How does DBISAM 4.x construct its fulltext index?
1) Does it have to create an index entry for each row in the table? So if there are
100,000 rows with 100 rows each, the index would have 10 million rows.
2) Or does it create a bitmap of all words used in the table? (similar to Rubicon)

TIA
Sam

Sam <sam@nothere.com> wrote:

How does DBISAM 4.x construct its fulltext index?
1) Does it have to create an index entry for each row in the table? So if there are
100,000 rows with 100 rows each, the index would have 10 million rows.
2) Or does it create a bitmap of all words used in the table? (similar to Rubicon)

TIA
Sam

Sorry, that should have read "1) Does it have to create an index entry for each row in the
table? So if there are
100,000 rows with 100 *words* in each memo, the index would have 10 million rows."

Sam

Sam,

<< How does DBISAM 4.x construct its fulltext index?
1) Does it have to create an index entry for each row in the table? So if
there are 100,000 rows with 100 rows each, the index would have 10 million
rows. >>

This is how it does it (with your later correction to the math).  It then
uses these index entries to build bitmaps as necessary to represent the sets
of rows selected.  However, remember that the index entries are:

a) Unique words only
b) Compressed using both prefix and trailing space index key compression

--
Tim Young
Elevate Software
www.elevatesoft.com