export command is slower than checkout

[QbProg]

2 improvements for the export command:
I actually tryed it with "dir" output , not with compressed files.

1 - Progress status , it would be usefoul for large repositories.
2 - Make it faster . Even checkout is faster , from my POV should be at least as fast as checkout. I'm using it via local network.

I'm using Windows version 1.2.

Thank you

[Robert Collins (lifeless)]

So putting status aside for a second, I have some thoughts on speed. Basically the problem is we're doing size(tree) lookups. I propose that we walk the tree gathering file versions we want to export (for exporting a working tree this suggests using CURRENT to me). iter_changes is probably the right interface to use. Then we can emit all the directories to the exporter at once; after that we can walk the files to export via tree.iter_files_bytes().

[Martin Pool (mbp)]

Changed summary to make it specific to speed and falsifiable. I'm not sure if it's actually still slow or not.

[John A Meinel (jameinel)]

Is this now a dupe of bug #608640 (or vice versa?)

[Martin Pool (mbp)]

I think they're related but not necessarily the same, if it turns out
we do inefficient rpcs and also have some inefficiency locally.

[John A Meinel (jameinel)]

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

Martin Pool wrote:
> I think they're related but not necessarily the same, if it turns out
> we do inefficient rpcs and also have some inefficiency locally.
>

I'm wondering if it wasn't the fact that building the local tree wasn't
batching the requests (instead exporting one-by-one). Which has now been
'fixed' though we still use inefficient rpcs. Which would make this a
dupe of a different bug.

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

iEYEARECAAYFAkxrJNwACgkQJdeBCYSNAAP5mwCfS85AlXLpCLr4iHcgoI6Vt3lq
xp4Ani8oHtnIbJ9/6ZFC5p5aRd3+4ujT
=nY49
-----END PGP SIGNATURE-----