Merge lp://qastaging/~jimbaker/pyjuju/generate-html into lp://qastaging/pyjuju/ftests

[1]

This tool is doing three different things:

1. Running bzr update
2. Running churn
3. Converting churn output to html

The tool we discussed should do 3 only.

Let it take as input a directory full of churn output directories, and
produce the output based on the directory name. So we run it like:

$ butler.py watefall/ > waterfall.html

Given that within the waterfall dir one might expect:

    /waterfall
        /341
            some-test-name.out
            other-test-name.out.FAILED
        /342
            some-test-name.out
            other-test-name.out

etc. butler will then have the 341, 342, etc in the rows, and the test
names in the columns, with OK/FAILED in the cells, as you've done.

Once this is in place, let's create a tiny but opinionated driver
that knows how to update a branch *revision by revision* and drive
churn and butler from outside. This driver must get as input _just_
the build directory, and will have three things inside it.

So

  $ whatever build/

Generates

  build/
    ensemble/
    waterfall/
    waterfall.html

But once you run

  $ whatever build/

again, it will notice that there's already content inside it,
and will just update revision by revision the branch between
the latest revision in the waterfall and the current known
tip of the trunk.

With that in place, we can put this logic in a machine, and have it
rsyncing this content out into a public web server.

Do not use bzrlib, btw. Use bzr itself. We're not doing anything
fancy, and the command line API won't change.

[2]

Please put a copy of pystache within the tree.

[3]

96 + <td>{{#failed}}<a href="{{path}}"><strong>FAILED</strong></a>{{/failed}}
97 + {{^failed}}Succeeded{{/failed}}

We have the succeeded output as well. Let's offer it so people can
investigate differences.

It might also be better to have "ok" (lowercased) rather than "Succeeded", to make the
"FAILED" one much more outstanding.

On Fri, Sep 2, 2011 at 6:42 PM, Gustavo Niemeyer <email address hidden>wrote:

> Review: Needs Fixing
> [1]
>
> This tool is doing three different things:
>
> 1. Running bzr update
> 2. Running churn
> 3. Converting churn output to html
>
> The tool we discussed should do 3 only.
>
> Let it take as input a directory full of churn output directories, and
> produce the output based on the directory name. So we run it like:
>
> $ butler.py watefall/ > waterfall.html
>
> Given that within the waterfall dir one might expect:
>
> /waterfall
> /341
> some-test-name.out
> other-test-name.out.FAILED
> /342
> some-test-name.out
> other-test-name.out
>
> etc. butler will then have the 341, 342, etc in the rows, and the test
> names in the columns, with OK/FAILED in the cells, as you've done.
>

Done. I believe the mechanism I have used in working with mustache templates
to generate this table is the best approach.

>
> Once this is in place, let's create a tiny but opinionated driver
> that knows how to update a branch *revision by revision* and drive
> churn and butler from outside. This driver must get as input _just_
> the build directory, and will have three things inside it.
>
> So
>
> $ whatever build/
>
> Generates
>
> build/
> ensemble/
> waterfall/
> waterfall.html
>
> But once you run
>
> $ whatever build/
>
> again, it will notice that there's already content inside it,
> and will just update revision by revision the branch between
> the latest revision in the waterfall and the current known
> tip of the trunk.
>
> With that in place, we can put this logic in a machine, and have it
> rsyncing this content out into a public web server.
>
> Do not use bzrlib, btw. Use bzr itself. We're not doing anything
> fancy, and the command line API won't change.
>

Sounds good. Given that butler is supposed to be managing the results of
everything, I renamed butler.py to waterfall.py, leaving the name butler
available for this opinionated driver. That will be done in a separate
branch.

>
> [2]
>
> Please put a copy of pystache within the tree.
>

Done. This actually involved some debugging effort because pystache, in its
interaction with MarkupSafe, was somehow behaving differently when placed
here, vs normal package install with tools like pip. Modified accordingly.
So pystache as added here just has package code + (renamed) license file.

> [3]
>
> 96 + <td>{{#failed}}<a
> href="{{path}}"><strong>FAILED</strong></a>{{/failed}}
> 97 + {{^failed}}Succeeded{{/failed}}
>
> We have the succeeded output as well. Let's offer it so people can
> investigate differences.
>

Done

>
> It might also be better to have "ok" (lowercased) rather than "Succeeded",
> to make the
> "FAILED" one much more outstanding.
>

Done. The mustache template will need some more work re styling, but it's
functional now. This can be done in a future branch.

Looks great, thanks Jim. Only a couple of minor comments, and it can go in IMO.

Then, for the driver! I hope we can see a waterfall working this week still!

[4]

=== added file 'pystache-LICENSE'

Please rename this to LICENSE and put it within the pystache dir itself.

[5]

+OUTPUT_STATUS_RE = re.compile(r"(?P<name>.*?)(?P<status>\.out.FAILED|\.out)")

This regex makes things more complex than it helps. E.g.:

FAILED_SUFFIX = ".out.FAILED"
OK_SUFFIX = ".out"
(...)
    if name.endswith(FAILED_SUFFIX):
        test_name = name[:-len(FAILED_SUFFIX)]

etc.

On Thu, Sep 8, 2011 at 6:10 PM, Gustavo Niemeyer <email address hidden>wrote:

> Review: Approve
>
> Looks great, thanks Jim. Only a couple of minor comments, and it can go in
> IMO.
>
> Then, for the driver! I hope we can see a waterfall working this week
> still!
>

Please see the corresponding butler branch in review

>
> [4]
>
> === added file 'pystache-LICENSE'
>
> Please rename this to LICENSE and put it within the pystache dir itself.
>

Renamed accordingly

>
> [5]
>
> +OUTPUT_STATUS_RE =
> re.compile(r"(?P<name>.*?)(?P<status>\.out.FAILED|\.out)")
>
> This regex makes things more complex than it helps. E.g.:
>
> FAILED_SUFFIX = ".out.FAILED"
> OK_SUFFIX = ".out"
> (...)
> if name.endswith(FAILED_SUFFIX):
> test_name = name[:-len(FAILED_SUFFIX)]
>
> etc.
>
>
Done

Thanks Jim, please merge it on ftests.