Loggerhead directions

[John Arbash Meinel]

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

Gary van der Merwe wrote:
> On 15/04/2010 03:54, John Arbash Meinel wrote:
>> loggerhead is in a bit of a pickle, trying to stay stateless and yet
>> handle cache state... I don't have a great answer here.
> 
> At one stage (edged on by Martin Albisetti) I looked at porting qlog to
> loggerhead. I ran into 2 problems:
> * I could not find a nice way to render the graph in html.
> * Moving to stateless. qlog calculates lots of info at load, and caches
> that in memory. Each time you collapse, or expand a merge, the graph
> layout is recalculated, but because alot of the info needed is cached,
> this is very fast. How do we port this to a stateless environment?

Loggerhead currently does its own caching using an LRUCache structure.
So when you need state, *if* you are looking at the same location, it is
cheap.

One option would be to find out what info qlog needs, and preprocess the
graph, and shove it in a disk cache (like bzr-history-db).

I don't know what specific data you compute, though.

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

iEYEARECAAYFAkvHMD0ACgkQJdeBCYSNAAPRTgCfRV8K0ZqWLO/0Gex5Q5rom7nl
HJYAoI/QSWaEcrsw3Yc7E10t9vzPR4pP
=2Uqe
-----END PGP SIGNATURE-----