Use of keywords in DESCRIPTION and in search

[maelle]

I've tried finding https://cran.r-project.org/web/packages/paws/index.html via pkgsearch::ps("aws", size = 100). It didn't appear. I assume it's because the aws abbreviation is only present in an URL? Is this an example of why authors should think about keywords folks might use to find their package? E.g. in this case would it be sufficient to replace "Interface to Amazon Web Services" with "Interface to Amazon Web Services ('AWS')" in the package Description?

[gaborcsardi]

Yeah, keywords would be great.

…

On Tue, May 21, 2019 at 8:53 AM Maëlle Salmon ***@***.***> wrote: I've tried finding https://cran.r-project.org/web/packages/paws/index.html via pkgsearch::ps("aws", size = 100). It didn't appear. I assume it's because the aws abbreviation is only present in an URL? Is this an example of why authors should think about keywords folks might use to find their package? E.g. in this case would it be sufficient to replace "Interface to Amazon Web Services" with "Interface to Amazon Web Services ('AWS')" in the package Description? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[maelle]

So it's something the authors should pay attention to, right? It's something I'll mention in the blog post but I wonder whether it could/should be part of the package docs too. 🤔

[gaborcsardi]

IDK, what is the best way to put keywords into DESCRIPTION?

[maelle]

You can use X.schema.org stuff cf e.g. https://github.com/ropensci/codemetar/blob/0630dc38c48f594b48ad708707248a37d129dba6/DESCRIPTION#L100 and CRAN won't complain (thanks to @cboettig). Such info doesn't appear on the CRAN page, but could the search API use it?

[gaborcsardi]

The search server indexes all fields, so yeah they will be found. But should we advertise this as the correct way to add keywords? If this is the standard way, then we should up-weight the search hits for this field.

[maelle]

there's no standard way AFAIK. Maybe we should rather recommend to pack as many keywords as possible in Description.

[gaborcsardi]

I think having an explicit keyword field would be nice.

[gaborcsardi]

But until then we can also use the X.schema.org stuff....

[maelle]

especially as it was already CRAN-approved. In codemetar we even discussed a possible helper function that'd take GitHub topics of a repo and write them as X.schema.org keywords in DESCRIPTION cf ropensci/codemetar#152 but I'm not sure where such a function should live, probably not pkgsearch.

[maelle]

Interestingly at the time of that convo in codemetar I noted that having (schema.org) keywords in DESCRIPTION was useless if not for inclusion in the codemeta.json (whereas GitHub topics are useful or at least can be, for folks who browse topics), but if the search API used those, then keywords in DESCRIPTION would be useful for sure!

[maelle]

at the moment not that many results for https://github.com/search?p=1&q=X-schema.org-keywords+user%3Acran&type=Code, and most of them seem to be rOpenSci packages.

I'm not sure we can revolutionize CRAN before the next pkgsearch release but in the release blog post I could show the use of advanced search on that field and mention package developers could use this field for keywords. 🤔

[gaborcsardi]

Closing, as it does not seem to be actionable.