yarn-lock-0.3.2: src/Yarn/Lock/Parse.hs
{-# LANGUAGE NoImplicitPrelude, GeneralizedNewtypeDeriving, OverloadedStrings #-}
{-|
Module : Yarn.Lock.Parse
Description : Parser for yarn.lock files
Maintainer : Profpatsch
Stability : experimental
This module provides a parser for the AST of @yarn.lock@ files.
-}
module Yarn.Lock.Parse
( PackageFields(..), Package
-- * Parsing
-- ** Re-export
, Parser
-- ** Parsers
, packageList
, packageEntry
-- * Internal Parsers
, field, nestedField, simpleField
, packageKeys, packageKey
) where
import Protolude hiding (try)
import qualified Data.Char as Ch
import qualified Data.Text as T
import qualified Data.Map.Strict as M
import Control.Monad (fail)
import Text.Megaparsec as MP hiding (space)
import Text.Megaparsec.Text
import qualified Text.Megaparsec.Lexer as MPL
-- import qualified Data.MultiKeyedMap as MKM
-- import Data.Proxy (Proxy(..))
import qualified Yarn.Lock.Types as YLT
-- | The @yarn.lock@ format doesn’t specifically include a fixed scheme,
-- it’s just an unnecessary custom version of a list of fields.
--
-- An field can either be a string or more fields w/ deeper indentation.
--
-- The actual conversion to semantic structures needs to be done afterwards.
newtype PackageFields = PackageFields (Map Text (Either Text PackageFields))
deriving (Show, Eq, Monoid)
-- | A parsed 'Package' AST has one or more keys, a position in the original files
-- and a collection of fields.
type Package = YLT.Keyed (SourcePos, PackageFields)
-- | Parse a complete yarn.lock into an abstract syntax tree,
-- keeping the source positions of each package entry.
packageList :: Parser [Package]
packageList = many $ (skipMany (comment <|> eol)) *> packageEntry
where comment = char '#' *> manyTill anyChar eol
-- | A single Package.
--
-- Example:
--
-- @
-- handlebars@^4.0.4:
-- version "4.0.6"
-- resolved "https://registry.yarnpkg.com/handlebars/-/handlebars-4.0.6.tgz#2ce4484850537f9c97a8026d5399b935c4ed4ed7"
-- dependencies:
-- async "^1.4.0"
-- optimist "^0.6.1"
-- source-map "^0.4.4"
-- optionalDependencies:
-- uglify-js "^2.6"
-- "
-- @
packageEntry :: Parser (YLT.Keyed (SourcePos, PackageFields))
packageEntry = label "package entry" $ do
pos <- getPosition
-- A package entry is a non-indented
(keys, pkgs) <- nonIndented
-- block that has a header of package keys
-- and an indented part that contains fields
$ indentedFieldsWithHeader packageKeys
pure $ YLT.Keyed keys (pos, pkgs)
-- | The list of PackageKeys that index the same Package
--
-- @
-- align-text@^0.1.1, align-text@^0.1.3:\\n
-- @
packageKeys :: Parser [YLT.PackageKey]
packageKeys = label "package keys" $ do
firstEls <- many (try $ lexeme $ packageKey ":," <* char ',')
lastEl <- packageKey ":" <* char ':'
pure $ firstEls ++ [lastEl]
-- | A packageKey is @\<package-name\>\@\<semver\>@;
--
-- If the semver contains spaces, it is also quoted with @"@.
packageKey :: [Char] -> Parser YLT.PackageKey
packageKey separators = inString (pkgKey "\"")
-- if no string delimiters is used we need to check for the separators
-- this file format is shit :<
<|> pkgKey separators
<?> "package key"
where
pkgKey :: [Char] -> Parser YLT.PackageKey
pkgKey valueChars = label "package key" $ do
key <- someTextOf (noneOf valueChars)
-- okay, here’s the rub:
-- `@` is used for separation, but package names containing `@`
-- are totally a thing. As first character, as well.
-- As are package keys without any semver whatsoever, thus ending
-- with `@`.
-- Did I mention this file format is big chunks of elephant shit?
case (\(n, v) -> (T.dropEnd 1 n, v)) $ T.breakOnEnd "@" key of
("", _) -> fail "packageKey: package name can not be empty"
(n, "") -> pure $ YLT.PackageKey n ""
(n, v) -> pure $ YLT.PackageKey n v
-- | Either a simple or a nested field.
field :: Parser (Text, Either Text PackageFields)
field = try nested <|> simple <?> "field"
where
simple = fmap Left <$> simpleField
nested = fmap Right <$> nestedField
-- | A key-value pair, separated by space.
-- Key any value may be enclosed in "".
-- Returns key and value.
simpleField :: Parser (Text, Text)
simpleField = (,) <$> lexeme (strSymbolChars <|> symbolChars)
-- valueChars may be in Strings or maybe not >:
-- this file format is absolute garbage
<*> (strValueChars <|> valueChars)
<?> "simple field"
where
valueChars, strValueChars :: Parser Text
valueChars = someTextOf (noneOf ("\n\r\"" :: [Char]))
strSymbolChars = inString $ symbolChars
strValueChars = inString $ valueChars
-- as with packageKey semvers, this can be empty
<|> (pure T.empty <?> "an empty value field")
-- | Similar to a @simpleField@, but instead of a string
-- we get another block with deeper indentation.
nestedField :: Parser (Text, PackageFields)
nestedField = label "nested field" $
indentedFieldsWithHeader (symbolChars <* char ':')
-- internal parsers
-- | There are two kinds of indented blocks:
-- One where the header is the package
-- and one where the header is already a package field key.
indentedFieldsWithHeader :: Parser a -> Parser (a, PackageFields)
indentedFieldsWithHeader header = indentBlock $ do
-- … block that has a header of package keys
hdr <- header
-- … and an indented part that contains fields
pure $ MPL.IndentSome Nothing
(\fields -> pure (hdr, toPfs fields)) field
where
toPfs :: [(Text, Either Text PackageFields)] -> PackageFields
toPfs = PackageFields . M.fromList
-- | Characters allowed in key symbols.
--
-- TODO: those are partly npm package names, so check the allowed symbols, too.
--
-- Update: npm doesn’t specify the package name format, at all.
-- Apart from the length.
symbolChars :: Parser Text
symbolChars = label "key symbol" $ someTextOf $ satisfy
(\c -> Ch.isAscii c &&
(Ch.isLower c || Ch.isUpper c || Ch.isNumber c || c `elem` special))
where special = "-_.@/" :: [Char]
-- text versions of parsers & helpers
someTextOf :: Parser Char -> Parser Text
someTextOf c = T.pack <$> some c
-- | parse everything as inside a string
inString :: Parser a -> Parser a
inString = between (char '"') (char '"')
-- lexers
-- | Parse whitespace.
space :: Parser ()
space = MPL.space (void MP.spaceChar)
(MPL.skipLineComment "# ")
(void $ satisfy (const False))
-- | Parse a lexeme.
lexeme :: Parser a -> Parser a
lexeme = MPL.lexeme space
-- | Ensure parser is not indented.
nonIndented :: Parser a -> Parser a
nonIndented = MPL.nonIndented space
indentBlock :: ParsecT Dec Text Identity (MPL.IndentOpt (ParsecT Dec Text Identity) a b)
-> ParsecT Dec Text Identity a
indentBlock = MPL.indentBlock space