RE: CHAIN Versus SETLL and READ When Data Needed

From: Alan Campin

I know this is a question about which is more efficient but wouldn't the
answer be neither?

In other words, wouldn't this be much more efficient to use SQL
especially if you have multiple records?

Well, it MIGHT perform better on SQL for multiple updates, but you'd have to
be very careful in your coding.

It would have to be something like:

UPDATE filename SET field=newvalue
  WHERE key1=tgtkey1 and key2=tgtkey2 and key3=tgtkey3

Followed by:

UPDATE filename SET field=newvalue
  WHERE key1=tgtkey1 and key2=tgtkey2
  And NOT EXISTS
    (SELECT 1 from filename
     WHERE key1=tgtkey1 and key2=tgtkey2 and key3=tgtkey3)

Followed by:

UPDATE filename SET field=newvalue
  WHERE key1=tgtkey1
  And NOT EXISTS
    (SELECT 1 from filename
     WHERE key1=tgtkey1 and key2=tgtkey2)


Ignoring issues of atomicity (where a records gets added, updated or deleted
between two of those statements), you still have to figure out how to modify
each of those statements to process multiple records (the tgtkeyX fields
would actually have to come from another file).  I think if you actually
tried to write the code, you'd find it quite a challenge.


What Shannon is asking for is far more a traditional database-driven
solution, where the logic executed depends on records or fields in the
database.  This continues to remain the domain of native I/O.

Joe