[darcs-users] [patch272] Limit index updates to relevant subtree in a few cases.

Eric Kow kowey at darcs.net
Tue Jun 8 09:54:56 UTC 2010


On Mon, Jun 07, 2010 at 20:01:43 +0000, Petr Ročkai wrote:
> This has been sitting around on my mainline branch for a while. Guess I just
> forgot to send it... Should be self-explanatory (I hope).

Seems to make sense.  I was going to ask "what if the
user does not pass in any paths on the command line?", but I think
this tells restrictSubpaths helper tells me the answer,

      restrictPaths :: FilterTree t m => t m -> t m
      restrictPaths = if null subpaths then id else filter (filterPaths anchored)

By the way, I seem to remember having seen and reported a recent bug
where 'whatsnew .' would break but 'whatsnew' without any arguments
would be fine.  Do you know what I'm referring to?  I can't find it on
the tracker or in my local list!

The comments below are just Eric-reads-the-patch, no action on your part
needed.

> Sun May  9 12:22:48 CEST 2010  Petr Rockai <me at mornfall.net>
>   * Limit index updates to relevant subtree in a few cases.

Limit index updates to relevant subtree in a few cases.
-------------------------------------------------------
> Petr Rockai <me at mornfall.net>**20100509102248
>  Ignore-this: fea041133d039cecead73935f0cd6762
> ] hunk ./src/Darcs/Commands/Diff.lhs 210

> -               restrict <- restrictSubpaths repository subpaths
> -               restrict `fmap` readUnrecorded repository >>= (flip writePlainTree (toFilePath ndir))
> +               readUnrecorded repository subpaths >>= (flip writePlainTree (toFilePath ndir))

Basically, it looks like we're taking these smarts from the diff command
and refactoring them into readUnrecorded.

The smarts in question are explained in Petr's haddock:
>  -- | Obtains a Tree corresponding to the "unrecorded" state of the repository:
> +-- the working tree plus the "pending" patch. The optional list of paths (it is
> +-- ignored if empty) allows to restrict the query to a subtree.
> +--
> +-- Limiting the query may be more efficient, since hashes on the uninteresting
> +-- parts of the index do not need to go through an up-to-date check (which
> +-- involves a relatively expensive lstat(2) per file.

The commands that consequently inherit said smarts are

> hunk ./src/Darcs/Commands/Remove.lhs 101
> hunk ./src/Darcs/Commands/Replace.lhs 171
> hunk ./src/Darcs/Commands/WhatsNew.lhs 113

There's also this bit which does not change any behaviour

> hunk ./src/Darcs/Repository/Merge.hs 71
>       debugMessage "Checking for unrecorded conflicts..."
>       have_unrecorded_conflicts <- checkUnrecordedConflicts opts pc
>       debugMessage "Reading working directory..."
> -     working <- readUnrecorded r
> +     working <- readUnrecorded r []

> -readUnrecorded :: (RepoPatch p) => Repository p C(r u t) -> IO (Tree IO)
> -readUnrecorded repo = readIndex repo >>= I.updateIndex

> +readUnrecorded :: (RepoPatch p) => Repository p C(r u t) -> [SubPath] -> IO (Tree IO)
> +readUnrecorded repo paths = do
> +  relevant <- restrictSubpaths repo paths
> +  readIndex repo >>= I.updateIndex . relevant

And here's the meat of the change.

-- 
Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow>
PGP Key ID: 08AC04F9
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.osuosl.org/pipermail/darcs-users/attachments/20100608/c9d2d78d/attachment.pgp>


More information about the darcs-users mailing list