On Thu, 21 Jun 2018 10:45:01 +1000, Andrew Rowley wrote:
>
>Unfortunately [ ... ] the symbols in SYSIN are not
>substituted the same way as in the regular JCL. I have been playing
>around with them and my conclusion is that the whole feature seems to
>have been badly thought out.
>
Yes.
>- Symbols in in stream data are not necessarily substituted with the
>same value as the same symbol in the JCL of the same step
>
o First, PROCs are expanded by the converter at the point of EXEC PRC=...
o Then, symbols in JCL statements are resolved to the last value assigned
before the point of reference.
o Symbols in SYSIN are resolved to the last value assigned before the next
EXEC PGM... statement. Those steeped in MVS internals ("Kool-Aid") say
this makes perfect sense. To normal people it's utter nonsense.
>- Symbols "leak" from step to step and from PROCs into the main JCL, so
>if you modify or add a symbol in a PROC you can break subsequent steps
>in any job that calls that PROC.
>
I believe that symbols assigned values in a PROC statement or in EXEC PRC=
are local to that PROC (and its children?) Symbols assigned values in a SET
statement within a PROC are global. Ugh! It may be better to eschew SET
within a PROC; rather supply default values in the PROC header.
>If IBM had omitted some features, e.g. system symbols on the execution
>system and instead substituted the symbols at the same time as the
>symbols in the rest of the JCL, they would have lost maybe 10% of the
>usefulness but decreased the astonishment by 90%.
>
Be careful. If the SYSIN appears within a PROC and different symbol values
are in effect at various EXEC PRC=... statements, multiple copies of that
SYSIN might need to be supplied for the various proc steps. But the effect
could be realized by passing along with each reference to that SYSIN a list
of name-value pairs of symbols used in that SYSIN, with the values at the
point //SYSIN DD * appears in the JCL source.
>The good news is that the syntax probably allows an easy fix in a new
>version: a new value such as SYMBOLS=(JCLCNVT) where symbols are
>substituted at the same time with the same logic as the rest of the JCL.
>
>Please IBM?
>
They'd only need to do it right.
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
