[Rd] MetaCran website v1.0.0-alpha

Uwe Ligges ligges at statistik.tu-dortmund.de
Mon May 25 10:59:48 CEST 2015



On 25.05.2015 02:29, Gábor Csárdi wrote:
> On Sun, May 24, 2015 at 7:40 PM, Uwe Ligges
> <ligges at statistik.tu-dortmund.de
> <mailto:ligges at statistik.tu-dortmund.de>> wrote:
>
>     Thanks for letting us know about the new website. Some comments:
>
>     - Download statistics: Where are they from? CRAN does not monitor
>     downloads generally, maybe some selected mirrors do.
>
>
> It's the RStudio mirror only. It is mentioned on the Services page and
> on the page of the download count API. It should probably be on the main
> page, but 'Most downloaded from the RStudio mirror' sounded very clumsy,
> even if it is correct..... Anyway, I'll put it there in some form in a
> minute.
>
> Btw. I do not think this statistics is the best:
> - single mirror, although this is the least of my concerns to be honest.
> - if a package is updated frequently, then it is downloaded more
> frequently, even if it is not used by more people.
> - if a popular package depends on another one, that is downloaded often,
> too. Even if it is not used much directly. You would probably want to
> rank down these packages  a bit in the search result list. A good
> example is my 'rversions' package, that is trending like crazy, but
> nobody really wants it, only the last version of 'devtools' depends on it.
>
> Ideally, I would want to count the number of 'install.packages()'
> invocations, and this is actually technically easy without tracking
> anyone. Is CRAN open to talk about this?
>
>     - Section "Recently updated" can only hold 9 packages, but
>     frequently more than 9 get accepted even within an hour, hence not
>     sure if this makes sense.
>
>
> Click on "Recently updated", and you get the latest 100. You can get
> even more with the API, if you want.
>
>     - The links "Download R" and "CRAN homepage" point to a particular
>     link of a mirror rather than the official pages. I'd appreciate if
>     you do not link to selected mirrors. The R homepage on purpose links
>     to http://cran.r-project.org/mirrors.html when pointing to CRAN.
>
>
> Good points, I'll fix these. I put in that mirror, because it was fast,
> and it is bound to be fast from anywhere in the world, whereas the main
> site was sluggish from my location at that time.


OK, thanks!

Uwe

> Thanks!
> Gabor
>
>     Best,
>     Uwe Ligges
>
>
>
>
>     On 24.05.2015 11 <tel:24.05.2015%2011>:44, Rainer M Krug wrote:
>
>         Gábor Csárdi <csardi.gabor at gmail.com
>         <mailto:csardi.gabor at gmail.com>> writes:
>
>             Dear All,
>
>             [ I was wondering if this should have gone to the new
>             mailing list. Maybe. ]
>
>             As some of you maybe know from my earlier posts, I am
>             building a simple
>             search engine for R packages. Now the search engine has a
>             proper web site,
>             where you can also browse CRAN packages.
>
>             http://www.r-pkg.org/
>
>             As I see the value is in
>             1. package search (search box on top right)
>             2. APIs, see http://www.r-pkg.org/services
>
>             It is in alpha version, meaning that things seem to work,
>             some pages are a
>             bit slow and there are a lot of glitches to fix.
>
>
>         I had a quick peek, and it looks really nice! I particularly
>         think the
>         github integration for diff-ing versions can be very use full!
>
>         It might be an idea, to also add R itself to the github repo for
>         diff-ing?
>
>         Thanks a lot,
>
>         Rainer
>
>
>             Please tell me what you think.
>
>             Best,
>             Gabor
>
>                      [[alternative HTML version deleted]]
>
>
>
>
>         ______________________________________________
>         R-devel at r-project.org <mailto:R-devel at r-project.org> mailing list
>         https://stat.ethz.ch/mailman/listinfo/r-devel
>
>



More information about the R-devel mailing list