[Grok-dev] Grok 1.1 the plan

[Martijn Faassen]

Hey Paul,

Thanks for the progress report!

Paul Wilson wrote:
> On 1 February 2010 17:23, Martijn Faassen <faassen at startifact.com> wrote:
>> * Martian testing changes (the Manuel work)
> 
> This is almost ready. I had a number of problems relating to the fact
> that Martian was searching '__builtin__' for implicit directives
> (since 'object' is often in the MRO), and I was using it to store
> doctest-defined objects in manuelpi.fakemodule. This has forced me,
> rightly or wrongly, to guard Martian from scanning __builtin__ at all.

I think it's correct not to worry about __builtin__ (or object, for that 
matter, I think).

> I don't think this should cause any issues in general. This is sorted
> now  -- all that remains is to double check the doctests for
> formatting. I'll notify the list when its ready for review.

Thanks very much on working on this. Our tests should become a lot more 
comprehensible thanks to this.

> I guess the next job would be to update the grokcore.* suite where
> needed. Any volunteers? :)

I think the best place to start would be grokcore.component.

Currently the tests have this structure:

"""
The doctest
"""

The module under test.


We also have cases where an external fixture is in use and referred to 
from the doctest.

These can typically change to the fakemodule work that Paul has done, 
where the module is simply inline in the doctest. Martian already has 
some tests like that, but using an earlier hackish approach. I'll leave 
it up to Paul to write a more complete tutorial for those who want to 
help. :)

Regards,

Martijn