re:re tsearch2 problem

THANKS very very much again for Teodor Sigaev 's help.

really thanks.

>> I only know that GiST is also  "Inverse Order OF Idexing",but of course
>> it is a tree NOT as lucene does

>Tsearch2 doesn't use inverted index (yet), and so it easy to update (your index
>is always full - online indexing), but it hasn't good performance on large data
>set. The limit is about 10^5 - 10^6 rows in practice.

I know the LUCENE " inverted index " is just like follows:

lexeme     doc_id[frequence]   location
guangzhou   1[2]               3，6
he          2[1]               1
i           1[1]               4
live        1[2],2[1]           2，5，2
shanghai     2[1]               3

so ,before I  thought that A node In GiST tree is very like the above :

In the tree's node , it has several value for storing its's

turple_id[frequence]   location ,

and that is why I thoutht the " inverted index " IN tsearch2.

BUT now , I'm lost again ,how does Gist tree store lexem and how to find his correspoding tuples in database( including  frequence   location)?Or to say ,how TSearch2  implement the Lucene's "inverted index" mean ?