>Stephen,
>
>>Index on deleted() even if none are. Index on every column sepeciall if this is your trasnaction table.
>
>Well, Your perception of the deleted() tag is an old one. Recently we believe that the deleted() tag is killing performance especially on large tables. Besides that if there are no deleted records there is no way that vfp could benefit from the deleted tag except for the following exceptions:
>
>- You want to use FILTERED sql results.
> (Alternative Use the table again and set a filter)
>
>- You want to optimize the COUNT command.
>
>In fact index on all columns is not a good idea. If you only index the most important (and selective) columns, you won't need the other columns (as they could be more like a burden than it helps you). When using a trancaction table it is generally a bad idea to have many indexes since this slows INSERT, UPDATE and DELETE actions significant.
>
>Recently an article in fox advisor (or was it foxtalk ?) explained this whole topic.

But in this case of Statistical exam of a transaction table with 15 mill rows, only 8 cols' wide in this case, you can and should index on all cols. It's not like your indexing on description, qty ordered, qty shipped, etc.

And when you are doing decission support exercises you will slice and dice the data set every which way. So doing it on the front end will save you time.


Not that I have not heard about the Detelted Tag, change, but did anyone inquire if my insert routine used a deleted() tag?

So thanks to all of you who pointed out the change in thinking, now I have to decide if I want to address my whole insert into approach. As it stands now NO. NO Way, Don't Fix It If It Aint Broke, all come to mind.

I'm not yelling at you, more at the community at large. Just casue M$ makes a subtle {yeah right} change in the core of data handling, doesn't mean that I throw away years of working code in tooooooo many apps and places. to quote an older Bush, "Wouldn't be prudent."

I'll bring this issue up at our SIG this Thurs here in Memphis.

We will attempt to see speed of large data sets.