Package comparisons and reviews

General
#1

Hi,

CRAN Task Views is great to find a list of packages to perform a certain task. However, once you have the list, how do you weed it down to the one package you will use? Comparing them all requires too much time for me. I'm looking for something like a Rotten Tomatoes for packages :grin: https://www.crantastic.org/ had this kind of goal, I think, but a lot of good packages have 0 reviews here.

I like https://www.rdocumentation.org/taskviews, or in general looking up packages on https://www.rdocumentation.org/, because of the "percentile of downloads" statistic. However, this is the percentile of total downloads, i.e., it includes dependency downloads, not only direct downloads. Thus, if package A has fallen out of grace in the community, but it has been around for a long time, or if it's a dependency of package B which is newer and more commonly used these days, it gets an unfair advantage over package B. Also, https://www.rdocumentation.org/ doesn't seem to have user reviews/comparisons among similar packages.

So, what's your process to choose among packages?

cross-posting disclaimer: similar question asked also in the #r4ds Slack community https://rfordatascience.slack.com/archives/C6VCZPGPR/p1524298419000029

#2

To complete the list of website with this kind of information you have also METACRAN
I use it to search :package: sometimes

There is also this to help you choose

unnamed-chunk-3-1.png1200×1408 111 KB

3 Likes
#3

METACRAN looks nice, but it seems very similar to https://www.rdocumentation.org/ to me: they both give information on the total number of downloads. No surprise Rcpp comes on top of downloaded packages - nearly all computationally efficient R packages depend on it.

https://github.com/ropenscilabs/packagemetrics looks great! I severely misunderstood this project - I read the introductory rOpenSci post last year, but I thought it only compared table-related packages - instead it collects statistics about all CRAN packages. The table packages are the only ones for which you find a human-compiled comparison of different packages, but the statistics are there for all packages. AND you can make compare one package against another (or multiple packages at the same time) - that's excellent. I'm surprised it's not more famous!

#4

Does it also compare not-on cran package ? I am not sure...

#5

I don't know, but even if it compares CRAN-only packages, it's already a valuable tool IMO.

1 Like