From: Joel VanderWerf <vjoel@...>
Date: 2006-05-18T01:58:43+09:00
Subject: Re: ruby sdbm library reliable?

Kirk Haines wrote:
> On Wednesday 17 May 2006 7:48 am, Jamey Cribbs wrote:
> 
>> I've been doing some thinking about Logan's idea of offering a choice of
>> multiple backends.  That sounds intriguing.  I could see that idea,
>> coupled with something like qdbm, to provide different options for the
>> user.
> 
> I wrote something a while back that might be helpful (or not).  I wanted a 
> lightweight mechanism to persist data to disk via a hashlike interface.  I 
> didn't want to use pstore because one pays a pretty high cost for each 
> transaction once the data store gets any significant amount of data in it, 
> and one has to go through extra effort to make pstore multiprocess safe.  I 
> also wanted it to have flat RAM usage, regardless of the number of items in 
> the data store.
> 
> What I ended up with is something that can be used as an LRU cache or simply a 
> hash.  It is completely disk based and is multithread/multiprocess safe.  It 
> still needs some love, but it works.
> 
> It's slower than pstore (when starting a transaction, performaing an action, 
> closing a transaction) for small numbers of items in the data store, and much 
> faster than pstore once the data store has a few thousand items.
> 
> I'm planning on packaging it up as a seperate micropackage in the IOWA area on 
> Rubyforge.  It's probably no faster than Kirbybase is now, but if it might be 
> useful in some capacity, I'll bump that up on my priority list.

Sounds interesting. I wrote FSDB with some of the same ideas in mind,
but probably we used different approaches. Looking forward to seeing it
when it comes out.

-- 
      vjoel : Joel VanderWerf : path berkeley edu : 510 665 3407