Skip to main content.
home | support | download

Back to List Archive

Re: Indexing MySQL data

From: Bill Moseley <moseley(at)not-real.hank.org>
Date: Fri Jul 20 2001 - 16:14:07 GMT
At 03:03 AM 7/20/2001 -0700, tooms wrote:
>Support of Mysql in 2.0.5 version ?
>
>Hi, I'm currently installing swish-e in order to
>index data from Mysql, but I use 2.0.5 version of
>swish-e. I was told mysql indexing can only be indexed
>since 2.2 version (Is not 2.0.5 the last stable
>version ?)

Yes and no.  You could use filters to index data in a MySQL database in
2.0.5.  You can create a file for each record of the database and use a
filter program to query the database.  You can also use a single file to
trigger a filter, which then indexes all the records using the LST parser
which splits one large XML file into sections (swish records the offset of
the "record" within the XML file).

Neither of those methods are really that satisfactory.  You will be much
better off using the dev version and writing a program to query the MySQL
database using the -S prog feature.  There is at least one person on this
list doing that now.

Don't confuse "stable" with "bug free".  There have been many bug and logic
fixes to the 2.0.5 code that are now in the development code.  I think the
biggest risk to using the dev code is some new config directive my change
after you start using it.  Bugs end up in the dev snapshots, but they tend
to get fixed really fast - especially when someone using the dev version
finds a bug.  For indexing MySQL data I think you really need to use the
dev version.

The reason the development code has not been "released" is not because it's
not stable or that there are known bugs, but rather because of design
issues and available development time to address those design issues.
Namely, there are plans to improve the ranking code, and adjustments to the
index file format to make merging faster and to reduce the memory limiting
factor of indexing very large data sets.  Those are issues that come up on
this list often.



Bill Moseley
mailto:moseley@hank.org
Received on Fri Jul 20 16:14:25 2001