Talk:Searching R Packages

Learning Project

This should be moved under R (programming language). People looking for R won't be able to find it if the title starts with Searching, and most people looking for search information aren't looking for information on R programming. -- Dave Braunschweig (discuss • contribs) 15:47, 24 January 2018 (UTC)

Yes. How do you suggest that be accomplished? Thanks, DavidMCEddy (discuss • contribs) 18:39, 24 January 2018 (UTC)

@Dave Braunschweig: I added a link from R (programming language) to here, inviting readers to contribute here to a proposal for improving existing search capabilities that might be funded by, e.g., the R Foundation. If you have other ideas, please let me know.

However, because this is intended to be a platform for developing a proposal -- as well as a tutorial on existing search capabilities -- I think the current article needs to be a separate article. DavidMCEddy (discuss • contribs) 05:12, 25 January 2018 (UTC)

General comments

(Ben Bolker) --24.36.66.178 (discuss) 16:59, 27 January 2018 (UTC)

• I find sos output much less useful than it used to be (previously gave details of functions, now reports only for functions in installed packages?)

Could you please run a test and look again? If I understand your concern, the web page with detail by help page sorted by package should still be there, but concealed by a second page with one row for each package.

I changed it 2017-07-03 to produce TWO web pages, one with detail by help page as before, the second a package summary: The package summary has always been available, but it had previously required more work to get it. I apologize if that "improvement" wasn't an improvement. What would you suggest we do about that? Revert that "improvement"?

(BMB) My bad! New users are more likely to read the documentation, so they're less likely than I to get to confused. This is great news (and I'll point it out explicitly when I'm teaching)

Someone suggested a way to get the detail I wanted by package without installing them. Sadly, I misplaced the email with that suggestion. If you know how to get that additional information, please let me know. If you know how and would like to help with that, please "Request to join" at "https://r-forge.r-project.org/projects/rsitesearch/". I will approve it instantly from you, Ben.

However, I don't plan to work on that in the short term. Longer term, I hope we'll get a project funded, with you, me, and others writing specs.

(BMB) Sorry, don't know (seems hard?)

(sg): Just read "https://cloud.r-project.org/", find the packages of interest, then read the web page for each package desired and scrape what is needed from that. I've done that kind of thing before. I didn't think if it until after I wrote the line above. DavidMCEddy (discuss • contribs) 19:37, 28 January 2018 (UTC)

• Strongly prefer to have a command-line feature continue to be available (i.e. not just RStudio add-ins)

I agree -- except that if at most 2.2 percent of users actually use this, it may be hard to get someone else to pay for someone to recode "sos" to use an improved and common database. We might have to claim that the survey was biased. Otherwise, we'd need a volunteer like you to improve it, etc.

(BMB) Agree that we should make this a desideratum. I don't think it's too hard to argue for a flexible/open API (i.e., not just "make command-line users happy", but "provide utility for future, unforeseen interfaces to the information")

• idea of having usage monitoring has been discussed many times in the past; R-core has always been reacted with somewhere between "meh" and "no" (concerns about privacy, difficulties of interpreting convenience sample, difficulty of creating a good implementation, etc.)

Yes, but now with Metacran, crantastic, CRANsearcher, RDocumentation, and depsy all providing additional information, if we can pull all that together into one common database with good tools for extracting what someone wants in different ways, it should be easier to rewrite sos to produce a data.frame of class 'findFn' containing whatever information a user might want in a form that would allow different 'findFn' objects to be combined via union, intersection, etc. (I've used "union" a lot myself, because similar search terms often return different results. I've not found "intersection" so useful.)

(Spencer Graves)

(BMB) Yes. I was surprised at how much impact it made to have the RStudio folks open/provide an API to their download logs (leading to the cranlog package etc). It would be great if there were an API we could use to look for scholarly/peer-reviewed records (although I guess we could re-use Depsy's citation counts)

DavidMCEddy (discuss •

contribs) 18:52, 27 January 2018 (UTC)