I151 use GitHub topics as keywords

[maelle]

No description provided.

[maelle]

Appveyor build fail is very annoying but it's a false positive, many packages would apparently need to be reinstalled.

[katrinleinweber]

Dear @maelle,

Do I understand correctly, that this code combines the GitHub Topics and the manually specified X-schema.org-keywords as the keywords list in codemeta.json?

I was initially thinking, that codemetar can copy the repo's topics into the DESCRIPTION's X-schema.org-keywords list, but that's not the case, is it?

Thanks and greetings :-)

[maelle]

Yes keywords are not copied in DESCRIPTION, because the only reason to keep keywords in DESCRIPTION was to add them to codemeta.json, there's no other use cases for them that I know of? What do you think? Happy to hear any suggestions!

I find that adding keywords as repo topics is more natural, and I'd be glad to add whatever the equivalent is on other platforms (Gitlab &co)

[katrinleinweber]

I totally agree :-) Just wanted to double-check before I accidentally teach codemetar's approach to schema.org the wrong way around. Thanks a lot for the reply!

I could imagine that it might in some cases be useful to populate both DESCRIPTION and codemeta.json from only the repo topics. Once GitLab has them as well, that approach may be more useful than manually curating the keywords in two places. Or, the repo providers could try and extract the topics from DESCRIPTION.

[maelle]

Thanks for starting the discussion!

Currently I'd actually advise to store no keyword in DESCRIPTION when the repo is on GitHub. I was imagining that one could re-populate the place for topics on another platform (say Gitlab after a migration) from codemeta.json?

Cc @cboettig

[katrinleinweber]

Hm, but most of ROpenSci's packages seem to contain X-schema.org-keywords. Are those added manually? I followed the trail to #32.

[maelle]

Most? I think only @cboettig and @sckott do that, they started doing that before I implemented the repo topic ingestion I think? Or they prefer the DESCRIPTION approach maybe 😉

[sckott]

I don't see any equivalent to github topics in gitlab (don't know about bitbucket). Do you?

[katrinleinweber]

There are "tags" under a project's Settings > General, see https://gitlab.com/gitlab-org/gitlab-ce/issues/2978. However, just staying on GitHub: Can they be auto-generated somehow? I'm asking for TIBHannover/BacDiveR#86 ;-) Thanks for any hints!

[maelle]

You mean auto-generating GitHub repo topics? No they can't but GitHub gives you suggestions. The way I've added topics to repos was using what I identified as keywords, and what other repos seem to be using (e.g. "rstats", "r" and "r-package" for R packages, and the different forms of library names for say magick which interfaces Image Magick). I am not sure whether topics help SEO, but they sure help browsing e.g. https://github.com/topics/bacterial-database

If you mean auto-generating DESCRIPTION keywords from GitHub repo topics/codemeta.json, we don't have that yet. codemeta.json will contain the repo topics and the keywords from DESCRIPTION (removing duplicates) but things only work in this direction.

We could add code to generate DESCRIPTION schema.org keywords from GitHub repo topics, but I'm just not sure it'd be useful? Since the only use of DESCRIPTION schema.org keywords is as far as I know to be fed into codemeta.json? I'm happy to have my mind changed though. 😉

[katrinleinweber]

I meant the latter: generating the X-schema.org-keywords from GitHub Topics. Because

GitHub gives you suggestions

and that makes it IMHO the best place to define them manually, but only once :-)

but I'm just not sure it'd be useful?

Me neither. Presumably, Google can extract the GitHub topics directly. But seeing all the X-schema.org that have been added around here, it surely would be better automatically, than manually, no? Or are those lines part of some cookiecutter/template? I looked through https://github.com/ropensci/dev_guide/search?q=keywords&unscoped_q=keywords.

[maelle]

Not sure who added these X-schema.org stuff, was it you @sckott and how did you do that?

[sckott]

Carl introduced me to them, so I just kept going with them - @cboettig would know more

[cboettig]

Yup, X- is CRAN's approved way of adding additional fields that come from approved namespaces, and Kurt approved X-schema.org for me in email. The thought here was that a DESCRIPTION file is still the definitive metadata record for an R package (after all, all R packages must have them!) and so that ideally any metadata would be added and maintained there (after all, codemetar is also all about not having to write/edit the codemeta.json file manually). So this approach seemed a good way to add any codemeta fields that didn't already crosswalk into DESCRIPTION in some other way, -- keywords is just one example.

[katrinleinweber]

OK, thanks for explaining. But, they were added manually, or copy pasted, but not automatically transferred from some source (file or GitHub topics), correct?

[cboettig]

correct, they were added manually to the DESCRIPTION