apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mladen Turk" <mt...@mappingsoft.com>
Subject RE: [PATCH] apr_hash.c -- Make table ordered
Date Thu, 11 Oct 2001 16:15:34 GMT

> -----Original Message-----
> From: Justin Erenkrantz [mailto:jerenkrantz@ebuilt.com]
> Sent: 10. listopad 2001 23:10
> To: Mladen Turk
> Cc: APR Dev List
> Subject: Re: [PATCH] apr_hash.c -- Make table ordered
> 
> On Thu, Oct 11, 2001 at 07:58:04AM +0200, Mladen Turk wrote:
> > The question is where the hash table is used and for what purpose,
not for
> > the sake of the theory itself.
> > How many entries are in the table itself? It is not used to store
couple of
> > thousand records, not even couple of hundreds.
> 
> That can't be realistically answered by the APR library.
> 
> In point of fact, mod_mbox uses the hash table to store all of the
> message IDs when computing threads.  (It uses DBMs to store the
> complicated information, but it uses the hash when doing the threading
> calculation because of the insanely fast in-memory lookups keyed off
> of the Message-IDs.)
> 
> We have at least one mbox file with about 2,000 messages.  -- Justin

It was just a figure of speech :)

An excellent example!

What is the real table size (containing messages) compared to 2000
pointers?
What is the overhead of converting hash tables to a SLL (MBOX_LIST) you
are using now, compared to simply sorting pointers?  
The patch does not interfere with the in-memory lookups at all.

MT.


Mime
View raw message