Jason,

I just did a test on a local table:

create table maintable ( pk i, cdata c(254) )

I created an updateable view on it, and ran several tests:

insert 10000 rows, tableupdate all at once = 41 seconds
insert 100 batches of 100 rows, tableupdate each batch = 22.37 seconds
insert 10 batches of 1000 rows, tableupdate each batch = 22.43 seconds
insert 1000 batches of 10 rows, tableupdate each batch = 22.55 seconds
insert 10000 rows, tableupdate each row = 23.06 seconds

insert 10000 rows directly into maintable = 3.1 seconds

so a couple of things stand out in this test:

1) pviews are at best 1/8th the speed of working with the table itself.

2) there appears to be a non-linearity based on the number of rows involved in the tableupdate.

You might benefit from adding an index:

index on deleted() tag delindex

Or by processing the files and periodically committing the updates during the scan.


>It is not unusual for a single location to have a load file that has 10000 lines (a line representing one product). Most of the parameterization in this project is to seperate the individual locations from one another.

>The adding of records seems to be reasonable, but the deletion of records makes the tableupdate drag.
>
>There must be something I am doind wrong with views.