Progress 4GL Performance Gains By Avoiding Database Write Operation

Question

At my company we've recently been working on a modification to an existing program that used the following logic to perform some database field assignments:

db-buffer1.field1 = if db-buffer1.field1 <> db-buffer2.field2
                        then db-buffer2.field2
                        else db-buffer1.field1

I'm not 100% sure of the original programmer's intention here. Can anyone help me understand the value in comparing whether or not the field values are different before deciding whether to assign the new field?

If the comparison is 'false' and we end up assigning db-buffer1.field1 = db-buffer1.field1 do we avoid a write operation on the database?

Also note that this example is part of a larger ASSIGN statement that contains several fields running similar assignment/comparison logic. Does that have any effect on the value of this additional code? (i.e. Do all comparisons in the ASSIGN statement have to succeed in order to avoid a DB write?)

Tom Bascom · Accepted Answer · 2013-07-29T14:04:21.657

0

An "IF function" on the right hand side of an expression is a lot like a "ternary operator" in C or Javascript (the "?:" construct).

It only impacts the portion of the ASSIGN that it is part of. When I write code using those I always encase it in parens to make it clear. Like so:

assign
  a = ( if x = y then b else c )
  z = 2
.

Whether or not a WRITE happens depends a lot on the rest of the code.

Judging only by this snippet you are going to write SOMETHING (at least logically). db-buffer.field1 is going to get a value assigned to it regardless. The IF logic on the right-hand side is just picking if it will be field1 or field2. In the case that boils down to field1 = field1 you might hope that some lower layer will optimize the write out of existence. I don't see your Progress version posted but if it is v9 or better then it may very well be optimized away. (Prior to v9 it would not have been.)

If you really want to "avoid the write" at the application level you should code it like this:

if field1 <> field2 then
  assign
    field1 = field2
  .

This form does not use the IF function, it is just a normal IF ... THEN statement. It is much clearer and it does not depend on optimizations at a lower level. Of course the snippet shown is said to be part of a larger ASSIGN -- so it may or may not be sensible to break it out and write it as I've shown. But it would be worth thinking about.

edited Jul 29 '13 at 14:04

answered Jul 25 '13 at 20:09

Tom Bascom

13,405
2
27
33

"In the case that boils down to field1 = field1 you might hope that some lower layer will optimize the write out of existence" - That's pretty much what I was wondering. We're running 10.05B.32. Unless Progress is optimizing the write away - I don't see any reason to have the code convoluted with additional IF/THEN logic since in the end field1 will have the same value regardless. – pmartin Jul 25 '13 at 20:30
...and even if Progress does optimize the write at a lower layer we're really just replacing the write operation with a comparison operation. Certainly the comparison is faster but I'm not convinced that the performance gains are so significant that it warrants code that is far less readable and understandable for future maintenance programmers. – pmartin Jul 25 '13 at 20:32
I agree. It's silly, pretty much pointless and not at all clear. – Tom Bascom Jul 25 '13 at 20:52
1

If the content of field1 is unchanged by the write (assigned to itself) and nothing else changes that fact the field won't be written. You can test it out by adding a write-trigger in your code before the assign `ON WRITE OF db-buffer1 DO: MESSAGE "Writing db-buffer1" VIEW-AS ALERT-BOX. RETURN. END.` – Jensd Jul 26 '13 at 13:37

score 0 · Answer 2 · edited May 23 '17 at 12:21

It appears that the programmer is checking if field1 is not equal to field2 go ahead and update it.

They might as well simply use the following since either way the field is being assigned:

   assign db-buffer1.field1 = db-buffer2.field2.

When using the ASSIGN statement (recommended always), from a logical perspective you have to keep in mind that each line is assigned separately.

Tom has a great answer here that gives more information regarding efficiency/history of the ASSIGN statement.

score 0 · Answer 3 · edited Jun 18 '14 at 18:33

0

I would be very interested in seeing the rest of the assign statement, but I would also get rid of the if logic in this case as it adds no value, just and extra instruction. As Terry stated above

 ASSIGN db-buffer1.field1 = db-buffer2.field2.

edited Jun 18 '14 at 18:33

CRABOLO

8,605
39
41
68

answered Jul 29 '13 at 09:30

AquaAlex

360
7
17

Progress 4GL Performance Gains By Avoiding Database Write Operation

3 Answers3