Code review comment for lp:~ian-clatworthy/bzr/faster-tags

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ian Clatworthy wrote:
> John A Meinel wrote:
>> This only benefits the case of "no actual tags present". I thought we
>> actually had something like this, but whatever.
>>
> Not true. OOo has tags (10-20) and its works fine.
>> I'll mention that "bzr tags --show-ids" should always be the same speed
>> (fast), and this is just avoiding loading the revision_id=>revno map
>> until actually needed.
> Right. But showing revnos is default behaviour so saying to users "but
> tags --show-ids is fast" doesn't seem a good enough response.
>> Unless revision_id_to_revno() also knows to try to determine mainline
>> revnos first (which it can do by walking backwards from tip). So there
>> is the chance that if all your tags are on the mainline (and recent) it
>> won't load as much history.
>>
> It does exactly that.
>
> Tagging things along the mainline is pretty common, particularly when
> there's only a mainline. :-) So I think this is a win for many projects
> using tags.
>
> Ian C.

Perhaps. Though having 'really old' tags in a long mainline is likely to
also be common.

If 'list(iter_reverse_revision_history(branch.last_revision()))' is much
faster than 'tsort.merge_sort' on very long linear histories, then I
think we have something wrong going on that is worth looking into.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkofnCAACgkQJdeBCYSNAAMGaACeMvKXKZVYzYbD9j22WOF9oHYg
64gAn34WHY6fejM21ipjQ88RlIftZR3G
=UGUR
-----END PGP SIGNATURE-----