I fear that you have driven yourself into a corner. Instead of deleting records, you should have a logical field called for instance Reusable. So instead of deleting a record, you set this field to true. And instead of inserting a new record, you first look for a reusable record. This is how for instance SQL server works internally, AFAIK. Something like this

select top 1 pk from mytable where reusable order by reusable into array laPk
if _tally>0 && A reusable record is found
   update mytable set field1=myvalue1, field2=myvalue2, reusable=.T. where pk=laPk
else
   insert into mytable (field1, field2, pk) values (mavalue1, mavalue2, generateGuid())
endif

My pk is always a guid, created by the generateGuid() function. Remember, a Primary key shall not be meaningful, but guaranteed to be unique, which is what guid stands for, Globally Uniqie IDentifier.


>Thank you. I will have to figure out a different approach. I will try PACK.
>
>UPDATE: Pack does not work either.
>
>>No. ZAP cannot be used against a buffered cursor/table.
>>
>>>Do you know (of course, I will test it today) if ZAP can be issued against a Cursor Adapter cursor? Because I am trying to avoid opening the table with USE command but rather do everything via CA cursor.
>>>
>>>>After DELETE you'll have table with 10,000 deleted records which can make even SEEK() slow. On other hand ZAP requires exclusive access to the table. I would go with later, if possible.
>>>>>
>>>>>I am opening and updating tables using Cursor Adapter. In one of the procedures I need to zap all records in a table. I can fill the CA cursor with all records, delete records, and then call tableupdate(). But my questions is, how practical this approach if the table can have up to (max) 10,000 records. This is a VFP database (not SQL Server).