Re: Changes in /contrib/fulltextindex

Hi Florian,

> I have done some changes and improvements to the fulltextindex trigger
> (contrib/fulltextindex). As these changes affects among other things the
> parameter list I would like to ask the maintainer about his thoughts
before
> I submit a patch. And there comes the problem - there is no explizit
> maintainer listed in the accompanying docs. The last changes according to
> the cvs log has been done by Bruce Momjian.

The most recent patches were submitted by me, so I guess you could call me
the defacto "maintainer".

> Maybe the original author is subscribed to this list and can get in touch
> with me because I'm kind of new to incorporation patches to a open source
> project.

Cool, reply to me personally if you need technical help, or if you want me
to review your patch and use the list to discuss things that affect the
project...

> The changes made include:
>
> + Changed the split up behaviour from checking via isalpha to
> using a list of delimiters as isalpha is a pain used with
> data containing german umlauts, etc. ATM this list contains:
>
> " ,;.:-_#/*+~^°!?\"\\§$%&()[]{}=<>|0123456789\n\r\t(at)µ"

Good idea. Is there a locale-aware version of isalpha anywhere?

> + If the do be indexed field has'nt changed, the indexing will
> not be done. This way unnecessary redindexing of fields not
> affected by an update can by omited.

Fantastic! That was on my list of things to do!

> + There is a new field 'word' of type BOOL in the index table.
> Using this field, it is possible to do 'full word' and
> 'substring' searches.

Hehe - that was another idea I had as well. Breaks back compatibility.

> + The text gets no longer lowercased before it is written into
> the index table. This way it is possible to to casesensitive
> and incasesenstive (via a functional index using lower) searches.

ok

> + Added functionality to prevent indexing of duplicate words
> (this is one item of the todo list in fti.c) using a hash
> table. As this is chained to a significant loos of performance
> and depends on the indexed data, this can be turned on or off
> via a newly introduced parameter.

ok.

OK Florian, can you please send me your new contrib/fulltextindex directory
tarred up?

List: what should we do about the backward compatibility problem?

Chris