Mark Grosberg <mark@nolab.conman.org> writes:
> > it strikes me that you're complicating things enormously, *and* you
> > introduced a quadratic algorithm for tagging the array.
>
> Why?
>
> As for the algorithm on the array, I thought about using a hash. But I
> figured that would waste memory. Usually the list of comitted files isn't
> that large that it would matter.
>
> But I'll do it and submit another patch and see how it works. Should I
> make a subpool for the hash or just use the pool passed in to the log
> getter?
The great thing about the hash is that you can alloc just the hash
itself, but not bother duplicating the stuff in the array into the
hash's pool. That is, there should be no lifetime concerns. So I
don't think you'll be, say, doubling the amount of allocation for the
commit items or anything.
> > By the way, your method for deleting stiff from the array seems wrong;
> > you never check the flag on the items you move from the end of the array
>
> I'll change it to use qsort.
Actually, I was a bit confused as to why the deletion step occurred at
all. Can the commit driver not simply skip over those things as it is
looping over the list?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri May 9 02:31:11 2003