Stack cmd to produce docset(s) for https://kapeli.com/dash or http://zealdocs.org/

[dysinger]

I would like to be able to quickly generate docsets for zeal (or dash.app) for a project. Offline hotkey search of APIs is handy.

Currently I use the cabal based dash-haskell with some command line foo:
cabal list --installed|grep -e '^\*'|sed 's/\* //g'|parallel --no-notice dash-haskell -o ~/.local/share/Zeal/Zeal/docsets
... but it doesn't yet work with cabal 1.22 and I'm lazily relying on my .zsh_history to remember how to type that.

It would be handy to have stack do this sort of thing for me.

[snoyberg]

If we can get this added to stackage-curator, then we can generate that output for all users with each build. What is the resulting artifact from the above command? Is it a single file, a directory, etc?

[dysinger]

stackage-curator is probably the best place for this (not stack). the output is a directory.

[snoyberg]

Since this will be part of stackage-curator instead, closing.

[PierreR]

What about generating docsets for a local library ?

Any plan to have docsets for all stackage packages ? I am using Zeal but unfortunately Haskell docsets are quite incomplete (and out-of-dated).

Is stackage-curator able to generate docsets currently ?

Updated As a note, the best bet is probably to use https://github.com/philopon/haddocset for now

[angerman]

Bump. What happened to this? Is there a corresponding stockage-curator issue somewhere?

[angerman]

We currently have two docset generators:

• https://github.com/jfeltz/dash-haskell
  and (as @PierreR noted)
• https://github.com/philopon/haddocset
  as well.
  Where haddocset seems to be the slightly more active one. Where I asked about converting the whole documentation database to docset (Generate docset for all of a stackage lts? philopon/haddocset#37) which is out of the haddocset scope though.

[mgsloan]

Considering that stack has haddocks support, it seems consistent to also have something for these tools, if there's demand.

We'd want to do it elsewhere than stackage-curator so that such lookup is available for private / local projects.

[angerman]

You can actually have multiple docsets.
I would go with:

• Stackage (lts-xxx or nightly).docset
• MyProject.docset

Here's basically the haddocset work flow (though incomplete):

$ stack exec -- haddocset -t stackage-lts create
$ stack exec -- haddocset -t stackage-ltd.docset add -f $(stack path --snapshot-pkg-db)/*.conf
$ stack exec -- haddocset -t myproj create
$ stack exec -- haddocset -t myproj.docset add -f $(stack path --local-pkg-db)/*.conf

After that I'm having two docsets, one which contains the stockage docs for the packages I depend on only. (Hence I won't be able to lookup any functionality that would be in stockage, but is not yet a dependency in the cabal file.) And another one which contains my local project and dependencies.
I'd much prefer to get the full stackage docset from stackage and just install it.

I guess I could fake it, by creating a project that depended on all packages in stackage; but this seems like something that should be shared, instead of having everyone compile this for their own.

[PierreR]

When I try it (on nixos) I have got a puzzling error message:

stack exec -- haddocset -t stackage-lts create
[1/5] Create Directory.
[2/5] Writing plist.
[3/5] Migrate Database.
    Global package directory: /nix/store/zzagkgsmb66gxng1819fyqrz7n08l7j4-ghc-7.10.3/lib/ghc-7.10.3/package.conf.d
    Global package count:     25
[4/5] Copy and populate Documents.
    filepath-1.4.0.0 .haddocset: stackage-lts.docset/Contents/Resources/Documents/filepath-1.4.0.0/doc-index.html: commitAndReleaseBuffer: invalid argument (invalid character)

but it works outside a project (the project is nix enable)