Mikhail Ramendik <[EMAIL PROTECTED]> said:

> Aaron Stone wrote:
> 
>> I haven't read the patch over yet, but I think Paul has (?). Wasn't there
>> a thread about Firebird and/or Outlook problems with this or another
>> patch? Let's make sure that all of those are resolved.
> 
> This was with the CVS HEAD version. My existing speedup was intended for
> 2.0.x only, and so the patch is against 2.0 not CVS HEAD, and the
> changes are minimal. 

Oh, cool. So that I don't have to go digging in the archives, would you be
able to post the latest versions of these to your website?

[snip]
> Actually there are questions for the 2.1 speedup - basically, how far
> are we willing to go with changing the database interface in 2.1. The
> "maximum" variant involves using a big query instead of many small ones,
> but at least for mysql this means a whole set of new/changed db basic
> API functions, to enable the use of mysql_use_query instead of
> mysql_set_query for this. I'm not sure about pgsql yet.

I think that 2.1 should require MySQL 4.1, which means sub-selects and
transactions can become an integral part of our queries.

[snip]
> Actually I have two ideas on SEARCH. One is a change in header storage,
> similar to what you described (involving three tables instead of two,
> but this is a technical detail that can be worked out later). But I'm
> not sure this will work for 2.1, that is, before a BIG partial rewrite
> that is certainly needed (the language is another question) and is going
> to be -ng or 3.0. 

If we can't get three header tables in 2.1, then I'll settle for two ;-) 
We've had big partial rewrites already, and experience shows that they can
take up to a year each. This is to be expected, I think.

> But I also have a simpler idea that will still speed things up. It
> involves using a regexp SQL fulltext search. While it's definitely slow
> compared to a changed storage (the db won't be able to use indexes),
> it's still faster because it does not involve any parsing, and uses the
> well-tuned SQL server code to do fulltext searching. 
> 
> The advantage is that it will work on current storage (with is_header),
> and therefore can be a better interim solution for 2.1, befire a new
> storage layout is fully supported. 

We could make this change during 2.0.x, actually. It's only code and
queries that are changing, not the database structure or other user
interfaces.

Aaron

--

Reply via email to