-language: Haskell2010 ghc-options: -Wall - - build-depends: scientific - , base >= 4.3 && < 4.11 - , binary >= 0.4.1 && < 0.9 - , tasty >= 0.5 && < 0.12 - , tasty-ant-xml >= 1.0 && < 1.2 - , tasty-hunit >= 0.8 && < 0.10 - , tasty-smallcheck >= 0.2 && < 0.9 - , tasty-quickcheck >= 0.8 && < 0.10 - , smallcheck >= 1.0 && < 1.2 - , QuickCheck >= 2.5 && < 2.11 - , text >= 0.8 && < 1.3 - - if flag(bytestring-builder) - build-depends: bytestring >= 0.9 && < 0.10.4 - , bytestring-builder >= 0.10.4 && < 0.11 - else - build-depends: bytestring >= 0.10.4 && < 0.11 + build-depends: + base + , binary + , bytestring + , QuickCheck >=2.14.2 + , scientific + , smallcheck >=1.0 + , tasty >=1.4.0.1 + , tasty-hunit >=0.8 + , tasty-quickcheck >=0.8 + , tasty-smallcheck >=0.2 + , text benchmark bench-scientific type: exitcode-stdio-1.0 @@ -121,6 +140,7 @@ main-is: bench.hs default-language: Haskell2010 ghc-options: -O2 - build-depends: scientific - , base >= 4.3 && < 4.11 - , criterion >= 0.5 && < 1.3 + build-depends: + base + , criterion >=0.5 + , scientific diff --git a/src/Data/ByteString/Builder/Scientific.hs b/src/Data/ByteString/Builder/Scientific.hs --- a/src/Data/ByteString/Builder/Scientific.hs +++ b/src/Data/ByteString/Builder/Scientific.hs @@ -1,4 +1,4 @@ -{-# LANGUAGE CPP, OverloadedStrings #-} +{-# LANGUAGE OverloadedStrings, Safe #-} module Data.ByteString.Builder.Scientific ( scientificBuilder @@ -17,18 +17,7 @@ import Utils (roundTo, i2d) -#if !MIN_VERSION_base(4,8,0) -import Data.Monoid (mempty) -#endif - -#if MIN_VERSION_base(4,5,0) import Data.Monoid ((<>)) -#else -import Data.Monoid (Monoid, mappend) -(<>) :: Monoid a => a -> a -> a -(<>) = mappend -infixr 6 <> -#endif -- | A @ByteString@ @Builder@ which renders a scientific number to full @@ -70,11 +59,10 @@ byteStringCopy (BC8.replicate dec' '0') <> byteStringCopy "e0" _ -> - let - (ei,is') = roundTo (dec'+1) is - (d:ds') = map i2d (if ei > 0 then init is' else is') - in - char8 d <> char8 '.' <> string8 ds' <> char8 'e' <> intDec (e-1+ei) + let (ei,is') = roundTo (dec'+1) is + in case map i2d (if ei > 0 then init is' else is') of + [] -> mempty + d:ds' -> char8 d <> char8 '.' <> string8 ds' <> char8 'e' <> intDec (e-1+ei) Fixed -> let mk0 ls = case ls of { "" -> char8 '0' ; _ -> string8 ls} @@ -100,8 +88,7 @@ in mk0 ls <> (if null rs then mempty else char8 '.' <> string8 rs) else - let - (ei,is') = roundTo dec' (replicate (-e) 0 ++ is) - d:ds' = map i2d (if ei > 0 then is' else 0:is') - in - char8 d <> (if null ds' then mempty else char8 '.' <> string8 ds') + let (ei,is') = roundTo dec' (replicate (-e) 0 ++ is) + in case map i2d (if ei > 0 then is' else 0:is') of + [] -> mempty + d:ds' -> char8 d <> (if null ds' then mempty else char8 '.' <> string8 ds') diff --git a/src/Data/Scientific.hs b/src/Data/Scientific.hs --- a/src/Data/Scientific.hs +++ b/src/Data/Scientific.hs @@ -1,9 +1,11 @@ -{-# LANGUAGE CPP #-} {-# LANGUAGE DeriveDataTypeable #-} {-# LANGUAGE BangPatterns #-} {-# LANGUAGE ScopedTypeVariables #-} {-# LANGUAGE UnboxedTuples #-} {-# LANGUAGE PatternGuards #-} +{-# LANGUAGE Trustworthy #-} +{-# LANGUAGE DeriveLift #-} +{-# LANGUAGE StandaloneDeriving #-} -- | -- Module : Data.Scientific @@ -22,6 +24,11 @@ -- aren't truly arbitrary precision. I intend to change the type of the exponent -- to 'Integer' in a future release. -- +-- /WARNING:/ Although @Scientific@ has instances for all numeric classes the +-- methods should be used with caution when applied to scientific numbers coming +-- from untrusted sources. See the warnings of the instances belonging to +-- 'Scientific'. +-- -- The main application of 'Scientific' is to be used as the target of parsing -- arbitrary precision numbers coming from an untrusted source. The advantages -- over using 'Rational' for this are that: @@ -41,16 +48,9 @@ -- to @'Integral's@ (like: 'Int') or @'RealFloat's@ (like: 'Double' or 'Float') -- will always be bounded by the target type. -- --- /WARNING:/ Although @Scientific@ is an instance of 'Fractional', the methods --- are only partially defined! Specifically 'recip' and '/' will diverge --- (i.e. loop and consume all space) when their outputs have an infinite decimal --- expansion. 'fromRational' will diverge when the input 'Rational' has an --- infinite decimal expansion. Consider using 'fromRationalRepetend' for these --- rationals which will detect the repetition and indicate where it starts. --- -- This module is designed to be imported qualified: -- --- @import Data.Scientific as Scientific@ +-- @import qualified Data.Scientific as Scientific@ module Data.Scientific ( Scientific @@ -66,8 +66,14 @@ , isInteger -- * Conversions + -- ** Rational + , unsafeFromRational , fromRationalRepetend + , fromRationalRepetendLimited + , fromRationalRepetendUnlimited , toRationalRepetend + + -- ** Floating & integer , floatingOrInteger , toRealFloat , toBoundedRealFloat @@ -94,18 +100,15 @@ import Control.Exception (throw, ArithException(DivideByZero)) import Control.Monad (mplus) -import Control.Monad.ST (runST) import Control.DeepSeq (NFData, rnf) import Data.Binary (Binary, get, put) import Data.Char (intToDigit, ord) import Data.Data (Data) -import Data.Function (on) import Data.Hashable (Hashable(..)) import Data.Int (Int8, Int16, Int32, Int64) import qualified Data.Map as M (Map, empty, insert, lookup) import Data.Ratio ((%), numerator, denominator) import Data.Typeable (Typeable) -import qualified Data.Primitive.Array as Primitive import Data.Word (Word8, Word16, Word32, Word64) import Math.NumberTheory.Logarithms (integerLog10') import qualified Numeric (floatToDigits) @@ -116,22 +119,10 @@ import Text.ParserCombinators.ReadP ( ReadP ) import Data.Text.Lazy.Builder.RealFloat (FPFormat(..)) -#if !MIN_VERSION_base(4,8,0) -import Data.Functor ((<$>)) -import Data.Word (Word) -import Control.Applicative ((<*>)) -#endif - -#if MIN_VERSION_base(4,5,0) -import Data.Bits (unsafeShiftR) -#else -import Data.Bits (shiftR) -#endif - -import GHC.Integer (quotRemInteger, quotInteger) -import GHC.Integer.Compat (divInteger) -import Utils (roundTo) +import GHC.Integer.Compat (quotRemInteger, quotInteger, divInteger) +import Utils (maxExpt, roundTo, magnitude) +import Language.Haskell.TH.Syntax (Lift (..)) ---------------------------------------------------------------------- -- Type @@ -156,6 +147,20 @@ -- in 'toDecimalDigits'. -- -- Use 'normalize' to do manual normalization. + -- + -- /WARNING:/ 'coefficient' and 'base10exponent' violate + -- substantivity of 'Eq'. + -- + -- >>> let x = scientific 1 2 + -- >>> let y = scientific 100 0 + -- >>> x == y + -- True + -- + -- but + -- + -- >>> (coefficient x == coefficient y, base10Exponent x == base10Exponent y) + -- (False,False) + -- , base10Exponent :: {-# UNPACK #-} !Int -- ^ The base-10 exponent of a scientific number. @@ -166,50 +171,80 @@ scientific :: Integer -> Int -> Scientific scientific = Scientific - ---------------------------------------------------------------------- -- Instances ---------------------------------------------------------------------- +-- | @since 0.3.7.0 +deriving instance Lift Scientific + instance NFData Scientific where rnf (Scientific _ _) = () +-- | A hash can be safely calculated from a @Scientific@. No magnitude @10^e@ is +-- calculated so there's no risk of a blowup in space or time when hashing +-- scientific numbers coming from untrusted sources. +-- +-- >>> import Data.Hashable (hash) +-- >>> let x = scientific 1 2 +-- >>> let y = scientific 100 0 +-- >>> (x == y, hash x == hash y) +-- (True,True) +-- instance Hashable Scientific where - hashWithSalt salt = hashWithSalt salt . toRational + hashWithSalt salt s = salt `hashWithSalt` c `hashWithSalt` e + where + Scientific c e = normalize s +-- | Note that in the future I intend to change the type of the 'base10Exponent' +-- from @Int@ to @Integer@. To be forward compatible the @Binary@ instance +-- already encodes the exponent as 'Integer'. instance Binary Scientific where - put (Scientific c e) = do - put c - -- In the future I intend to change the type of the base10Exponent e from - -- Int to Integer. To support backward compatability I already convert e - -- to Integer here: - put $ toInteger e - + put (Scientific c e) = put c *> put (toInteger e) get = Scientific <$> get <*> (fromInteger <$> get) +-- | Scientific numbers can be safely compared for equality. No magnitude @10^e@ +-- is calculated so there's no risk of a blowup in space or time when comparing +-- scientific numbers coming from untrusted sources. instance Eq Scientific where - (==) = (==) `on` toRational - {-# INLINABLE (==) #-} - - (/=) = (/=) `on` toRational - {-# INLINABLE (/=) #-} + s1 == s2 = c1 == c2 && e1 == e2 + where + Scientific c1 e1 = normalize s1 + Scientific c2 e2 = normalize s2 +-- | Scientific numbers can be safely compared for ordering. No magnitude @10^e@ +-- is calculated so there's no risk of a blowup in space or time when comparing +-- scientific numbers coming from untrusted sources. instance Ord Scientific where - (<) = (<) `on` toRational - {-# INLINABLE (<) #-} - - (<=) = (<=) `on` toRational - {-# INLINABLE (<=) #-} + compare s1 s2 + | c1 == c2 && e1 == e2 = EQ + | c1 < 0 = if c2 < 0 then cmp (-c2) e2 (-c1) e1 else LT + | c1 > 0 = if c2 > 0 then cmp c1 e1 c2 e2 else GT + | otherwise = if c2 > 0 then LT else GT + where + Scientific c1 e1 = normalize s1 + Scientific c2 e2 = normalize s2 - (>) = (>) `on` toRational - {-# INLINABLE (>) #-} + cmp cx ex cy ey + | log10sx < log10sy = LT + | log10sx > log10sy = GT + | d < 0 = if cx <= (cy `quotInteger` magnitude (-d)) then LT else GT + | d > 0 = if cy > (cx `quotInteger` magnitude d) then LT else GT + | otherwise = if cx < cy then LT else GT + where + log10sx = log10cx + ex + log10sy = log10cy + ey - (>=) = (>=) `on` toRational - {-# INLINABLE (>=) #-} + log10cx = integerLog10' cx + log10cy = integerLog10' cy - compare = compare `on` toRational - {-# INLINABLE compare #-} + d = log10cx - log10cy +-- | /WARNING:/ '+' and '-' compute the 'Integer' magnitude: @10^e@ where @e@ is +-- the difference between the @'base10Exponent's@ of the arguments. If these +-- methods are applied to arguments which have huge exponents this could fill up +-- all space and crash your program! So don't apply these methods to scientific +-- numbers coming from untrusted sources. The other methods can be used safely. instance Num Scientific where Scientific c1 e1 + Scientific c2 e2 | e1 < e2 = Scientific (c1 + c2*l) e1 @@ -264,36 +299,66 @@ "realToFrac_toRealFloat_Float" realToFrac = toRealFloat :: Scientific -> Float #-} --- | /WARNING:/ 'recip' and '/' will diverge (i.e. loop and consume all space) --- when their outputs are . +-- | /WARNING:/ 'recip' and '/' will throw an error when their outputs are +-- . -- --- 'fromRational' will diverge when the input 'Rational' is a repeating decimal. --- Consider using 'fromRationalRepetend' for these rationals which will detect --- the repetition and indicate where it starts. +-- These methods also compute 'Integer' magnitudes (@10^e@). If these methods +-- are applied to arguments which have huge exponents this could fill up all +-- space and crash your program! So don't apply these methods to scientific +-- numbers coming from untrusted sources. +-- +-- 'fromRational' will throw an error when the input 'Rational' is a repeating +-- decimal. Consider using 'fromRationalRepetend' for these rationals which +-- will detect the repetition and indicate where it starts. instance Fractional Scientific where recip = fromRational . recip . toRational - {-# INLINABLE recip #-} - x / y = fromRational $ toRational x / toRational y - {-# INLINABLE (/) #-} + Scientific c1 e1 / Scientific c2 e2 + | d < 0 = fromRational (x / (fromInteger (magnitude (-d)))) + | otherwise = fromRational (x * fromInteger (magnitude d)) + where + d = e1 - e2 + x = c1 % c2 - fromRational rational - | d == 0 = throw DivideByZero - | otherwise = positivize (longDiv 0 0) (numerator rational) + fromRational rational = + case mbRepetendIx of + Nothing -> s + Just _ix -> error $ + "fromRational has been applied to a repeating decimal " ++ + "which can't be represented as a Scientific! " ++ + "It's better to avoid performing fractional operations on Scientifics " ++ + "and convert them to other fractional types like Double as early as possible." where - -- Divide the numerator by the denominator using long division. - longDiv :: Integer -> Int -> (Integer -> Scientific) - longDiv !c !e 0 = Scientific c e - longDiv !c !e !n - -- TODO: Use a logarithm here! - | n < d = longDiv (c * 10) (e - 1) (n * 10) - | otherwise = case n `quotRemInteger` d of - (#q, r#) -> longDiv (c + q) e r + (s, mbRepetendIx) = fromRationalRepetendUnlimited rational - d = denominator rational +-- | Although 'fromRational' is unsafe because it will throw errors on +-- , +-- @unsafeFromRational@ is even more unsafe because it will diverge instead (i.e +-- loop and consume all space). Though it will be more efficient because it +-- doesn't need to consume space linear in the number of digits in the resulting +-- scientific to detect the repetition. +-- +-- Consider using 'fromRationalRepetend' for these rationals which will detect +-- the repetition and indicate where it starts. +unsafeFromRational :: Rational -> Scientific +unsafeFromRational rational + | d == 0 = throw DivideByZero + | otherwise = positivize (longDiv 0 0) (numerator rational) + where + -- Divide the numerator by the denominator using long division. + longDiv :: Integer -> Int -> (Integer -> Scientific) + longDiv !c !e 0 = Scientific c e + longDiv !c !e !n + -- TODO: Use a logarithm here! + | n < d = longDiv (c * 10) (e - 1) (n * 10) + | otherwise = case n `quotRemInteger` d of + (#q, r#) -> longDiv (c + q) e r --- | Like 'fromRational', this function converts a `Rational` to a `Scientific` --- but instead of diverging (i.e loop and consume all space) on + d = denominator rational + +-- | Like 'fromRational' and 'unsafeFromRational', this function converts a +-- `Rational` to a `Scientific` but instead of failing or diverging (i.e loop +-- and consume all space) on -- -- it detects the repeating part, the /repetend/, and returns where it starts. -- @@ -335,7 +400,18 @@ -> Rational -> Either (Scientific, Rational) (Scientific, Maybe Int) -fromRationalRepetend mbLimit rational +fromRationalRepetend mbLimit rational = + case mbLimit of + Nothing -> Right $ fromRationalRepetendUnlimited rational + Just l -> fromRationalRepetendLimited l rational + +-- | Like 'fromRationalRepetend' but always accepts a limit. +fromRationalRepetendLimited + :: Int -- ^ limit + -> Rational + -> Either (Scientific, Rational) + (Scientific, Maybe Int) +fromRationalRepetendLimited l rational | d == 0 = throw DivideByZero | num < 0 = case longDiv (-num) of Left (s, r) -> Left (-s, -r) @@ -345,11 +421,38 @@ num = numerator rational longDiv :: Integer -> Either (Scientific, Rational) (Scientific, Maybe Int) - longDiv n = case mbLimit of - Nothing -> Right $ longDivNoLimit 0 0 M.empty n - Just l -> longDivWithLimit (-l) n + longDiv = longDivWithLimit 0 0 M.empty - -- Divide the numerator by the denominator using long division. + longDivWithLimit + :: Integer + -> Int + -> M.Map Integer Int + -> (Integer -> Either (Scientific, Rational) + (Scientific, Maybe Int)) + longDivWithLimit !c !e _ns 0 = Right (Scientific c e, Nothing) + longDivWithLimit !c !e ns !n + | Just e' <- M.lookup n ns = Right (Scientific c e, Just (-e')) + | e <= (-l) = Left (Scientific c e, n % (d * magnitude (-e))) + | n < d = let !ns' = M.insert n e ns + in longDivWithLimit (c * 10) (e - 1) ns' (n * 10) + | otherwise = case n `quotRemInteger` d of + (#q, r#) -> longDivWithLimit (c + q) e ns r + + d = denominator rational + +-- | Like 'fromRationalRepetend' but doesn't accept a limit. +fromRationalRepetendUnlimited :: Rational -> (Scientific, Maybe Int) +fromRationalRepetendUnlimited rational + | d == 0 = throw DivideByZero + | num < 0 = case longDiv (-num) of + (s, mb) -> (-s, mb) + | otherwise = longDiv num + where + num = numerator rational + + longDiv :: Integer -> (Scientific, Maybe Int) + longDiv = longDivNoLimit 0 0 M.empty + longDivNoLimit :: Integer -> Int -> M.Map Integer Int @@ -362,22 +465,6 @@ | otherwise = case n `quotRemInteger` d of (#q, r#) -> longDivNoLimit (c + q) e ns r - longDivWithLimit :: Int -> Integer -> Either (Scientific, Rational) (Scientific, Maybe Int) - longDivWithLimit l = go 0 0 M.empty - where - go :: Integer - -> Int - -> M.Map Integer Int - -> (Integer -> Either (Scientific, Rational) (Scientific, Maybe Int)) - go !c !e _ns 0 = Right (Scientific c e, Nothing) - go !c !e ns !n - | Just e' <- M.lookup n ns = Right (Scientific c e, Just (-e')) - | e <= l = Left (Scientific c e, n % (d * magnitude (-e))) - | n < d = let !ns' = M.insert n e ns - in go (c * 10) (e - 1) ns' (n * 10) - | otherwise = case n `quotRemInteger` d of - (#q, r#) -> go (c + q) e ns r - d = denominator rational -- | @@ -393,6 +480,11 @@ -- -- * @r < -(base10Exponent s)@ -- +-- /WARNING:/ @toRationalRepetend@ needs to compute the 'Integer' magnitude: +-- @10^^n@. Where @n@ is based on the 'base10Exponent` of the scientific. If +-- applied to a huge exponent this could fill up all space and crash your +-- program! So don't apply this function to untrusted input. +-- -- The formula to convert the @Scientific@ @s@ -- with a repetend starting at index @r@ is described in the paper: -- @@ -443,6 +535,10 @@ nines = m - 1 +-- | /WARNING:/ the methods of the @RealFrac@ instance need to compute the +-- magnitude @10^e@. If applied to a huge exponent this could take a long +-- time. Even worse, when the destination type is unbounded (i.e. 'Integer') it +-- could fill up all space and crash your program! instance RealFrac Scientific where -- | The function 'properFraction' takes a Scientific number @s@ -- and returns a pair @(n,f)@ such that @s = n+f@, and: @@ -566,52 +662,14 @@ -- | Precondition: the 'Scientific' @s@ needs to be an integer: -- @base10Exponent (normalize s) >= 0@ toIntegral :: (Num a) => Scientific -> a -toIntegral (Scientific c e) = fromInteger c * fromInteger (magnitude e) +toIntegral (Scientific c e) = fromInteger c * magnitude e {-# INLINE toIntegral #-} ----------------------------------------------------------------------- --- Exponentiation with a cache for the most common numbers. ----------------------------------------------------------------------- --- | The same limit as in GHC.Float. -maxExpt :: Int -maxExpt = 324 -expts10 :: Primitive.Array Integer -expts10 = runST $ do - ma <- Primitive.newArray maxExpt uninitialised - Primitive.writeArray ma 0 1 - Primitive.writeArray ma 1 10 - let go !ix - | ix == maxExpt = Primitive.unsafeFreezeArray ma - | otherwise = do - Primitive.writeArray ma ix xx - Primitive.writeArray ma (ix+1) (10*xx) - go (ix+2) - where - xx = x * x - x = Primitive.indexArray expts10 half -#if MIN_VERSION_base(4,5,0) - !half = ix `unsafeShiftR` 1 -#else - !half = ix `shiftR` 1 -#endif - go 2 -uninitialised :: error -uninitialised = error "Data.Scientific: uninitialised element" --- | @magnitude e == 10 ^ e@ -magnitude :: Int -> Integer -magnitude e | e < maxExpt = cachedPow10 e - | otherwise = cachedPow10 hi * 10 ^ (e - hi) - where - cachedPow10 = Primitive.indexArray expts10 - - hi = maxExpt - 1 - - ---------------------------------------------------------------------- -- Conversions ---------------------------------------------------------------------- @@ -754,10 +812,21 @@ {-# SPECIALIZE toBoundedInteger :: Scientific -> Maybe Word32 #-} {-# SPECIALIZE toBoundedInteger :: Scientific -> Maybe Word64 #-} --- | @floatingOrInteger@ determines if the scientific is floating point --- or integer. In case it's floating-point the scientific is converted --- to the desired 'RealFloat' using 'toRealFloat'. +-- | @floatingOrInteger@ determines if the scientific is floating point or +-- integer. -- +-- In case it's floating-point the scientific is converted to the desired +-- 'RealFloat' using 'toRealFloat' and wrapped in 'Left'. +-- +-- In case it's integer to scientific is converted to the desired 'Integral' and +-- wrapped in 'Right'. +-- +-- /WARNING:/ To convert the scientific to an integral the magnitude @10^e@ +-- needs to be computed. If applied to a huge exponent this could take a long +-- time. Even worse, when the destination type is unbounded (i.e. 'Integer') it +-- could fill up all space and crash your program! So don't apply this function +-- to untrusted input but use 'toBoundedInteger' instead. +-- -- Also see: 'isFloating' or 'isInteger'. floatingOrInteger :: (RealFloat r, Integral i) => Scientific -> Either r i floatingOrInteger s @@ -885,13 +954,19 @@ -- Pretty Printing ---------------------------------------------------------------------- +-- | See 'formatScientific' if you need more control over the rendering. instance Show Scientific where - show s | coefficient s < 0 = '-':showPositive (-s) - | otherwise = showPositive s + showsPrec d s + | coefficient s < 0 = showParen (d > prefixMinusPrec) $ + showChar '-' . showPositive (-s) + | otherwise = showPositive s where - showPositive :: Scientific -> String - showPositive = fmtAsGeneric . toDecimalDigits + prefixMinusPrec :: Int + prefixMinusPrec = 6 + showPositive :: Scientific -> ShowS + showPositive = showString . fmtAsGeneric . toDecimalDigits + fmtAsGeneric :: ([Int], Int) -> String fmtAsGeneric x@(_is, e) | e < 0 || e > 7 = fmtAsExponent x @@ -961,11 +1036,10 @@ case is of [0] -> '0' :'.' : take dec' (repeat '0') ++ "e0" _ -> - let - (ei,is') = roundTo (dec'+1) is - (d:ds') = map intToDigit (if ei > 0 then init is' else is') - in - d:'.':ds' ++ 'e':show (e-1+ei) + let (ei,is') = roundTo (dec'+1) is + in case map intToDigit (if ei > 0 then init is' else is') of + [] -> "" + d:ds' -> d:'.':ds' ++ 'e':show (e-1+ei) fmtAsFixedDecs :: Int -> ([Int], Int) -> String fmtAsFixedDecs dec (is, e) = @@ -977,11 +1051,10 @@ in mk0 ls ++ (if null rs then "" else '.':rs) else - let - (ei,is') = roundTo dec' (replicate (-e) 0 ++ is) - d:ds' = map intToDigit (if ei > 0 then is' else 0:is') - in - d : (if null ds' then "" else '.':ds') + let (ei,is') = roundTo dec' (replicate (-e) 0 ++ is) + in case map intToDigit (if ei > 0 then is' else 0:is') of + [] -> "" + d:ds' -> d : (if null ds' then "" else '.':ds') where mk0 ls = case ls of { "" -> "0" ; _ -> ls} diff --git a/src/Data/Text/Lazy/Builder/Scientific.hs b/src/Data/Text/Lazy/Builder/Scientific.hs --- a/src/Data/Text/Lazy/Builder/Scientific.hs +++ b/src/Data/Text/Lazy/Builder/Scientific.hs @@ -1,4 +1,4 @@ -{-# LANGUAGE CPP, OverloadedStrings #-} +{-# LANGUAGE OverloadedStrings, Safe #-} module Data.Text.Lazy.Builder.Scientific ( scientificBuilder @@ -16,14 +16,7 @@ import qualified Data.Text as T (replicate) import Utils (roundTo, i2d) -#if MIN_VERSION_base(4,5,0) import Data.Monoid ((<>)) -#else -import Data.Monoid (Monoid, mappend) -(<>) :: Monoid a => a -> a -> a -(<>) = mappend -infixr 6 <> -#endif -- | A @Text@ @Builder@ which renders a scientific number to full -- precision, using standard decimal notation for arguments whose @@ -62,11 +55,10 @@ case is of [0] -> "0." <> fromText (T.replicate dec' "0") <> "e0" _ -> - let - (ei,is') = roundTo (dec'+1) is - (d:ds') = map i2d (if ei > 0 then init is' else is') - in - singleton d <> singleton '.' <> fromString ds' <> singleton 'e' <> decimal (e-1+ei) + let (ei,is') = roundTo (dec'+1) is + in case map i2d (if ei > 0 then init is' else is') of + [] -> mempty + d:ds' -> singleton d <> singleton '.' <> fromString ds' <> singleton 'e' <> decimal (e-1+ei) Fixed -> let mk0 ls = case ls of { "" -> "0" ; _ -> fromString ls} @@ -90,8 +82,7 @@ in mk0 ls <> (if null rs then "" else singleton '.' <> fromString rs) else - let - (ei,is') = roundTo dec' (replicate (-e) 0 ++ is) - d:ds' = map i2d (if ei > 0 then is' else 0:is') - in - singleton d <> (if null ds' then "" else singleton '.' <> fromString ds') + let (ei,is') = roundTo dec' (replicate (-e) 0 ++ is) + in case map i2d (if ei > 0 then is' else 0:is') of + [] -> mempty + d:ds' -> singleton d <> (if null ds' then "" else singleton '.' <> fromString ds') diff --git a/src/GHC/Integer/Compat.hs b/src/GHC/Integer/Compat.hs --- a/src/GHC/Integer/Compat.hs +++ b/src/GHC/Integer/Compat.hs @@ -1,7 +1,14 @@ {-# LANGUAGE CPP #-} +{-# LANGUAGE Trustworthy #-} -module GHC.Integer.Compat (divInteger) where +module GHC.Integer.Compat (divInteger, quotRemInteger, quotInteger) where +import GHC.Integer (quotRemInteger, quotInteger) + +#if MIN_VERSION_base(4,15,0) +import GHC.Integer (divInteger) +#else + #ifdef MIN_VERSION_integer_simple #if MIN_VERSION_integer_simple(0,1,1) @@ -20,4 +27,5 @@ divInteger = div #endif +#endif #endif diff --git a/src/Utils.hs b/src/Utils.hs --- a/src/Utils.hs +++ b/src/Utils.hs @@ -1,14 +1,24 @@ +{-# LANGUAGE CPP #-} {-# LANGUAGE BangPatterns #-} {-# LANGUAGE MagicHash #-} {-# LANGUAGE UnboxedTuples #-} +{-# LANGUAGE Trustworthy #-} +{-# LANGUAGE ScopedTypeVariables #-} module Utils ( roundTo , i2d + , maxExpt + , magnitude ) where import GHC.Base (Int(I#), Char(C#), chr#, ord#, (+#)) +import qualified Data.Primitive.Array as Primitive +import Control.Monad.ST (runST) + +import Data.Bits (unsafeShiftR) + roundTo :: Int -> [Int] -> (Int, [Int]) roundTo d is = case f d True is of @@ -34,3 +44,40 @@ {-# INLINE i2d #-} i2d :: Int -> Char i2d (I# i#) = C# (chr# (ord# '0'# +# i# )) + +---------------------------------------------------------------------- +-- Exponentiation with a cache for the most common numbers. +---------------------------------------------------------------------- + +-- | The same limit as in GHC.Float. +maxExpt :: Int +maxExpt = 324 + +expts10 :: Primitive.Array Integer +expts10 = runST $ do + ma <- Primitive.newArray maxExpt uninitialised + Primitive.writeArray ma 0 1 + Primitive.writeArray ma 1 10 + let go !ix + | ix == maxExpt = Primitive.unsafeFreezeArray ma + | otherwise = do + Primitive.writeArray ma ix xx + Primitive.writeArray ma (ix+1) (10*xx) + go (ix+2) + where + xx = x * x + x = Primitive.indexArray expts10 half + !half = ix `unsafeShiftR` 1 + go 2 + +uninitialised :: error +uninitialised = error "Data.Scientific: uninitialised element" + +-- | @magnitude e == 10 ^ e@ +magnitude :: Num a => Int -> a +magnitude e | e < maxExpt = cachedPow10 e + | otherwise = cachedPow10 hi * 10 ^ (e - hi) + where + cachedPow10 = fromInteger . Primitive.indexArray expts10 + + hi = maxExpt - 1 diff --git a/test/test.hs b/test/test.hs --- a/test/test.hs +++ b/test/test.hs @@ -1,4 +1,3 @@ -{-# LANGUAGE CPP #-} {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE FlexibleContexts #-} {-# LANGUAGE MultiParamTypeClasses #-} @@ -10,16 +9,12 @@ module Main where -#if !MIN_VERSION_base(4,8,0) -import Control.Applicative -#endif import Control.Monad import Data.Int import Data.Word import Data.Scientific as Scientific import Test.Tasty -import Test.Tasty.Runners.AntXML -import Test.Tasty.HUnit (testCase, (@?=), Assertion) +import Test.Tasty.HUnit (testCase, (@?=), Assertion, assertBool) import qualified Test.SmallCheck as SC import qualified Test.SmallCheck.Series as SC import qualified Test.Tasty.SmallCheck as SC (testProperty) @@ -34,10 +29,60 @@ import qualified Data.ByteString.Lazy.Char8 as BLC8 import qualified Data.ByteString.Builder.Scientific as B import qualified Data.ByteString.Builder as B +import Text.ParserCombinators.ReadP (readP_to_S) main :: IO () main = testMain $ testGroup "scientific" - [ smallQuick "normalization" + [ testGroup "DoS protection" + [ testGroup "Eq" + [ testCase "1e1000000" $ assertBool "" $ + (read "1e1000000" :: Scientific) == (read "1e1000000" :: Scientific) + ] + , testGroup "Ord" + [ testCase "compare 1234e1000000 123e1000001" $ + compare (read "1234e1000000" :: Scientific) (read "123e1000001" :: Scientific) @?= GT + ] + + , testGroup "RealFrac" + [ testGroup "floor" + [ testCase "1e1000000" $ (floor (read "1e1000000" :: Scientific) :: Int) @?= 0 + , testCase "-1e-1000000" $ (floor (read "-1e-1000000" :: Scientific) :: Int) @?= (-1) + , testCase "1e-1000000" $ (floor (read "1e-1000000" :: Scientific) :: Int) @?= 0 + ] + , testGroup "ceiling" + [ testCase "1e1000000" $ (ceiling (read "1e1000000" :: Scientific) :: Int) @?= 0 + , testCase "-1e-1000000" $ (ceiling (read "-1e-1000000" :: Scientific) :: Int) @?= 0 + , testCase "1e-1000000" $ (ceiling (read "1e-1000000" :: Scientific) :: Int) @?= 1 + ] + , testGroup "round" + [ testCase "1e1000000" $ (round (read "1e1000000" :: Scientific) :: Int) @?= 0 + , testCase "-1e-1000000" $ (round (read "-1e-1000000" :: Scientific) :: Int) @?= 0 + , testCase "1e-1000000" $ (round (read "1e-1000000" :: Scientific) :: Int) @?= 0 + ] + , testGroup "truncate" + [ testCase "1e1000000" $ (truncate (read "1e1000000" :: Scientific) :: Int) @?= 0 + , testCase "-1e-1000000" $ (truncate (read "-1e-1000000" :: Scientific) :: Int) @?= 0 + , testCase "1e-1000000" $ (truncate (read "1e-1000000" :: Scientific) :: Int) @?= 0 + ] + , testGroup "properFracton" + [ testCase "1e1000000" $ properFraction (read "1e1000000" :: Scientific) @?= (0 :: Int, 0) + , testCase "-1e-1000000" $ let s = read "-1e-1000000" :: Scientific + in properFraction s @?= (0 :: Int, s) + , testCase "1e-1000000" $ let s = read "1e-1000000" :: Scientific + in properFraction s @?= (0 :: Int, s) + ] + ] + , testGroup "toRealFloat" + [ testCase "1e1000000" $ assertBool "Should be infinity!" $ isInfinite $ + (toRealFloat (read "1e1000000" :: Scientific) :: Double) + , testCase "1e-1000000" $ (toRealFloat (read "1e-1000000" :: Scientific) :: Double) @?= 0 + ] + , testGroup "toBoundedInteger" + [ testCase "1e1000000" $ (toBoundedInteger (read "1e1000000" :: Scientific) :: Maybe Int) @?= Nothing + ] + ] + + , smallQuick "normalization" (SC.over normalizedScientificSeries $ \s -> s /= 0 SC.==> abs (Scientific.coefficient s) `mod` 10 /= 0) (QC.forAll normalizedScientificGen $ \s -> @@ -55,10 +100,18 @@ , testCase "reads \"(1.3 )\"" $ testReads "(1.3 )" [(1.3, "")] , testCase "reads \"((1.3))\"" $ testReads "((1.3))" [(1.3, "")] , testCase "reads \" 1.3\"" $ testReads " 1.3" [(1.3, "")] + , testCase "read \" ( (( -1.0e+3 ) ))\"" $ testRead " ( (( -1.0e+3 ) ))" (-1000.0) + , testCase "scientificP \"3\"" $ testScientificP "3" [(3.0, "")] + , testCase "scientificP \"3.0e2\"" $ testScientificP "3.0e2" [(3.0, "e2"), (300.0, "")] + , testCase "scientificP \"+3.0e+2\"" $ testScientificP "+3.0e+2" [(3.0, "e+2"), (300.0, "")] + , testCase "scientificP \"-3.0e-2\"" $ testScientificP "-3.0e-2" [(-3.0, "e-2"), (-3.0e-2, "")] ] , testGroup "Formatting" [ testProperty "read . show == id" $ \s -> read (show s) === s + , testCase "show (Just 1)" $ testShow (Just 1) "Just 1.0" + , testCase "show (Just 0)" $ testShow (Just 0) "Just 0.0" + , testCase "show (Just (-1))" $ testShow (Just (-1)) "Just (-1.0)" , testGroup "toDecimalDigits" [ smallQuick "laws" @@ -91,6 +144,17 @@ -- show d ] + , testGroup "Eq" + [ testProperty "==" $ \(s1 :: Scientific) (s2 :: Scientific) -> + (s1 == s2) == (toRational s1 == toRational s2) + , testProperty "s == s" $ \(s :: Scientific) -> s == s + ] + + , testGroup "Ord" + [ testProperty "compare" $ \(s1 :: Scientific) (s2 :: Scientific) -> + compare s1 s2 == compare (toRational s1) (toRational s2) + ] + , testGroup "Num" [ testGroup "Equal to Rational" [ testProperty "fromInteger" $ \i -> fromInteger i === fromRational (fromInteger i) @@ -213,11 +277,20 @@ ] testMain :: TestTree -> IO () -testMain = defaultMainWithIngredients (antXMLRunner:defaultIngredients) +testMain = defaultMainWithIngredients defaultIngredients testReads :: String -> [(Scientific, String)] -> Assertion testReads inp out = reads inp @?= out +testRead :: String -> Scientific -> Assertion +testRead inp out = read inp @?= out + +testShow :: Maybe Scientific -> String -> Assertion +testShow inp out = show inp @?= out + +testScientificP :: String -> [(Scientific, String)] -> Assertion +testScientificP inp out = readP_to_S Scientific.scientificP inp @?= out + genericIsFloating :: RealFrac a => a -> Bool genericIsFloating a = fromInteger (floor a :: Integer) /= a @@ -391,8 +464,4 @@ bigIntGen = QC.sized $ \size -> QC.resize (size * 1000) intGen intGen :: QC.Gen Int -#if MIN_VERSION_QuickCheck(2,7,0) intGen = QC.arbitrary -#else -intGen = QC.sized $ \n -> QC.choose (-n, n) -#endif