On Thursday 28 November 2002 16:19, Bill Moseley wrote:
> I don't think that makes sense.  You can't sort a list by two different
> primaray keys.

	True, i didn't explain myself to well but by the issue of only having one 
property value per property per file this isn't really relevant any more :)

> But note that combining more than one -L
> select will AND the selects.  When you use -L the code is throwing out
> results that do not fit within the range, so any -L that fails will remove
> the result from the final result list.
>
> So you can't say select releaseA between X and Y OR releaseB between X and
> Y.
>
> If that's what you want then you might consider splitting each record into
> multiple "files" (as far as swish-e think) and index each one with a date.
>
> So if some record has two event dates then you index the record twice,
> except you use different release dates for each one.  Not perfect, of
> course.  You will need to worry about duplicate results.

	ANDing the selects is OK. I now have a number of different release fields 
that hold the min and max values and depending on what search is required a 
different release property is used and for the initial testing that i have 
done so far works well.

> > 	Also how is the MetaNamesRank directive going? This sounds interesting
> > and could well turn out to be a worthwhile feature.
>
> Eh, well, ah, need to get back to that some day soon.  What I did was to
> use -S prog and then format the data as HTML and then used tags like <h1>
> and <em> to make some text have a higher rank value than others.

Will this work when using 'Document-Type: XML2' when passing the data via 
stdin? Will example #1 below have a higher score then example #2 when 
searching using -w 'recording.title=bush'?

Also is there a way to produce a high score for phrases that match the whole 
value in a meta name. Can example #1 get a higher score then example #3 when 
searching using -w 'recording.title="bush"'?

	Example #1
		<recording.title><h1>Bush</h1></recording.title>

	Example #2
		<recording.title>Bush</recording.title>

	Example #3
		<recording.title>Kate Bush</recording.title>

Hmmm, now i wonder if anybody can guess what type of data im indexing :)

And for what its worth (dont want to seem like all i do is take) i have a php 
class on my web site (in the 'source code' section) that seems to do a good 
enough job of wrapping swish-e and its results. There is no documentation at 
the moment but if you understand php and my style of coding, heh, then 
hopefully you will work out how it works.

-- 
Regards,
	William Bailey.
	Pro-Net Internet Services Ltd.
	http://www.pro-net.co.uk
	http://wb.pro-net.co.uk