This is a prettyprinting library designed for laying out
plain-text documents. It originated in the pandoc module
Text.Pandoc.Pretty, and its development has been guided by
pandoc’s needs in rendering wrapped textual documents.
In supports wrapping of text on breaking spaces, indentation
and other line prefixes, blank lines, and tabular content.
Example:
Text.DocLayout> mydoc = hang 2 "- " (text "foo" <+> text "bar")
Text.DocLayout> putStrLn $ render (Just 20) mydoc
- foo bar
Text.DocLayout> putStrLn $ render (Just 10) (prefixed "> " (mydoc $+$ mydoc))
> - foo
> bar
>
> - foo
> bar
The Doc type may be parameterized to either String or
(strict or lazy) Text, depending on the desired render target.
Changes
doclayout
0.5
Extract existing HasChars definition into a module
Text.DocLayout.HasChars (Evan Silberman).
Add a build method to the signature of HasChars [API change]
(Evan Silberman). This has a default definition, and is only needed with
the ANSI renderer, so existing users should not need to add anything to
their HasChars instances.
Introduce support for ANSI-styled output (Evan Silberman) [API change].
renderPlain and renderANSI are now exported; the old render
is a synonym of renderPlain. In addition, various functions are
exported to add ANSI formatting (including bold, italics,
underline, strikeout, links, and colors) to a Doc. The Attributed
type is also now exported.
Change type of Block constructor, replacing [a] with [Attributed a],
which carries a Font along with an inner string type [API change]
(Evan Silberman).
Introduce FlatDocs and use them for rendering (Evan Silberman).
This is an internal concept, not part of the public API.
FlatDoc is an “intermediate representation” for the Doc “interpreter”.
The general design is that any Doc can be turned into a list of FlatDocs
that carry equivalent information. The main point of doing this
is to replace the nested Styled and Linked Docs, which form a more
complicated tree structure than previously existed in DocLayout.
Deprecate unfoldD.
0.4.0.1
Add clause for Empty to renderList (#22).
Remove upper bound for criterion
0.4
Expose unfoldD [API change].
Remove realLengthNoShortcut, isEmojiModifier, and
isEmojiJoiner [API change] (Stephen Morgan).
Drop support for ghc <= 8.4, add test for ghc 9.2.
Don’t collapse the CarriageReturn + Newline combination (#20).
We want to ensure that a literal starting with a Newline
doesn’t lose the newline if it occurs after a CarriageReturn.
This affects code blocks in pandoc that begin with newlines.
Improve performance of NoShortcut code (Stephen Morgan).
Simplify emoji processing (Stephen Morgan).
Add benchmarking for code with no shortcuts (Stephen Morgan).
Add unicodeWidth.inc to cabal file.
Fix offset, minOffset, updateColumn so they don’t re-render.
Get unicode block widths directly from the Unicode specification, rather
than writing it out ourselves (Stephen Morgan).
Resolve the width of ambiguous characters based on their context
(Stephen Morgan).
Spacing marks should have nonzero width, even though they are combining
characters (Stephen Morgan).
Add shortcuts for extended Latin, Arabic, Cyrillic, Greek,
Devangari, Bengali, Korean, Telugu, and Tamil (Stephen Morgan).
Fix location of extra-source-files in cabal.
update.hs: require text package
Handle emoji variation modifiers specially, so the keypad emoji can be
ignored (Stephen Morgan). This results in a 16% speedup of realLength
on ascii text.
Add benchmarks for all scripts used by more than 50 million people, plus
a couple more. (#9, Stephen Morgan).
0.3.1.1
Fix the end of the block of zero width characters which contains
the zero-width joiners and directional markings (Stephen Morgan, #5).
This fixes a regression introduced in 0.3.1, affecting code
points 0x2010 to 0x2030.
0.3.1
Improved handling of emojis. Emojis are double-wide, but
previously this library did not treat them as such. We now
have comprehensive support of emojis, including variation
modifiers and zero-width joiners, verified by a test suite.
Performance has been confirmed to be no worse for text without emojis.
(Stephen Morgan, #1). API changes: export realLengthNoShortcut,
isEmojiModifier, isEmojiVariation, isEmojiJoiner.
0.3.0.2
NOINLINE literal instead of fromString (#2, sjakobi).
This produces a further reduction in allocations and
pandoc compile time.
0.3.0.1
NOINLINE fromString (#1).
@sjakobi reports that this change reduced total allocations
for building pandoc-2.12 with GHC 8.10.4 by 8.5% and reduced
peak allocations are reduced from 3854MB to 3389MB.
0.3
Add foldlChar to signature of HasChars [API change].
Use foldlChar in realLength. This avoids a stack overflow
we were getting with long strings in the previous version
(with foldrChar). See jgm/pandoc#6031.
Replace isBlank with isBreakable and improved startsWithBlank.
Previously isBlank was used in the layout algorithm where
what we really wanted was isBreakable.
Avoid unnecessary calculation in updateColumns.
Replace a right fold with a strict left fold.
Add strictness annotations in realLength and updateColumn.
0.2.0.1
Made realLength smarter about combining characters.
If a string starts with a combining character, that character
takes up a width of 1; if the combining character occurs after
another character, it takes 0. See jgm/pandoc#5863.
Improve isBlank, re-use in rendering code for BreakingSpace.
Fixed incorrect Text width in renderig blocks.
0.2
Add instances for Doc: Data, Typeable, Ord, Read, Generic.
Add literal (like text, but polymorphic).
Change some IsString constraints to HasChars.
Add some default definitions for methods in HasChars.
Change offset and minOffset to be more efficient (in
simple cases they no longer render and count line lengths).
Add updateColumn.
Fix problem with lblock/cblock/rblock when chop is
invoked. This caused very strange behavior in which text
got reversed in certain circumstances.
