Tim

Any idea about comparative speeds for DBISAM and ElevateDB especially in the full text indexing arena and extra especially in terms of deleting records?

Roy Lambert

Roy,

<< Any idea about comparative speeds for DBISAM and ElevateDB especially in
the full text indexing arena >>

Full-text indexing should be faster with EDB, especially if you pay close
attention to the most desirable word size setting.

<< and extra especially in terms of deleting records? >>

Much faster, especially with BLOBs and lots of indexes.

--
Tim Young
Elevate Software
www.elevatesoft.com

Tim

><< and extra especially in terms of deleting records? >>
>
>Much faster, especially with BLOBs and lots of indexes.


Yipeee a VERY positive reason to change

Roy Lambert

Tim


Been looking at the OLM and there's only the one length criteria for words so I'm guessing we can specify the max length but not the minimum?

Any (high) chance of a minimum word length in the future?

It also looks like Stop Words, Space Characters and Include Characters have vanished. Is that right?

Roy Lambert

Roy,

<< Been looking at the OLM and there's only the one length criteria for
words so I'm guessing we can specify the max length but not the minimum?

Any (high) chance of a minimum word length in the future? >>

You're mixing up two separate things.  The indexed word length is only there
for the purposes of determining the size of the index keys.  To control how
words are actually parsed, etc. you would use a custom word generator.  The
next build will include the extensions in the object repositiory for
creating a custom word generator module (DLL) for use with the text
indexing.

<< It also looks like Stop Words, Space Characters and Include Characters
have vanished. Is that right? >>

Yep, same thing - word generator module.

A quick note - the default stop words, etc. will be documented in the
manual, and basically they are pretty standard fare so they should cover any
English and Western European indexing fairly well.

--
Tim Young
Elevate Software
www.elevatesoft.com

Tim

I said I needed a course in how to read manuals


Roy Lambert

Tim


>You're mixing up two separate things. The indexed word length is only there
>for the purposes of determining the size of the index keys.

Sorry, I've just read the OLM again and I don't understand this can you explain for me please?

>To control how
>words are actually parsed, etc. you would use a custom word generator. The
>next build will include the extensions in the object repositiory for
>creating a custom word generator module (DLL) for use with the text
>indexing.

OK I can "JUST" port my current code

Roy Lambert

Roy,

<< Sorry, I've just read the OLM again and I don't understand this can you
explain for me please? >>

Well, the OLM is basically non-existent on this subject, so I wouldn't
expect you to get much from it.

The default word generator is built into EDB and covers the English language
and most common stop words, etc.  It's a general-purpose word generator.  In
order to affect the way the word generation is performed, you would create a
custom word generator using the Delphi object repository template that we
will have in the next build.  It only has one event handler, and that event
handler will allow you to spit back words to EDB as required when EDB needs
to break up any text into words for indexing.  After compiling the word
generator module, you would then just drop it into the same directory as the
EDB configuration file being used, and then you can create the word
generator in EDB using this SQL statement:

http://www.elevatesoft.com/edb1sql_create_word_generator.htm

Then, just simply create the index using this SQL statement:

http://www.elevatesoft.com/edb1sql_create_text_index.htm

being sure to specify that the word generator you want to use is

--
Tim Young
Elevate Software
www.elevatesoft.com

Tim


Obviously didn't explain well - what's baffling me is the indexed word length....

Roy Lambert

Roy,

<< Obviously didn't explain well - what's baffling me is the indexed word
length.... >>

The index keys are not variable-length in terms of how they are stored.
Therefore, even full-text indexes need a length to use for the size of the
full-text index keys.  Normally, the index key size is determined by the
size of the columns that comprise the index.  However, this won't work for
full-text indexes for obvious reasons, so EDB needs you to tell it what the
index key size should be.  20-30 characters is usually a good indexed word
length to use, with the lower the number, the smaller the index.

--
Tim Young
Elevate Software
www.elevatesoft.com