diff --git a/CHANGELOG.md b/CHANGELOG.md --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,3 +1,17 @@ +## MMark 0.0.5.1 + +* The parser can now recover from block-level parse errors in tables and + continue parsing. + +* Pipes in code spans in table cells are not considered as table cell + delimiters anymore. + +* Table sub-parser now faster rejects inputs that do not look like a table, + this improves overall performance. + +* Better handling of the cases when a block can be interpreted as a list and + as a table at the same time. + ## MMark 0.0.5.0 * Documentation improvements. diff --git a/README.md b/README.md --- a/README.md +++ b/README.md @@ -268,11 +268,11 @@ Library | Parsing library | Execution time | Allocated | Max residency --------------------|---------------------|---------------:|------------:|-------------: -`cmark-0.5.6` | Custom C code | 316.1 μs | 228,440 | 9,608 -`mmark-0.0.4.3` | Megaparsec | 7.280 ms | 32,207,640 | 37,856 -`cheapskate-0.1.1` | Custom Haskell code | 10.30 ms | 44,686,272 | 799,200 -`markdown-0.1.16` † | Attoparsec | 13.58 ms | 69,261,816 | 699,656 -`pandoc-2.0.5` | Parsec | 36.04 ms | 141,868,840 | 1,471,080 +`cmark-0.5.6` | Custom C code | 323.4 μs | 228,440 | 9,608 +`mmark-0.0.5.1` | Megaparsec | 7.027 ms | 26,180,272 | 37,792 +`cheapskate-0.1.1` | Custom Haskell code | 10.76 ms | 44,686,272 | 799,200 +`markdown-0.1.16` † | Attoparsec | 14.13 ms | 69,261,816 | 699,656 +`pandoc-2.0.5` | Parsec | 37.90 ms | 141,868,840 | 1,471,080 *Results are ordered from fastest to slowest.* diff --git a/Text/MMark/Parser.hs b/Text/MMark/Parser.hs --- a/Text/MMark/Parser.hs +++ b/Text/MMark/Parser.hs @@ -170,11 +170,11 @@ [ Just <$> pThematicBreak , Just <$> pAtxHeading , Just <$> pFencedCodeBlock + , Just <$> pTable , Just <$> pUnorderedList , Just <$> pOrderedList , Just <$> pBlockquote , pReferenceDef - , Just <$> pTable , Just <$> pParagraph ] _ -> Just <$> pIndentedCodeBlock @@ -441,6 +441,8 @@ pTable :: BParser (Block Isp) pTable = do (n, headerRow) <- try $ do + pos <- L.indentLevel + option False (T.any (== '|') <$> lookAhead nonEmptyLine) >>= guard let pipe' = option False (True <$ pipe) l <- pipe' headerRow <- NE.sepBy1 cell (try (pipe <* notFollowedBy eol)) @@ -448,18 +450,22 @@ let n = NE.length headerRow guard (n > 1 || l || r) eol <* sc' + L.indentLevel >>= \i -> guard (i == pos || i == (pos <> pos1)) lookAhead nonEmptyLine >>= guard . isHeaderLike return (n, headerRow) - caligns <- rowWrapper (NE.fromList <$> sepByCount n calign pipe) - otherRows <- many $ do - lookAhead (option True (isBlank <$> nonEmptyLine)) >>= guard . not - rowWrapper (NE.fromList <$> sepByCount n cell pipe) - Table caligns (headerRow :| otherRows) <$ sc + withRecovery recover $ do + sc' + caligns <- rowWrapper (NE.fromList <$> sepByCount n calign pipe) + otherRows <- many $ do + endOfTable >>= guard . not + rowWrapper (NE.fromList <$> sepByCount n cell pipe) + Table caligns (headerRow :| otherRows) <$ sc where cell = do startPos <- getPosition txt <- fmap (T.stripEnd . T.pack) . foldMany' . choice $ [ (++) . T.unpack <$> hidden (string "\\|") + , (++) . T.unpack <$> pCodeSpanB , (:) <$> label "inline content" (satisfy cellChar) ] return (IspSpan startPos txt) cellChar x = x /= '|' && notNewline x @@ -488,6 +494,13 @@ 8 % 10 isHeaderConstituent x = isSpace x || x == '|' || x == '-' || x == ':' + endOfTable = + lookAhead (option True (isBlank <$> nonEmptyLine)) + recover err = + Naked (IspError (replaceEof "end of table block" err)) <$ + manyTill + (optional nonEmptyLine) + (endOfTable >>= guard) <* sc -- | Parse a paragraph or naked text (is some cases). @@ -530,6 +543,23 @@ (IspSpan startPos (assembleParagraph (l:ls []))) <$ sc ---------------------------------------------------------------------------- +-- Auxiliary block-level parsers + +-- | 'match' a code span, this is a specialised and adjusted version of +-- 'pCodeSpan'. + +pCodeSpanB :: BParser Text +pCodeSpanB = fmap fst . match . hidden $ do + n <- try (length <$> some (char '`')) + let finalizer = try $ do + void $ count n (char '`') + notFollowedBy (char '`') + skipManyTill (label "code span content" $ + takeWhile1P Nothing (== '`') <|> + takeWhile1P Nothing (\x -> x /= '`' && notNewline x)) + finalizer + +---------------------------------------------------------------------------- -- Inline parser -- | The top level inline parser. @@ -581,6 +611,8 @@ else pPlain -- | Parse a code span. +-- +-- See also: 'pCodeSpanB'. pCodeSpan :: IParser Inline pCodeSpan = do diff --git a/mmark.cabal b/mmark.cabal --- a/mmark.cabal +++ b/mmark.cabal @@ -1,5 +1,5 @@ name: mmark -version: 0.0.5.0 +version: 0.0.5.1 cabal-version: >= 1.18 tested-with: GHC==7.10.3, GHC==8.0.2, GHC==8.2.2 license: BSD3 diff --git a/tests/Text/MMarkSpec.hs b/tests/Text/MMarkSpec.hs --- a/tests/Text/MMarkSpec.hs +++ b/tests/Text/MMarkSpec.hs @@ -1805,7 +1805,7 @@ let o = "\n\n\n\n\n\n\n
Foo
foo
\n" "|Foo\n---\nfoo" ==-> o "Foo|\n---\nfoo" ==-> o - "| Foo |\n --- \n foo " ==-> o + "| Foo |\n --- \n foo " ==-> o "| Foo |\n| --- |\n| foo |" ==-> o it "reports correct parse errors when parsing the header line" $ (let s = "Foo | Bar\na-- | ---" @@ -1824,7 +1824,7 @@ "

Foo | Bar\nab- | ---

\n" it "demands that number of columns in rows match number of columns in header" $ (let s = "Foo | Bar | Baz\n--- | --- | ---\nfoo | bar" - in s ~-> err (posN 41 s) (ueof <> etok '|' <> eic)) + in s ~-> err (posN 41 s) (ulabel "end of table block" <> etok '|' <> eic)) >> (let s = "Foo | Bar | Baz\n--- | --- | ---\nfoo | bar\n\nHere it goes." in s ~-> err (posN 41 s) (utok '\n' <> etok '|' <> eic)) @@ -1837,6 +1837,21 @@ it "escaped pipes do not fool position tracking" $ let s = "Foo | Bar\n--- | ---\n\\| *fo | bar" in s ~-> err (posN 26 s) (ueib <> etok '*' <> elabel "inline content") + it "pipes in code spans in headers do not fool the parser" $ + "`|Foo|` | `|Bar|`\n--- | ---\nfoo | bar" ==-> + "\n\n\n\n\n\n\n
|Foo||Bar|
foobar
\n" + it "pipes in code spans in cells do not fool the parser" $ + "Foo | Bar\n--- | ---\n`|foo|` | `|bar|`" ==-> + "\n\n\n\n\n\n\n
FooBar
|foo||bar|
\n" + it "multi-line code spans are disallowed in table headers" $ + "`Foo\nBar` | Bar\n--- | ---\nfoo | bar" ==-> + "

Foo Bar | Bar\n--- | ---\nfoo | bar

\n" + it "multi-line code spans are disallowed in table cells" $ + let s = "Foo | Bar\n--- | ---\n`foo\nbar` | bar" + in s ~~-> + [ err (posN 24 s) (utok '\n' <> etok '`' <> elabel "code span content") + , err (posN 35 s) (ueib <> etok '`' <> elabel "code span content") + ] it "parses tables with just header row" $ "Foo | Bar\n--- | ---" ==-> "\n\n\n\n\n\n
FooBar
\n" @@ -1849,6 +1864,22 @@ [ err (posN 10 s) (ueib <> etok '*' <> eic) , err (posN 26 s) (ueib <> etok '_' <> eic) , errFancy (posN 32 s) (nonFlanking "_") ] + it "tables have higher precedence than unordered lists" $ do + "+ foo | bar\n------|----\n" ==-> + "\n\n\n\n\n\n
+ foobar
\n" + "+ foo | bar\n -----|----\n" ==-> + "\n\n\n\n\n\n
+ foobar
\n" + it "tables have higher precedence than ordered lists" $ do + "1. foo | bar\n-------|----\n" ==-> + "\n\n\n\n\n\n
1. foobar
\n" + "1. foo | bar\n ------|----\n" ==-> + "\n\n\n\n\n\n
1. foobar
\n" + it "if table is indented inside unordered list, it's put there" $ + "+ foo | bar\n ----|----\n" ==-> + "\n" + it "if table is indented inside ordered list, it's put there" $ + "1. foo | bar\n ----|----\n" ==-> + "
    \n
  1. \n\n\n\n\n\n\n
    foobar
    \n
    2. \n
\n" it "renders a comprehensive table correctly" $ withFiles "data/table.md" "data/table.html" context "multiple parse errors" $ do