At 12:56 PM 08/07/02 -0700, Frits van Latum wrote:
>I have created a small index of 6 xml-files (downloaded from my database) in
>a directory called "f", with command:

..
>had 1 record as result, which is correct. I tried several other queries and
>everything was ok.

..

>These 6 records are from my database which is indexed by the following
>The database has about 150.000 records. The indexfiles are 57Mb and 15 Mb.
>
>The same search as above gives 11 records, which is wrong.

Hum, well what I was hoping for was a small set of files I could install on
my machine and see the same problem you are seeing.

It makes it hard to debug if I can't duplicate the error, of course.  If
you can make it happen with just a few files (100?  1000?) and send them to
me in separate email that would be very helpful.

So try this: run this command and copy-n-past the results to the list.

  swish-e -w record=(Royalton-Kisch) -L begindatum '<=' 1900 -L einddatum
'>=' 1800 -x 'begin=<begindatum> end=<einddatum>\n'

You should not see any begin dates later that 1900 or end dates before 1800.

I do remember some problem with the <= or >= operators, but I can't find
any notes -- so I'm not sure if it was fixed or not (or if I just imagined
it).

Also try:

  swish-e -w record=(Royalton-Kisch) -L begindatum 0 1900 -L einddatum
1800 9999 -x 'begin=<begindatum> end=<einddatum>\n'

>So is a "-S prog index" different from an index based on files?

No, not really.

>Is the size of the index a problem?

Not that I can imagine, but bugs show up in weird ways.



-- 
Bill Moseley
mailto:moseley@hank.org