Storage requirements of tree transforms

[Denys Duchier]

Aaron Bentley <aaron.bentley at utoronto.ca> writes:

> I had originally planned that the iterators for new file contents would
> be held in memory until halfway through the transform application, and
> then applied.

I don't think that was very reasonable: content may be extremely large.

> The new plan
> ============
> The transform application uses a temp directory used to hold files
> undergoing moves/renames, called ".bzr/limbo".  I propose to stick the
> new contents there.

mmm... limbo! yummm... :-)

> Does this sound like a reasonable trade off?  Does anyone actually care
> about a small size increase during transform creation/application?

I think that is the only viable alternative.  It may not be possible for one
process to simultaneously hold all the new content.  For example, suppose you
have a branch holding a large number of big corpora and you want to apply a
changeset that adds a very small header to each one. boom!

Cheers,

--Denys