[Rpm-metadata] createrepo testing please.
Daniel Veillard
veillard at redhat.com
Wed Jan 14 09:11:04 UTC 2004
On Tue, Jan 13, 2004 at 10:28:56PM -0500, seth vidal wrote:
> On Tue, 2004-01-13 at 17:42, Daniel Veillard wrote:
> > On Tue, Jan 13, 2004 at 04:55:35PM -0500, seth vidal wrote:
> > > tell me anything that breaks.
> >
> > not really "breaking" but there are some usability improvement
> > which may or may not be possible:
> > - memory usage (150 MB indexing FC1 i386/os)
>
> fixed - by doing dumping it to a string and writing each package node
> out then node.unlinkNode() node.freeNode() I've gotten rid of all the
> built up memory. it now maxes out at 7MB instead of whatever the size of
> the xml doc is.
Cool :-)
> > - reusing the existing XML as a cache on subsequent runs
>
> this will take more time. But I'd like to do it.
Suggestions:
- scan initially the primary.xml.gz to extract just the pathnames
- do the file lookup (xmlReader interface sounds just fine).
- build a list of added packages and removed packages
- scan the added packages, dump the descriptions for those
- then rerun the reader on primary.xml.gz, expand the nodes that
are kept unmodified and use the same dump routine you use now
to output them (see http://xmlsoft.org/xmlreader.html#Mixing
for an example of the reader and expansion code).
I think that way the impact will be minimal, both in terms of speed,
memory usage and code changes for reindexing.
Daniel
--
Daniel Veillard | Red Hat Network https://rhn.redhat.com/
veillard at redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
More information about the Rpm-metadata
mailing list