[MERGE] GraphIndex bisection

Andrew Bennetts andrew at canonical.com
Tue Oct 9 03:57:29 BST 2007


Robert Collins wrote:
> 
> On Mon, 2007-10-08 at 16:42 +1000, Andrew Bennetts wrote:
> > bb:comment
> > 
> > This isn't a proper review, just a comment on something that jumped
> > out at me
> > while skimming:
> ...
> > The -1/False/+1/result interface seems a bit distasteful to me.  Partly this is
[...]
> 
> I agree that its a little untasty. I felt it was more tasty that a
> tuple: (-1, None) (1, None), (False, None), (True, result).
> 
> The docstring is already quite clear about what result can be IMO, I'm
> not sure what can be made clearer - clearly it cannot be -1, 1 or False,
> but any other value is return to the caller e.g. ('key', None) if you
> were to return ((loc, 'key'), None) from the content lookup function.

Fair enough.  If I could think of a clearly better alternative I'd protest
harder, but seeing as I can't, I won't :)

-Andrew.




More information about the bazaar mailing list