Skip to main content.
home | support | download

Back to List Archive

RE: Problems w/ wildcarding (follow-up).

From: Thomas M. Parris <parris(at)not-real.isciences.com>
Date: Thu Apr 05 2001 - 18:53:49 GMT
SWISHers:

As a follow up to my previous message on this subject ...

1. I see from the archives (and my own experience) that the problem is
platform specific, not data specific.  I have also narrowed the problem down
to the indexer, not the search engine.

2. I have done a bit testing comparing indexes produced by David Norris'
windows binary and the one I generated with my Cygwin build.  My analysis at
this point is only cursory, but it looks to me like the problem is likely to
be found in slightly different sorting behaviors (or an off by one error).
The indexes tend to be very similar except the order of the locations moves
around within the list for a given word.  The index produced by the windows
binary presents numbers in set of ordered sequences, where as the one
produced by Cygwin has sequences are rotated by one.  For example, sequences
such as A B C D (where A<B<C<D) are rotated in the Cygwin version to B C D
A.

3. I have seen David's messages in the archives disparraging Cygwin,  and
have opted to use a build based on MSVC for my project at hand.  None the
less, it seems that it would be useful to offer Cygwin to the list of
supported platforms (and finding the incompatability may fix other lurking
problems).  For those of you steeped in the internals of SWISH, I hope this
message gives a clue where to look.  The fact that packages as complex as
Perl, PostgresSQL and MySQL can run succesfully under Cygwin indicates to me
that it is likely SWISH could as well.

4. A minor note for completeness ... the problems I mentioned below persist
with gcc (2.95.3-2)/cygwin.

5. Please view the above as constructive.  I am most grateful for the work
you all put into SWISH.  It is proving to be an ideal tool for my task at
hand.

With best regards,
Tom Parris
------------------------------------------------------
Thomas M. Parris
Research Scientist & Executive Director, Boston Office
ISCIENCES, LLC
685 Centre Street, Suite 207
Jamaica Plain, MA  02130
voice: 617.524.8041        email: parris@isciences.com
http://www.isciences.com/    http://www.terraviva.net/
------------------------------------------------------


-----Original Message-----
From: swish-e@sunsite.berkeley.edu
[mailto:swish-e@sunsite.berkeley.edu]On Behalf Of Thomas M. Parris
Sent: Thursday, April 05, 2001 12:41 AM
To: Multiple recipients of list
Subject: [SWISH-E] Problems w/ wildcarding


SWISHers:

I'm just getting started with SWISH-E 2 (swish-e-2.0.5) on Windows 2000
built with gcc (2.95.3-1)/cygwin.  On the whole, things seem to be working
well.  The test suite works and I have been able to build and search some
indexes.  However, I am having a curious problem with wildcards.

Most wildcard searches work exactly as they should.  Examples include:

	swish-e -w 'deposit*' -f UNCLOS
	swish-e -w 'tec*' -f UNCLOS
	swish-e -w 'pa*' -f UNCLOS
	swish-e -w 'pq*' -f UNCLOS

However:

	swish-e -w 'prospect*' -f UNCLOS
	swish-e -w 'pr*' -f UNCLOS
	swish-e -w 'p*' -f UNCLOS
	swish-e -w 'te*' -f UNCLOS
	swish-e -w 't*' -f UNCLOS

all hang.

Any thoughts?  Is this a problem with SWISH-E 2, or is it a problem with my
platform?

-- Tom
------------------------------------------------------
Thomas M. Parris
Research Scientist & Executive Director, Boston Office
ISCIENCES, LLC
685 Centre Street, Suite 207
Jamaica Plain, MA  02130
voice: 617.524.8041        email: parris@isciences.com
http://www.isciences.com/    http://www.terraviva.net/
------------------------------------------------------
Received on Thu Apr 5 18:54:45 2001