On 21.09.20 08:37, Andreas Tille wrote: > Hi, > > On Sun, Sep 20, 2020 at 04:02:39AM +0200, Steffen Möller wrote: >> Hello, the git archive occupies 421M on my disk, the tarball is 68MB, >> the package only 29MB, which I think is worth it. We have the human >> already in the archive, I would like the others to follow - which means >> rat, fly and worm to me. > I admit I'm not yet conviced what this will bring for the user. Usually > we are packaging CRAN / BioC packages as *dependencies* of other software > so we can install a whole system that contains the code users want to > run without extra effort.
Please have a look at the "Suggests me" and "Imports me" at https://bioconductor.org/packages/release/data/annotation/html/org.Mm.eg.db.html and preferably also at the even more impressive list at https://bioconductor.org/packages/release/data/annotation/html/org.Hs.eg.db.html And since reverse translational genetics/medicine (https://www.nature.com/articles/454274a) has become a thing, we need these packages for all species - and 68MB plus 29MB (platform independent) is not that bad. Just that increasingly large git repository is annoying. > I also try to package Test-Depends - except if it these are larger data > sets. In this case I tend to skip the test. We could even access the > internet for autopkgtests and may be we should enhance our tests > regarding this. But I'm somehow reluctant with these large data packages > before we have discusses its advantage for user installations. These are not large data sets. It is just no overly comfy with salsa, yet. And we shall develop ideas to fix that. git-lfs helps to reduce bandwidth and local disk space, a first idea, maybe we develop some more. A bit annoying is that these BioConductor-provided data packages kind of circumvent our discussion about how to deal with (much larger) databases with Debian and their post-download processing. Steffen
