[MERGE] Make knits the default format

[Martin Pool]

On 20 Apr 2006, Robert Collins <robertc at robertcollins.net> wrote:

> > Is it safe to do __class__ on any object?  It seems to work even for
> > None or other primitives.
> 
> AFAIK yes, and even if it fails it will tell one something.

Yes, looks reasonable - Python None is really much less like null than
it might appear.

> > > === modified file 'a/bzrlib/tests/revisionstore_implementations/test_all.py'
> > > --- a/bzrlib/tests/revisionstore_implementations/test_all.py	
> > > +++ b/bzrlib/tests/revisionstore_implementations/test_all.py	
> > > @@ -107,7 +107,9 @@
> > >          # we get a revision count and a numeric size figure from total_size().
> > >          count, bytes = self.store.total_size(self.transaction)
> > >          self.assertEqual(0, count)
> > > -        self.assertEqual(0, bytes)
> > > +        # some stores use disk immediately that they are created so we just 
> > > +        # check that its an int.
> > > +        self.assertIsInstance(bytes, (int, long))
> > >          self.add_sample_rev()
> > >          count, bytes = self.store.total_size(self.transaction)
> > >          self.assertEqual(1, count)
> > 
> > This test is a bit deceptive - it looks like it checks they are an int
> > and a long, but it actually just checks its a tuple.  Being able to
> > check that something matches such a template might be nice, but
> > assertIsInstance isn't that method.  For clarity, either check them
> > separately or just check it's a tuple.
> 
> No, it checks its an int or a long. Check pydoc isinstance :).

Ah of course, I'd forgotten that this acts just like isinstance()

> I'll add a couple of obvious cases, but I'm not really focused on
> documenting the existing api with tests :).

No need to add it now - I just wondered if they would break because of
the thing[-1] reference but they seem fine because of the way str.split
works.

-- 
Martin