Re: [htdig-dev] Re: Found it! (Was: Bug :( )

[Lachlan A. <lh...@us...>]

On Wed, 2 Jul 2003 04:24, Neal Richter wrote:

>   It is possible to create two different environments for the
> different DBs in the htdig classes that control them.

True, but the "_weakcmpr" database is internally created by  mp_cmpr.c=20
for any compressed database -- ht://Dig knows nothing about it.  Of=20
course, we could *change* the API to pass in two database=20
environments (one for the database proper and one for _weakcmpr), but=20
that is far from fixing it at the API level.

>   I am hesitant to put lots of time into tweaking BDB code
> directly.
>   2)BDB is a VERY widely used piece of software and it is
> incredibly likely that most of the problems we encounter can be
> fixed at the BDB API level in our classes.  This of course excludes
> our (Loic's) hacks to have ZLIB page compression.

Yes, it is exactly Loic's code (mp_cmpr.c) that I was proposing to=20
fix.  In the past I hacked  mp_alloc  to try to work around the bug,=20
but I had planned to back those changes out once  b5  was out and=20
there was time for a real fix.

> The more we tweak the more we
> diverge from stock BDB code and the more work we make for ourselves
> long term.

Agreed.

>   I used an STL hash and improved insertion time in the WordDB
> considerably.  Probably due to queing up all the inserts in larger
> batches reduces overhead.

That sounds impressive.  OK -- I'll stop tinkering with the database=20
code...

Cheers :)
Lachlan

--=20
lh...@us...
ht://Dig developer DownUnder  (http://www.htdig.org)