> One of these days, I will looking into write a > program for IRXJCL which allows "in line" REXX code.
Have not fully thought this through but couldn't you do it as a PROC that copied inline Rexx into a temporary PDS and then ran it with IRXJCL? Kind of an unfortunate restriction on IRXJCL that it requires a PDS and a member name. Presumably would not have been terribly difficult for it to have supported a sequential dataset. Charles -----Original Message----- From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of John McKown Sent: Wednesday, January 06, 2016 11:26 AM To: [email protected] Subject: Re: Sort for not there? On Wed, Jan 6, 2016 at 1:04 PM, Paul Gilmartin < [email protected]> wrote: > (hyperspaced from ASSEMBLER-LIST) > > On 2016-01-06, at 11:45, John McKown wrote: > >> > >> //DD1 DD *,DLM=AA > > > > True. I've used $$ quite a bit myself. But in the _general_ case, > > how > can > > you guarantee that whatever characters you select will _not_ occur > > as the first two characters in the "sample" data? > > > Well, the Pigeonhole Principle guarantees that if the data don't > exceed > 65,535 lines a suitable value must exist. But how to find it? > > I might take this to IBM-MAIN; someone is apt to jump in with a > DFSORT/ICETOOL solution. > This _is_ IBM-MAIN. > > Hmmm. Count occurences of each initial digraph and select any zero value. > But are zeroes counted? > > Why isn't DLM allowed to be longer? If it were a few dozen characters, > there's be a guaranteed value for any plausible size data set. > > I suppose one could concatenate SYSINs with a different DLM for each, until > the concatenation limit was encountered. > > I hate JCL! It ain't 1965 no more; we shouldn't be burdened with resource > constraints designed for 1965. > OK. But I can just imagine the flames and wars (and the demands for a business case from IBM) for "something else". Oh, and be damn sure that the "something else" can be used with CA-11 restart management. So the simplest thing would be a new control language which can be "compiled" into "internal text" (like JCL is at present). But that would still have some of the problems that JCL has. E.g. no looping construct. There is no way to construct SWA to do "looping". Of course, you could abandon most of JCL by using Co:Z Batch Launcher to write a UNIX shell script. One of these days, I will looking into write a program for IRXJCL which allows "in line" REXX code. E.g. //STEP1 EXEC PGM=RUNREXX,REGION=0M //SYSPRINT DD SYSOUT=* //SYSERROR DD SYSOUT=* //SYSPROC DD DISP=SHR,DSN=SYS1.SBPXEXEC //SYSIN DD DLM='$$' /* REXX */ SAY "HELLO, WORLD!" CALL BPXWDYN "ALLOC DDN(X) DSN(X) OLD" CALL 'some.linklib(somepgm)' 'some parm' $$ // ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
