Don't have the base tests inherit from TestCase else they will be discovered by unittest and run even though they are not fully defined.

See http://bugs.python.org/issue16748 as the trigger for this issue.

Here's a patch:
1) the base class doesn't inherit from TestCase anymore -- the subclasses do;
2) added a skipUnless() decorator on the C subclass;
3) used the modern if __name__ == '__main__': unittest.main() idiom;
4) renamed the AcceleratedExampleTest to CExampleTest (I prefer the PyFoo/CFoo parallel, "Accelerated" doesn't necessarily imply that it's testing the C version);
5) added a paragraph to explain the idiom;
6) added Post-History;
7) did a couple of minor cleanup in the code;

Hey Ezio, you forgot to attach the patch

That would explain why no one was reviewing it :)

A few minor grammatical and style nits in the prose added at the end of the diff:

> +The test module defines a base class with the tests methods that access the
> +``heapq`` module through a ``self.heapq`` class attribute, and two subclasses
> +that set this attribute to either the Python and C versions of the module.

I think that should be "Python or C version" (and -> or; versions -> version).

> +Note that only the two subclasses inherit from ``unittest.TestCase`` -- this
> +prevents the ``ExampleTest`` class to be detected as a ``TestCase`` subclass

"to be" should be "from being", I believe.

> +by ``unittest`` test discovery.
> +A ``skipUnless`` decorator can be added to the class that tests the C code

This should either have another newline inbetween or be reflowed.  Either one paragraph or two makes sense to me, but I can't tell which way you actually meant it to go which makes for confusing reading of the ReST source.

As far as the example code changes, those look good to me :)

Thanks for the review, new patch attached.

> This should either have another newline inbetween or be reflowed. 
> Either one paragraph or two makes sense to me, but I can't tell which 
> way you actually meant

I meant one and half :)
I guess in HTML that would have been a <br> inside the <p>, but I don't think it really matters if it's rendered as a single paragraph.

> Thanks for the review, new patch attached.

You're quite welcome.  Is there anything I've missed in the process of reviewing itself?  This is the first time I've reviewed a patch here...

I did miss another nit in the prose, though; "the tests methods" in the first line isn't quite right, but I can't decide if it should be "the test's methods" (singular possessive), "the tests' methods" (plural possessive), or "test methods" (non-specific, non-possessive).  Any of the three that better gets your point across makes me happy :)

>> This should either have another newline inbetween or be reflowed. 
>> Either one paragraph or two makes sense to me, but I can't tell which 
>> way you actually meant
>
> I meant one and half :)
> I guess in HTML that would have been a <br> inside the <p>, but I don't 
> think it really matters if it's rendered as a single paragraph.

I see.  As is works for me if it works for you, though I might lean towards rounding it up to 2 paragraphs :)

Attached a new patch.

Looks good to me :)

LGTM as well. Feel free to commit it, Ezio, or assign to me and I will commit it later (probably this weekend).

Done in http://hg.python.org/peps/rev/3740f42d3b94, thanks for the reviews!