[csw-devel] Atom feed with releases to unstable

[Maciej Bliziński]

2011/3/19 Ben Walton <bwalton at opencsw.org>:
> Excerpts from Maciej Bliziński's message of Sat Mar 19 07:12:22 -0400 2011:
>
>> Implementing the atom feed would require setting up a new feed
>> served by http, and a glue code between the catalog notifier and the
>> feed data backend (e.g. mysql).
>
> It seems to me that this could be done by the notifier itself.  The
> catalog parser and comparator parts would remain the same.  The data
> that generates the email notification from the template would just
> need a second step that appends to the xml atom feed...or maybe calls
> out to a separate tool that does this.
>
> At that point, it's just a matter of making the atom feed visible.
>
> I think that right around here[1] a bit of code would be added that
> uses the catalog comparator code (already self contained) and spits
> makes the appropriate changes to the feed.  It might make sense to
> factor out the calls to catalog comparator from the
> NotificationFormatter too, but I didn't look too deeply at that.
>
> Maciej, is this roughly accurate?

Yes, that's it.

Catalog downloading and comparing results are best reused.  The
comparator class only compares a single catalog, specific to an
architecture and OS release.  During one run, the notifier downloads
and compares 6 (2 architectures × 3 OS releases) catalogs and stores
all diffs in a single data structure.  This data structure could be
reused.  You could for instance memoize the _GetPkgsByMaintainer
function and call it again, reusing the data.

You could add a new method, similar to FormatNotifications, calling
memoized _GetPkgsByMaintainer and doing something else with the data.
I've packaged an atom formatting library for Python: py_atomixlib.

To persist data between runs, the same simple method could be used:
pickle the data on disk, and read them during the next.  Something like:

unpickle old feed data
get new information by comparing catalogs
add new entries to the feed
remove old items from the feed
format feed
save feed
save data

Maciej