RFC: change revision serialiser
John Arbash Meinel
john at arbash-meinel.com
Thu Aug 14 05:21:45 BST 2008
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Robert Collins wrote:
> On Wed, 2008-08-13 at 22:48 -0500, John Arbash Meinel wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Robert Collins wrote:
>>> So, thinking about serialisers, it seems trivial to change the revision
>>> serialiser to something with less overhead than xml; use use utf8, so
>>> surely there is a simple fast utf8 format we can use (perhaps RIO?), we
>>> could put that into a development format pretty easily...
>>>
>>> -Rob
>> Back in the day... rio isn't faster than xml. I think it was faster at some
>> bits (writing) but not faster at others (reading, IIRC).
>>
>> We really have a quite tweaked system for xml right now, such that at least
>> with pure-python it is hard to beat.
> ...
>> It certainly isn't hard to drop something into a new format. I would
>> certainly
>> also recommend one that splits on newline more. At least, it would
>> help our
>> line-based delta compression. (certainly if your inventory suggestion
>> is
>> getting rid of that...)
>
> Well, we don't compress revision texts; so I'm not sure whether that
> would help or not. I mainly thought of this because of your comment that
> log FILE was slow being due to xml overhead.
>
> -Rob
>
I didn't lsprof it to find the root cause. I was just intrigued that 'bzr log
file' was actually faster than 'bzr log' for once, and that was the first
thing that came to mind.
John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iD8DBQFIo7LZJdeBCYSNAAMRAkKTAJ9XJc2GZrVm4PJFoH9bP3XsJu9gBQCaAtdH
aVTs3WDQ4xvjD8lGZSOSzr0=
=xzPl
-----END PGP SIGNATURE-----
More information about the bazaar
mailing list