add speech-tagger recipe

[cosmicexplorer]

speech-tagger is an extension to tag parts of speech in an emacs buffer
using stanford coreNLP. It can be found at
https://github.com/cosmicexplorer/speech-tagger. I am the maintainer of
this package.

[purcell]

Thanks, looks good. I'd strongly encourage you to follow the elisp coding conventions for symbol naming, though, because the clojure-style naming here is not at all idiomatic.

Additionally, triple-semicolon comments have a specific meaning, so only use them for standard section headings, e.g. Commentary:, Code:, Usage:, plus the first and last comment lines.

Hope that helps,

-Steve

[cosmicexplorer]

Fixed the clojure naming and kept triple-semicolon lines to bookends only. Thanks for the link to those conventions.

I've also made a change (which you can see in this pull request) to the packaging; instead of downloading the jar at runtime, it pulls it in as part of the package, which hopefully takes fewer people by surprise.

[cosmicexplorer]

Just fixed an issue where the "loading" overlay wasn't being removed by the clear-tags function, and made the jar path a defcustom to facilitate easier development.

[purcell]

Sorry for the slow feedback. The .el file looks super now, but we'd strongly prefer to build from your dev branch, not a build branch that is manually updated and may one day get stale without anybody noticing. So to that extent, it would be preferable for the code to come from master, and for the package to download the jar at runtime (or, probably, load time).

[cosmicexplorer]

Alright, that's what I had before. I'm not familiar with eval-when but I'll make it check the jar path for existence and download if not. That does make versioning the jar a tad more difficult, I'll try to see if I can quickly add a hashing scheme to detect versioning without too much hackery.

[cosmicexplorer]

Ok, it's proving harder than I thought to download a file over http; url-retrieve-synchronously is either dropping bytes, or emacs is converting the coding system incorrectly or something, but I'm still working on this. I could make every speech tag command shoot a network request to my own server, and then I wouldn't have to deal with this at all, but the latency would be super annoying. I'll comment again when I can make this work. Thanks a lot for your help.

[cosmicexplorer]

Ok, turned out the coding system was the problem; specifying 'raw-text as the coding system in the buffer returned by url-retrieve-synchronously seems to have resolved the hashing issues. I've run this through a few tests on an empty emacs and it seems to work just fine.

[purcell]

Cool, thanks - merged. :-)