>Wanted to determine exactly how much I would gain by replacing a lot of SEEK's with INDEXSEEK's, and much to my surprise the answer was zero.
>
>Here's what I did:
>
>Set up table 1, Field1 C(10),Filler C(100). Added 1 million records, populating Field1 with a random integer from 1 to 1 million. Indexed on field1.
>
>I then wrote a routine that did 10,000 SEEKs into Table 1, followed by 10,000 Indexseek's, and back and forth.
>
>This was done on a Compaq laptop, Win 98, 32MB RAM. The data table was 111MB and the CDX was 5MB.
>
>The only noticeable trend was that each run was faster than the one before, suggesting that the file cache has much more to do with this than anything else.
>
>The first run, 10,000 SEEKs, took 245 seconds. The second run, 10,000 INDEXSEEKs, took 190 seconds. Next was 10,000 SEEKS at 174 seconds, and 10,000 INDEXSEEKS at 160 seconds.
>
>This test may not approximate everybody's usage, but it does approximate the usage here in one spot: many hits to the same table by the same user. My conclusion there was that there was no advantage to changing SEEKs into INDEXSEEKs.

My understanding about indexseek was not that IT would be fasted but, it does not move the record pointer.

Jeff