Compressing weaved revisions?

[John A Meinel]

Martin Pool wrote:
> On 01/10/05, John A Meinel <john at arbash-meinel.com> wrote:
>
> Bear in mind that some filesystems (ext2/3, reiserfs, ntfs, probably
> some others) will store small files in less than a block, so the
> actual disk usage is not necessarily as bad as du indicates.  I think
> du can't detect fractional block usage.
>
> If we just want to store things in a single file to avoid using
> inodes/blocks and multiple http requests then weave seems like
> overkill.
>
>
>
>>using a zipfile is another reasonable possibility, as the file is
>>indexed, giving us a decent time to get to the entries. You can append
>>to it (though if you crash in the middle, most readers will consider it
>>completely corrupted).
>
>
> I think this is a pretty interesting idea, particularly as Python has
> a builtin zipfile class.  I don't know if zip is efficient at random
> access to lots of tiny files; it might not be.
>
> Another approach is to use a local database like tdb or sqlite.
>

Well, if you are interested in a local database, feel free to check out:

http://bzr.arbash-meinel.com/plugins/revstore2sql/

It is my current work to create a local database for revision-store and
inventory-store.

Right now, I am able to compress the entire work of bzr.newformat into a
5.7MB sqlite file. That is inventory as well as revisons. (Revisions end
up really tiny <200k.)

This isn't complete, as in you can't replace a current revision +
inventory with it yet. I'm just playing with storage and the schema
right now. (Especially since the schema might change with time while we
settle out .revision versus .text_version, etc.)

John
=:->

> --
> Martin
>
>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 253 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20051003/56c6cdea/attachment.pgp