API export of IDistroSeries.getBuildRecords() can't filter on active source publications

William, what are you doing with your application here? I wonder if we can be more helpful than simply exposing IDs.

I was intending to use it to more efficiently generate a full package vs. arch grid of FTBFSes for a distroseries. getBuildSummariesForSourceIds should allow me to do that without making 7 requests for each package.

On Monday 08 June 2009 23:54:07 William Grant wrote:
> I was intending to use it to more efficiently generate a full package
> vs. arch grid of FTBFSes for a distroseries.
> getBuildSummariesForSourceIds should allow me to do that without making
> 7 requests for each package.

How are you getting hold of the SPPH objects? ie. how are you filtering which
ones you get?

I find all failed, current builds, and get the SPPHs from them.

For reference:
http://bazaar.launchpad.net/~geser/%2Bjunk/qa-ftbfs/annotate/head%3A/source/build_status.py is the current source and http://qa.ubuntuwire.com/ftbfs/ the result of it.

Perhaps you have some ideas how to improve it or the LP API.

William

Celso says he showed you the launchpadlib code snippet that uses a different way to get the failed builds. He and I both think that this is more efficient that the current way you're using, but is still not optimal because you end up throwing away some builds that were downloaded since they no longer have published sources.

So, I will change the description on this bug to reflect that series.getBuildRecords() needs to also filter on active builds (i.e. published ones).

something as simple as this is not usable:

        for pkg in ppa.getBuildRecords():
            csp = pkg.current_source_publication
            if not csp:
                continue
            if pkg.buildstate == 'Successfully built':
                continue
            print " %-28s %-8s %-5s %-20s %s" % (csp.source_package_name, csp.distro_series.name, pkg.arch_tag, pkg.buildstate, pkg.build_log_url)

in my case, it takes ~10 minutes, does ~300 HTTP requests, and download more than 10MB, just to display between 0 and 12 lines.

I'm going to see how far I can get with this one...