"bzr mv considered harmful" or "IDE != command-line"

[James Westby]

On Fri, 2008-03-07 at 14:00 -0500, John Yates wrote:
> A bzr design goal is the ability to record information about file moves
> and renames.  This comes down to two tasks:
> 
> 1) providing a storage format that can record/track such information
> 
> 2) developing one or more UI models to acquire the information to be
> recorded

Hi John,

Thanks for your mail, it's good to get your views.

I am a fan of rename detection, and more. However I think that at commit
time is also a little limiting. However I see that with bzr's current
status, detection at commit time would be very useful for some tools
(and some command line users).

However, I think there are two issues with this.

The first is the heuristics that we use. There are obviously other 
implementations that we can learn from on this.

The second is performance, as it is a very expensive operation.
Obviously this can be enabled by a config option of a flag, but it
does still impose a cost on the people that want this feature. With
just renames it could perhaps be limited to cases where you have
adds and deletes in the tree, but if we have copies then any add
will trigger it.

However I don't think these problems are insurmountable, and there is
obviously a demand for it, so it should be considered.

Would you be able to offer any help in the implementation that could
speed this along? There are some things, for instance a python 
implementation of rename detection (perhaps ported from git) would
help to get the feature realised.

Thanks,

James