Daniel Burrows wrote: > I'm also kind of curious about darcs internals, so I was wondering if > you could tell me where I'm wrong in the following reasoning. > > > AIUI, saying that "patch A depends on patch B" is equivalent to saying > that A touches (modifies or deletes) lines introduced by B. [0] If > that's the case, then if M is the number of lines affected by A, surely > N <= M, and so the asymptotic performance/size requirements are O(M)? > > Put more practically, it seems like you could avoid having to > recalculate patches by caching which patch is responsible for any given > line of the current pristine tree. That would inflate repository sizes > by a constant factor but would mean that darcs didn't have to > constantly recalculate patch relationships.
what about initially: line= A patch 1 changes "A" to "B" patch 2 changes "B" to "C" patch 3 changes "C" to "D" now patch 3 indirectly depends on patch 1; I don't know what that means to darcs. It gets more complicated when each patch also modifies other parts of files, thereby dragging in other patches as dependencies, and apparently conflicts/resolution make this nontrivial to deal with? I don't know, I don't think I understand darcs internals any better than you. ~Isaac _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
