Re: How do I export Query Manager and Query/400 "files" to my -- MIDRANGE-L

Dave Odom wrote:

And yes, REXX is very powerful and common in IBM.  Makes me wonder why
people use CL.


Dave:

After you, you might not find a bigger fan of REXX/400 on this listthan me; but I'll regularly use CL, _especially_ ILE CL, instead ofREXX for the vast majority of functions when the choice is betweenthe two. (And if it involves _any_ product code that goes to acustomer, there won't be a hint of REXX in it.)


So, I'll step in to discuss from the other side.

I'm not aware of anything that can't be done in CL without requiringREXX, and I can think of a lot of things that can't be done in REXXwithout requiring CL. I have learned that Qshell utilities can beused in place of CL for most of those things, via [ address"QSHELL/QZSHQSHC" ], which is pretty cool, BTW. Still, a lot of theIFS is out of reach for REXX itself, a major irritation to me. Ortry simply changing the owner of an object in REXX without using CL,e.g., CHGOBJOWN or CHGOWN CL commands.

For many management functions, REXX does little more than provide aframework for executing CL, one interpreted command at a time. Allof the REXX tends then simply to obscure the real CL that does the work.

Another element, REXX _must_ have a source member in order to run.CL only needs the member at creation time. Once created, the sourcecan be kept separately. If necessary, CL can even generate a sourcefile and member at run-time and populate it with REXX statementsthat can then be passed to the QREXX API. Try doing somethingsimilar in REXX; if there is no source file and member at run-time,REXX won't get very far.

Rather than having a series of REXX statements that are interpretedand that themselves create strings that are then interpreted by thecommand environment and executed, a CL program has the stringsinterpreted by the compiler at compile-time. The resulting programruns at compiled speed without needing to wait for the REXXenvironment to be created. The fundamental parsing and syntaxchecking is long out of the way.

ILE CL goes far beyond OPM CL. It provides bindable procedures thatREXX can't touch at all.

With REXX needing to execute CL commands so often, a REXX programmerneeds to be familiar with both REXX and CL. Yet, a CL programmergets by without REXX easily. I'd guess that the majority of CLprogrammers _never_ need REXX. And I'd guess that the majority ofREXX/400 programmers _regularly_ need CL. (A REXX/400 programmer whonever uses CL is probably doing a very large amount of unnecessaryprogramming.)

To REXX's credit, many CL and other programmers who never use REXXare also doing extra work. REXX provides some superb shortcuts.That's why I use it. I have a few procs that I rely on over and overagain and wouldn't want to code in another language. I even have aREXX/400 version of the old ELIZA 'psychotherapist' proc that Iconverted years ago from some early form of PC BASIC. (I never didmuch beyond making it work and never got so far as making it workwell.) As simple as my ELIZA version is, it's clearly better suitedfor REXX than CL.

But IBM seems to treat REXX/400 with less respect than is shown bythose who use it. IBM came up with Object REXX and recently releasedit to the open source community. AFAICT, there was never any seriousthought to bringing it to OS/400 even though OS/400 is object-based.Nor has there been a hint of _anything_ else advancing in REXX/400since... hmmm... since...?

Well, who wants to get held back by some old "legacy" language thatcan't even invoke a procedure? If IBM ignores it, why care?


Tom Liotta