relval progress report

[Adam Williamson]

On Fri, 2014-10-03 at 09:00 -0400, Kamil Paral wrote:

> I wonder, would it make sense to put it to our "misc" repo, rather
> than having the projects scattered around the web? I think it would be
> a good fit.
> https://git.fedorahosted.org/cgit/fedora-qa.git/
> 
> It definitely makes the bookmarking process easier ;)

It's a lot faster for me on happyassassin, though :P

It might make sense, I'll look into it in a bit.

> testcase_stats is Josef's baby, but I think there wasn't really an
> intention to revive it for F21 due to so many Fedora.Next and wiki
> changes. And there was the possibility to use moztrap for F22. But if
> you think it's viable and helpful to revive testcase_stats, we can
> have a look at it.

> I used direct MW API in my stats-{wiki,bugzilla,bodhi} scripts and
> Josef got inspired by it in testcase_stats. I didn't know about higher
> level tools like mwclient, that's very helpful information. If we work
> on it further, I'm all for simplifying it by using a higher-level API.
> 
> Which reminds me, we totally failed to collect and send Alpha
> contributor stats. And the wiki script will probably need adjustments.

wikitcms (relval) now has a web page, releases, and a package
repository:

https://www.happyassassin.net/wikitcms/

Today I've been doing some major surgery in a feature branch (this stuff
isn't in the packaged version yet, you'll have to check out the branch
and run it from the checkout):

https://www.happyassassin.net/cgit/wikitcms/log/?h=wiki-stats

relval can now act as - I believe - a full replacement for stats-wiki,
plus Extra Capabilities, and it works for both 21 *and* earlier
releases. Try:

relval userstats --release 20 --milestone Final

or:

relval userstats --release 21

or:

relval userstats --release 21 --filter TC --filter RC

wikitcms now has the concept of a 'result object', wiki.find_results()
returns a list of the result objects found in a given chunk of wikitext,
and 'userstats' is a new relval sub-command that does the work of
stats-wiki using these features of wikitcms. There's some fairly ninja
list/set comprehension stuff going on there that I TOTALLY DIDN'T RIP
FROM STACKOVERFLOW AT ALL OH NO. (Anyone who can manage to do bugs[user]
as a single nested comprehension gets many bonus points, because I'm
damned if I can). Use of set comprehensions makes the code require
python 2.7, but all supported Fedoras have that version, so hey. I've
replaced most of kparal's implementation by this point, but the function
for producing the HTML-formatted output is mostly untouched (I just
changed its inputs a bit), so thanks for that.

I'm reasonably happy with the implementation at this point. So far as
I've tested, it replicates the results of stats-wiki.py when called with
the same constraints, while offering the additional capability of
querying by milestone (and the underlying functions have even more
flexibility; you can specify the result statuses to include, and whether
'previous (foo) run'-type results should be included).

I believe it should be relatively simple to re-implement testcase_stats
on top of this stuff (I'll look at that tomorrow). I also have a sort-of
plan for doing more sophisticated result parsing and adding
'environment' attributes to the result objects; I think it ought to be
possible to identify and correctly parse both '2D' and '1D' result
tables.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net