- Jul 20, 2017
-
-
Herbert Valerio Riedel authored
Also improves markup and removes/fixes redundant/obsolete parts [skip ci]
-
Herbert Valerio Riedel authored
-
Herbert Valerio Riedel authored
-
-
Alex Biehl authored
* Beef up haddock description * Handle empty lines
-
Herbert Valerio Riedel authored
This .cabal property has long been considered obsolete
-
Herbert Valerio Riedel authored
This addresses https://github.com/haskell/haddock/issues/638#issuecomment-309283297
-
Herbert Valerio Riedel authored
A practical benefit is that we can control the build-depends and also avoid some recompilation between library and test-suite.
-
Herbert Valerio Riedel authored
otherwise we can't test w/ e.g. GHC 7.4.2
-
Herbert Valerio Riedel authored
-
Alex Biehl authored
-
Herbert Valerio Riedel authored
-
- Jul 17, 2017
-
-
- Jul 07, 2017
-
-
Alex Biehl authored
* Avoid missing home module warning * Update haddock-library.cabal
-
Alex Biehl authored
-
- Jul 04, 2017
-
-
* Avoid errors on non UTF-8 Windows Problem ==== haddock exits with errors like below: `(1)` ``` haddock: internal error: <stderr>: hPutChar: invalid argument (invalid character) ``` `(2)` ``` haddock: internal error: Language\Haskell\HsColour\Anchors.hs: hGetContents: invalid argument (invalid byte sequence) ``` `(1)` is caused by printing [the "bullet" character](http://www.fileformat.info/info/unicode/char/2022/index.htm) onto stderr. For example, this warning contains it: ``` Language\Haskell\HsColour\ANSI.hs:62:10: warning: [-Wmissing-methods] • No explicit implementation for ‘toEnum’ • In the instance declaration for ‘Enum Highlight’ ``` `(2)` is caused when the input file of `readFile` contains some Unicode characters. In the case above, '⇒' is the cause. Environment ---- OS: Windows 10 haddock: 2.17.3 GHC: 8.0.1 Solution ==== Add `hSetEncoding handle utf8` to avoid the errors. Note ==== - I found the detailed causes by these changes for debugging: - https://github.com/haskell/haddock/commit/8f29edb6b02691c1cf4c479f6c6f3f922b35a55b - https://github.com/haskell/haddock/commit/1dd23bf2065a1e1f2c14d0f4abd847c906b4ecb4 - These errors happen even after executing `chcp 65001` on the console. According to the debug code, `hGetEncoding stderr` returns `CP932` regardless of the console encoding. * Avoid 'internal error: <stderr>: hPutChar: invalid argument (invalid character)' non UTF-8 Windows Better solution for 59411754a6db41d17820733c076e6a72bcdbd82b's (1)
-
- Jul 03, 2017
-
-
* Ignore .stack-work * Fix for windows: use nul instead of /dev/null * Fix for windows: canonicalize line separator * Also normalize osx line endings
-
Alex Biehl authored
-
- Jul 02, 2017
-
-
* Fix haddock: internal error: links: UnhelpfulSpan This fixes #554 for me. I believe this is another fall out of `wildcard-refactor`, like #549. * Comment to clarify why we take the methods name location
-
- Jun 29, 2017
-
-
Alex Biehl authored
* Make per-argument docs for class methods work again * Test case
-
- Jun 23, 2017
-
-
Alex Biehl authored
These are unecessary anyway and just blow up interface size
-
Alex Biehl authored
So we agree that reexported declarations which do not have subordinates (for example top-level functions) shouldn't have gotten fixities reexported according to the current logic. I wondered why for example Prelude.($) which is obviously reexported from GHC.Base has fixities attached (c.f. http://hackage.haskell.org/package/base-4.9.1.0/docs/Prelude.html#v:-36-). The reason is this: In mkMaps we lookup all the subordinates of top-level declarations, of course top-level functions don't have subordinates so for them the resulting list is empty. In #644 I established the invariant that there won't be any empty lists in the subordinate map. Without the patch from #642 top-level functions now started to fail reexporting their fixities.
-
- Jun 22, 2017
-
-
* Use new function getNameToInstancesIndex instead of tcRnGetInfo There is some significant performance improvement in the ghc testsuite. haddock.base: -23.3% haddock.Cabal: -16.7% haddock.compiler: -19.8% * Remove unused imports
-
- Jun 09, 2017
-
-
* Haddock support for bundled pattern synonyms * Add fixities to bundled pattern synonyms * Add bundled pattern synonyms to the synopsis * Store bundled pattern fixities in expItemFixities * Add test for bundled pattern synonyms * Stop threading fixities * Include bundled pattern synonyms for re-exported data types Sadly, fixity information isn't found for re-exported data types * Support for pattern synonyms * Modify tests after #631 * Test some reexport variations * Also lookup bundled pattern synonyms from `InstalledInterface`s * Check isExported for bundled pattern synonyms * Pattern synonym is exported check * Always look for pattern synonyms in the current module Another overlooked cornercase * Account for types named twice in export lists Also introduce a fast function for nubbing on a `Name` and use it throughout the code base. * correct fixities for reexported pattern synonyms * Fuse concatMap and map * Remove obsolete import * Add pattern synonyms to visible exports * Fix test * Remove corner case
-
- Jun 04, 2017
-
-
Alex Biehl authored
-
- Jun 03, 2017
-
-
Alex Biehl authored
-
- Jun 01, 2017
-
-
Alex Biehl authored
This lead to #629.
-
- May 30, 2017
-
-
Alex Biehl authored
-
Alex Biehl authored
-
- May 27, 2017
-
-
Specialize.hs and Ast.hs are modified to have their Syb code not recurse into Name or Id in HsSyn types. Specialize.hs is refactored to have fewer calls to Syb functions. Syb.hs has some foldl calls replaced with foldl' calls. There is still a lot of performance on the floor of Ast.hs. The RenamedSource is traversed many times, and lookupBySpan is very inefficient. everywhereBut and lookupBySpan dominate the runtime whenever --hyperlinked-source is passed.
-
- May 13, 2017
-
-
Alex Biehl authored
These allow attoparsec to do some clever lookup optimization
-
- May 12, 2017
-
-
Alex Biehl authored
We previously used withFile in conjunction with hGetContents. The list returned by the latter wasn't completely forced by the time we left the withFile block, meaning that we would try to read from a closed handle.
-
- May 11, 2017
-
-
* Improve documentation of Haddock markup. - document that Haddock supports inferring types top-level functions with without type signatures, but also explain why using this feature is discouraged. Looks like this feature has been around since version 2.0.0.0 in 2008! - rework the "Module description" section: - move the general discussion of field formatting to the section intro and add examples illustrating the prose for multiline fields. - mention that newlines are preserved in some multiline fields, but not in others (I also noticed that commas in the `Copyright` field are not preserved; I'll look into this bug later). - add a subsection for the module description fields documentation, and put the field keywords in code formatting (double back ticks) instead of double quotes, to be consistent with the typesetting of keywords in other parts of the documentation. - mention that "Named chunks" are not supported in the long-form "Module description" documentation. - fix formatting of keywords in the "Module attributes" section. Perhaps these errors were left over from an automatic translation to ReST from some other format as part of the transition to using Sphinx for Haddock documentation? Also, add a missing reference here; it just said "See ?"! - update footnote about special treatment for re-exporting partially imported modules not being implemented. In my tests it's not implemented at all -- I tried re-exporting both `import B hiding (f)` and `import B (a, b)` style partial imports, and in both cases got the same result as with full imports `import B`: I only get a module reference. * Rework the `Controlling the documentation structure` section. My main goal was to better explain how to use Haddock without an export list, since that's my most common use case, but I hope I improved the section overall: - remove the incomplete `Omitting the export list` section and fold it into the other sections. In particular, summarize the differences between using and not using an export list -- i.e. control over what and in what order is documented -- in the section lead. - add "realistic" examples that use the structure markup, both with and without an export list. I wanted a realistic example here to capture how it can be useful to explain the relationship between a group of functions in a section, in addition to documenting their individual APIs. - make it clear that you can associate documentation chunks with documentation sections when you aren't using an export list, and that doing it in the most obvious way -- i.e. with `-- |`, as you can in the export list -- doesn't work without an export list. It took me a while to figure this out the first time, since the docs didn't explain it at all before. - add a "no export list" example to the section header section. - add more cross references. * Add examples of gotchas for markup in `@...@`. I'm not sure this will help anyone, since I think most people first learn about `@...@` by reading other people's Haddocks, but I've documented the mistakes which I've made and then gotten confused by. * Use consistent Capitalization of Titles. Some titles were in usual title caps, and others only had the first word capitalized. I chose making them all use title caps because that seems to make the cross references look better.
-
- May 01, 2017
-
-
Alex Biehl authored
(#610) introduced lazy decoding for docs from InstalledInterface but forgot to remove the original calls to get and put_
-
Alex Biehl authored
These are only used in case of a doc reexport so most of the time decoding these is wasted work.
-
- Apr 30, 2017
-
-
Alex Biehl authored
Now the "collapsed" cookie stores which sections have changed state instead of which are collapsed.
-
- Apr 29, 2017
-
-
- Apr 26, 2017
-
-
Alex Biehl authored
-
-