Hi!

>Yes, I thought about it too, but unfortunately, we don't have them organized this way.
>The typical examples :BAD
>UAD, etc.

Is it problem to make simple routine to sort them? It is jusy 1 run for large table, but when it will be kept sorted in future, you will never need to do this again.

>BTW, I ran already bunch of tests, using $ operator and Sergey's ideas and found, what for 2 items $ operator is better, for 3 items first Sergey's idea works faster, but time difference is ~20 sec. and time values like:
>&& 3 items
>214 s && 2 chrtran
>243 s && 3 $
>227 s && occurs + chrtran

Did you tried to run query with 'LIKE' operator as I mentioned? I'm very curious about speed, is my idea worth enough?

>I run these quieries directly from the server and on one table (1,8 mln. records). From my local machine it should be worse...

Why you thinnk so? All these expressions don't use indexes for optimization (empty function or chrtran or '$' operator or LIKE operator), so no extra overhead because indexes. Delay will be just in data transferring.

In addition, you have good reason to move that database to SQL server... Just a thought.