[Yum-devel] yum devel series and help needed

seth vidal skvidal at linux.duke.edu
Sun May 28 04:53:21 UTC 2006

On Wed, 2006-05-24 at 15:44 -0400, seth vidal wrote:
> Hi folks,
>  we should probably start pruning through some of the things we talked
> about before for this devel series soon.
> Here are the items I think we talked about in brief-form. Pick from them
> at will:
> Stuff we talked about doing for this series:
> - remove libxml2

Hey folks,
 looking through the code the only remaining place where libxml2 is used
is the mdcache.py metadata parser. It is also the item which takes the
xml metadata and creates the pickles. There is currently only one
application that uses this which is repoclosure. It does it b/c the
file/provides lookups are faster if everything is in memory. This
obviously works not-so-well when you have a massive amount of packages.

I've done some work on repoclosure to optimize it to work better with
the sqlite backend and it's improved a good deal, but it's still slow to
resolve and check all the deps for every package in every repository.

So I'm asking this: is everyone comfortable with the idea that
repoclosure might be a bit slower but we'll at least be able to reduce
our metadata parsing down to Tambet's C parser and the celementtree
parser in mdparser.py? I know I am comfortable with those two options
but I wanted to make sure no one had any deep longing for the libxml2
parser before I start chopping up code.


More information about the Yum-devel mailing list