Skip to content

FYI: daily R source tarballs from ETH: *.xz instead of *.bz2)

4 messages · Martin Maechler, Brian Ripley, Hervé Pagès

#
A quick notice for anyone who uses cron-like scripts to get
R source tarballs from the ETH  R/daily/ s:

I've finally switched to replace *.bz2 by *.xz which does save
quite a bit of bandwidth.

Currently, you can see the 2 day old *.bz2 (and their sizes) and
compare with the new  *.xz one  (sorted newest first):

  https://stat.ethz.ch/R/daily/?C=M;O=D


Best,
Martin
#
Hi Martin,

Sounds good. Are there any plans to support the xz compression for 
package source tarballs?

Thanks,

H.
On 9/8/23 06:44, Martin Maechler wrote:

  
    
3 days later
#
On 09/09/2023 01:56, Herv? Pag?s wrote:
What makes you think it is not supported?

R CMD INSTALL happily installs .tar.xz files, and the name is not used 
to detect compression so .tar.gz files could be bzip2- or xz-compressed.

Note that tarball compression is pretty much irrelevant where the 
tarball contains large compressed files, for example .rda files or 
vendor.tar.xz files of Rust sources.  You have to arrange that the first 
compression is the bast possible.
#
On 9/11/23 22:39, Prof Brian Ripley wrote:

            
I guess because I've never seen source tarballs distributed as .xz files 
but it's good to know that 'R CMD build' and 'R CMD INSTALL' support that.

So let me reformulate my question: do CRAN have any plans to switch from 
.tar.gz to .xz for the distribution of source tarballs? Is this 
something that tools like write_PACKAGES(), available.packages(), and 
install.packages() would be able to handle? Would they be able to handle 
a mix of .tar.gz and .xz packages? (Which would be important for a 
smooth transition from .tar.gz to .xz across CRAN/Bioconductor.)

I'm just trying to get a sense if the effort to reduce bandwidth will go 
beyond the distribution of R source snapshots.

Thanks,

H.