At 8/3/2000 07:02 AM (Thursday), Benny Schaich wrote:

>To make it even harder, we put some text into blob fields and parsed
>20,000 of those with a LIKE statement. IB always came back in less then
>a second.

[snip]

> From an architectural point of view I'd like to know how that works. Did
>we trick ourselves somehow?

I cannot answer directly about how this is implemented, but I would point
out that 20,000 records, even with large blobs, is a small enough set that
they don't really represent a thorough test of the implementation. Today's
hardware can fit that many records into memory quite easily and memory
searching is damned fast today. So, I'm not surprised at the outcome.