Re: [rdiff-backup-users] The rdiff-backup poll!

[Ben Escoto]

>>>>> dean gaudet <address@hidden>
>>>>> wrote the following on Wed, 21 Dec 2005 19:15:42 -0800 (PST)
> 
> it looks like you're keeping data about every file in memory...  i hacked 
> around that in mine by trimming the base component of every path before 
> adding it into the hash tables... so there are only as many hash keys as 
> directories, which is a bit more manageable.
> 
> if you want a more general solution you might consider invoking external 
> sort(1) ... sort(1) is geared towards sorting inputs which could be as 
> large as /tmp or $TMPDIR without consuming all of memory.  but that can be 
> slower on smallish inputs because you end up parsing strings multiple 
> times.

Ok I used a more complicated algorithm so it runs in constant space
while still being able to report individual files and not just
directories.  I profiled it but even after a few optimizations it's
still about as slow as the original perl script.


-- 
Ben Escoto

pgpG9766D3AAX.pgp
Description: PGP signature