[gentoo-soc] G-CRAN weekly report #2

[Auke Booij]

Last week I've been developing the repository driver for G-CRAN,
focusing primarily on syncing CRAN-style repositories. Downloading a
list of packages is easily done using python's urllib, which means
that we'll also be able to sync from non-HTTP CRAN-style package
mirrors. That file listing packages uses the Debian Control File
syntax, which apparently is much like RFC822 syntax. Googling around
for a good solution to read these files, I considered both
python-debian, which is in the pypi package repository, and a python
library called rfc822.

Interestingly, if G-CRAN would use python-debian, it would indirectly
depend on G-PYPI, which is of course slightly dubious. Nonetheless,
I've taken a look at its code and decided it's not suited to my
purposes, since it pretty much assumes a valid format, which is not a
very fair assumption in the case of CRAN (remember, this is what held
the Paludis devs from supporting it any further).

The other option was rfc822.py, which initially appeared to be too
much geared towards emails, but after some tests it seems to be a good
candidate after all, and I've decided to go on with this.

What I've also been working on is a piece of code somewhat like
ebuild.sh for python... a way to deal with the PMS side of packages.
I'm keeping it very minimalistic until I need any serious features and
will use a mapping function to transform CRAN-style packages into
PMS-style packages.

During the next few days I'll be finishing the CRAN repository reading
code, and if it's finished before the next report, I'll try to build
some initial functionality to export it to g-common. It's slowly
getting time to start uploading code to some repository and this is
another item on my shortterm todo list. g-common is still on hold
since the three of us are still working on our individual drivers.