> ^- IMO: WorkingTree shouldn't be calling an '_' method on BzrDir,
Wow, we do that a lot already, what's the problem ?
The public methods or attributes are for bzrlib users while the private
ones are for bzrlib internal needs, we never say this applies to inter
class calls.
> so we should probably be making it public.
IMHO, methods should be public if there is a need for them outside of
bzrlib and we guarantee some stability and backward compatibility.
If there is no need, the maintenance cost is lowered.
> Then again, why is *BzrDir* finding the available name, and not
> working tree itself?
Well, I just touched this method lightly, I didn't research the
rationale in the corresponding merge proposal.
>>>>> John Arbash Meinel <email address hidden> writes:
> On 9/10/2010 9:40 AM, Vincent Ladeuil wrote: conflicts_ missing_ parent( self):
>> + def test_resolve_
> The available_ backup_ name seems ok. I'm not a big fan of 'transport.has' you-leap in general, but we were already doing it.
> or the look-before-
> I don't really understand what this is doing here, though: conflicts_ missing_ parent( self):
> + def test_resolve_
> It seems unrelated. Also why remove: no_parent( self):
> - def test_resolve_
I've replaced it by test_resolve_ conflicts_ missing_ parent clarifying it
while I was there.
> === modified file 'bzrlib/ workingtree. py' workingtree. py 2010-08-20 19:07:17 +0000 workingtree. py 2010-09-10 14:39:52 +0000 backup. append( path[1] )
> --- bzrlib/
> +++ bzrlib/
> @@ -2078,9 +2078,10 @@
> files_to_
> def backup( file_to_ backup) : generate_ backup_ name(file_ to_backup) _available_ backup_ name(file_ to_backup)
> - backup_name =
> self.bzrdir.
> + backup_name =
> self.bzrdir.
> ^- IMO: WorkingTree shouldn't be calling an '_' method on BzrDir,
Wow, we do that a lot already, what's the problem ?
The public methods or attributes are for bzrlib users while the private
ones are for bzrlib internal needs, we never say this applies to inter
class calls.
> so we should probably be making it public.
IMHO, methods should be public if there is a need for them outside of
bzrlib and we guarantee some stability and backward compatibility.
If there is no need, the maintenance cost is lowered.
> Then again, why is *BzrDir* finding the available name, and not
> working tree itself?
Well, I just touched this method lightly, I didn't research the
rationale in the corresponding merge proposal.