Reordering main repo?

Adrian Buehlmann adrian at cadifra.com
Mon Jul 20 03:30:17 CDT 2009


On 19.07.2009 18:43, Adrian Buehlmann wrote:
> On 19.07.2009 17:28, Greg Ward wrote:
>> On Sat, Jul 18, 2009 at 2:15 PM, Adrian Buehlmann<adrian at cadifra.com> wrote:
>>> As an experiment, I then started doing a couple of incremental pulls
>>> into a new clone, trying to reorder the sequence of revisions in the
>>> repo a bit.
>> Interesting idea.  As I understand it, repository order can have two
>> practical effects: the size of the manifest and the readability of
>> graphical representations.  (Well, and of course the revnum<->node
>> mapping.)
>>
>> So... obviously your ad-hoc reordering made a prettier picture.  What
>> effect did it have on manifest size?
> 
> File sizes in kB
>                     original    reordered
> 00manifest.i             572          572
> 00manifest.d           5'003        5'079
> 
> Remember, I just rearranged some recent ~30 csets a bit (which were
> the only ones that caused an ugly graph part in entire history of
> mercurial).
> 
> And the mercurial repo is a rather small one anyway (
> 9155 changesets with 18268 changes to 1246 files).
> 

I have applied my reordering procedure to the TortoiseHg stable
repo [1].

File sizes in kB
                     original    reordered
00manifest.i              198          198
00manifest.d              446          603

(repo size: 3153 changesets with 4691 changes to 399 files)

(note that Steve has already exchanged the official repo [1] with my
reordered one, so [1] *is* now reordered)

The original TortoiseHg stable repo had a really horribly graph when
viewed with "hgtk log" (the TortoiseHg log viewer): there were more
than 26 vertical lines (I couldn't count beyond 26, as the rest was
not even visible in the viewer).

The reordered TortoiseHg stable repo now has <= 5 vertical lines in
the graph (at the cost of a 35% increase of the size of 00manifest.d to
603 kB).

[1] http://bitbucket.org/tortoisehg/stable


More information about the Mercurial-devel mailing list