Johan Corveleyn wrote on Wed, Jun 13, 2012 at 15:32:56 +0200:
> On Wed, Jun 13, 2012 at 3:16 PM, Vladimir Shun'kov <shunya.v_at_gmail.com> wrote:
> > Would be enough have the number bytes of changes. I did disk usage in
> > pre-commit hook and disk usage in post-commit hook and then calculated
> > the difference. But I commited the new file with size 20Kb and size of
> > commit I recieved only 5Kb due some svn compression I guess.
> In 1.7, there's a new svnlook subcommand: 'svnlook filesize'. This can
> give you the filesize of any file (from a revision or a transaction),
> without having to 'svnlook cat' it (i.e. more efficiently).
> Maybe you can use that directly, or use it to calculate the difference?
> Like, in the pre-commit hook:
> 1) 'svnlook changed -t $TXN $REPOS' to get the list of changes in this
> 2) For every changed (or added) file, do 'svnlook filesize -t $TXN $REPOS'
> 3) For changed files, compare them to 'svnlook filesize -r $YOUNGEST
> $REPOS' where $YOUNGEST is the result of 'svnlook youngest $REPOS'.
> 4) Do some arithmetic on the result
$TXN may be based on a revision earlier than $YOUNGEST, and the file may
have changed between that base revision and youngest. If that happens,
the commit attempt will fail due to a conflict, but before that happens
the logic above would make calculations based on the wrong information.
I don't know if there is an svnlook API to print the base revision of
a given txn... but I guess there should be.
Received on 2012-06-15 20:43:13 CEST