Fix repository's incorrect language-classification on GitHub.com

[Alhadis]

Currently, GitHub classifies haskell-ci as a Roff repository:

This stems from the abundance of collections/cabal.project.* files with numeric suffixes, which GitHub mistakes as man pages. This PR is a small patch to correct that using an override.

NOTE: I'm assuming the files contained in collections/ are Haskell source. If they're not, please let me know and I'll mark them as ordinary Text, instead of being any particular language.

[phadej]

Technically they are Cabal configuration, so it's more like EDN or JSON than Clojure or JavaScript. I guess linguist doesn't know about Cabal files?

…

Sent from my iPhone

On 13 Feb 2019, at 15.02, John Gardner ***@***.***> wrote: Currently, GitHub classifies haskell-ci as a Roff repository: This stems from the abundance of collections/cabal.project.* files with numeric suffixes, which GitHub mistakes as man pages. This PR is a small patch to correct that using an override. NOTE: I'm assuming the files contained in collections/ are Haskell source. If they're not, please let me know and I'll mark them as ordinary Text, instead of being any particular language. You can view, comment on, or merge this pull request online at: #223 Commit Summary Fix incorrect language classification on GitHub File Changes A .gitattributes (2) Patch Links: https://github.com/haskell-CI/haskell-ci/pull/223.patch https://github.com/haskell-CI/haskell-ci/pull/223.diff — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[Alhadis]

I guess linguist doesn't know about Cabal files?

Erm, it actually doesn't. That actually needs to be fixed upstream; I'll look into submitting a PR to have Cabal added as a language.

Since I'm (clearly) Haskell illiterate, would you be able to tell me what filenames and file extensions should be associated with Cabal?

Regarding this PR: I'll mark the files as Cabal in advance, but also linguist-vendored, which will instruct Linguist not to take them into account when calculating the language stats.

[Alhadis]

@phadej Alright, PR for Cabal support has been submitted upstream. 👍 Once it's merged and released in a new release of Linguist, the linguist-vendored attribute added by this PR can be dropped.

[phadej]

@Alhadis extensions are a bit tricky

• foo.cabal are package definition files
• and additionally we have cabal.project, cabal.haskell-ci are files which share the same syntax
  • because the only <pkg>.cabal file in the directory is the definition of the package
  • .... so extension can be suffix or prefix

[phadej]

Yet, handling only *.cabal files is fine for Cabal, I guess.

[phadej]

I guess this is fine for now, we can discuss Cabal + linguist in your PR.

[Alhadis]

Thanks! 👍

As mentioned at github/linguist#4419, we don't have a way of matching prefixes, only suffixes or entire filenames. 😢 Having said that, I'd recommend adding this to .gitattributes too:

cabal.haskell-ci linguist-language=Cabal

This will activate correct syntax highlighting for the file once the PR is live on GitHub.

[phadej]

Fwiw, cabal.haskell-ci is (hopefully) become common-ish filename too, as people use this tool. (only 16 occurences atm)

[Alhadis]

I ran a search for that, but it only yielded 16 results. We require languages/filetypes to have considerable distribution in-the-wild before supporting them officially (at least 100 unique repos/users).