[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [RFC] A libsvn_wc properties optimisation

From: Philip Martin <philip_at_codematters.co.uk>
Date: 2004-11-06 15:28:34 CET

Barry Scott <barry@barrys-emacs.org> writes:

> The win is that you only have to do a file open on one file rather
> then lots of files/directory. I'm sure that parsing one file is far
> faster then open 10 files. The file sizes are trivial. I have 1856
> bytes for 54 files in my repo for example.

The "I don't have any large properties, why should anyone else?"
argument, that came up last time as well.

Suppose one of the files has a 1MB property, now every operation that
reads properties on any file now has to read that all that data.

> I not sure what you are getting at with caching in memory. A command
> like svn_client_proplist will open the file, read and parse and return
> the info in an apr array alloc'ed in a pool. Where is the caching?

Consider

 do_something(directory):
   foreach (item in directory)
     if (item is a file)
       do_stuff_that_uses_file_properties(item)
     else if (item is a directory)
       do_something(item)

How often does do_stuff_that_uses_file_properties need to read that
single properties file?

What about checkouts? If you simply re-write all the properties seen
so far as each file arrives then the total amount written goes up
quadratically (appending to a file is difficult to do atomically). We
had exactly that problem with the entries file.

I have no interest in implementing your idea. Patches welcome as
always!

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Nov 6 15:28:59 2004

This is an archived mail posted to the Subversion Dev mailing list.