Re: EVAL-CORR

[Barbara Morris <bmorris@xxxxxxxxxx>]

On 2011/1/3 9:02 AM, David FOXWELL wrote:
> I was looking at the code I wrote last year. Well, last week
> actually, but it seemed a long time ago. There I was, like the next
> developper called upon todo the maintenance, and I said to myself,
> "Why has he put EVAL-CORR? - that would suggest that the two DS were
> not the same". So, I went back to a move, as in my OP but then asked
> myself if was safe, hence my question.

Arguably, using EVAL or MOVE to assign data structures should be the 
case that causes head-scratching.

I realize that MOVE, MOVEL or EVAL were the only ways to assign two data 
structures for decades, so EVAL-CORR is naturally the odd case.  But 
just as the use of MOVE or MOVEL nowadays might cause "Huh? Why not 
EVAL?" head-scratching for ordinary fields, perhaps eventually EVAL-CORR 
will become the norm for data structures.

Using EVAL-CORR makes the code a little bit more self-documenting, for 
the cases where the names might not make it obvious that the variables 
are data structures.

I wouldn't worry too much about the performance impact of using 
EVAL-CORR vs EVAL or MOVE for the cases where EVAL-CORR would handle the 
subfields separately.  In those cases, where the data structures are not 
related by LIKEDS or LIKEREC, the possibility of error due to the 
structure definitions getting out of synch should outweigh the possible 
tiny gain to be had by moving all the subfields at once.

Rules to live by:
- Correctness beats performance.
- Explicit correctness beats happenstance correctness.