search

conda-forge / staged-recipes

A place to submit conda recipes before they become fully fledged conda-forge feedstocks
https://conda-forge.org
BSD 3-Clause "New" or "Revised" License
709 stars 4.96k forks source link

Package languagetool #12756

Open bollwyvl opened 4 years ago

bollwyvl commented 4 years ago

languagetool is a grammar/spelling checker, and can be used as a web service or CLI. It apparently only requires openjdk, but ships with support hunspell for spell-checking if detected.

languagetool doesn't provide a default convenience CLI (because who doesn't like java -jar $PREFIX/lib/languagetool-commandline.jar -h), so it may make sense to have an output for languagetool-cli which contains wrapper scripts which set up the *PATH properly.

See also:

asmeurer commented 4 years ago

Another thing to be aware of is for hunspell we ship the dictionary hunspell-en separately. Presumably languagetool has separate language packs as well?

I just did this to match with Linux distros do. I don't know if that's how conda-forge should be handling things. No one seems to have made any packages for other languages. Shipping every language would make the package large (though I actually don't really know how large). I'm actually questioning whether hunspell should just ship with English by default. That would make the default better for most users. Right now, it just prints a message in the post-link if you install hunspell telling you to install hunspell-en (https://github.com/conda-forge/hunspell-feedstock/blob/master/recipe/post-link.sh).

bollwyvl commented 4 years ago

I haven't had a chance to wade into maven-land to start doing this, but good points!

Presumably languagetool has separate language packs as well?

languagetool ships language-specific stuff in-tree, so they probably all just end up in the jar...

No one seems to have made any packages for other languages.

I'd love to see other folks get excited enough about internationalizing docs, UI, etc. to need deeper language tools, and have this problem for real!

I'm actually questioning whether hunspell should just ship with English by default.

Sure, we could move the current hunspell to hunspell-core and make a new hunspell depend on -core and hunspell-en... don't think we'd actually want them in the same package. But then, that's what hunspell-en effectively does...