Re: cat and varying

[Hans Boldt <boldt@xxxxxxxxxx>]

Steve Richter wrote:
> what's the story, does the cat opcode have a problem with varying strings?
>
> d sMsg            s             80a   varying
> d sMsg2           s            256a   varying
> c                   eval      sMsg = 'abc'
> c     sMsg          cat       'def':1       sMsg2
>
> the resulting value in sMsg2 is blanks.  I want it to be 'abc def'.
>
> I was hoping that the CAT opcode would work best when appending to a string:
> d sWord           s             80a   varying
> d sSentence       s           2000a   varying
>
> ** using CAT
> c                   cat       sWord:1       sSentence
>
> ** is more efficient than using "+" ?
> c                   eval      sSentence = sSentence + ' ' + sWord
>
> will the v5r2 += operator be more efficient than + ?
>    sSentence += sWord ;
>    sSentence = sSentence + sWord ;
If you want to work with varying length character variables, you
should really be using expressions, and not the old-style fixed form
calcs. I discussed the reason in another thread. Basically, because
of the weird semantics of character operands in fixed form calcs, we
defined the functionality such that a varying length character
variable would work exactly the same as a fixed length variable
defined with the same length as the current length of the varying
variable. Defining the functionality this way avoided the need for
complex rules with lots of special cases. (By default, when you code
a fixed length character variable as a result of a fixed-form calc,
the left-most or right-most characters might not be changed
depending on the lengths of the operands.)

In your example, variable SMSG2 starts out with a value of ''.
Coding it as the result of the CAT operation is equivalent to coding
a fixed length character variable defined with a length of zero.

And so, yes, the + operator is preferred over CAT when dealing with
varying length character. Not because it's necessarily faster, but
rather, it works as you would expect with varying length character
variables.

Is += more efficient? It could be if the evaluation of the target
involves a compute-intensive operation. For example, let's say you
code "ARR(PROC())+=STR;", and PROC() takes some time. Clearly, that
would be faster than coding "ARR(PROC())=ARR(PROC())+STR;", but only
because the procedure is called one less time. Otherwise, "A+=B;" is
exactly equivalent to "A=A+B;".

(Oh yeah, note that "A+=B;" is not allowed where A is a fixed length
character variable. If it were allowed, it would effectively be a
NO-OP anyways.)

Cheers!  Hans