Menu
▾
▴
Re: [netrexx-pipelines] Migrate OS/2 pipelines to NetRexx Pipes
|
From: Jeff H. <Je...@Je...> - 2022-06-23 03:09:42
|
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<p>Mike,</p>
<p>Thank you for your great comments and suggestions. I too, a
long time ago, used and liked CMS Pipelines (although not loving
the notation as much as Rexx's) and when the possibility arose for
PC use, I decided I'd add what I could. I have not had CMS access
for 30 years now, and never had any of IBM's sources. Ours is all
on the basis of the published specifications. Usually they are
readable. Occasionally they leave holes open to interpretation,
and access to a CMS machine for the definitive answer would be
useful.</p>
<p>Please read the NetRexx Pipelines documentation on the
differences from CMS Pipelines. There are a couple deep,
fundamental, unavoidable differences. There are also builtin
stages / options in each that will never be appropriate in the
other. (And, I like to think, some new NetRexx things that would
be useful if implemented in CMS!) <br>
</p>
<p>All of NetRexx Pipelines, and its stages, is written in NetRexx.</p>
<p>In the case of SPECS and adding WS, there is a section of code
right now that handles space delimited words, in a (possibly
nested) substring. It is surrounding that where I am doing the
substitution of SPACEs to a string of characters not in the record
(first trying \x00, then \x00 01, \x00 01 02, etc) for SPACEs,
then the WSs to SPACEs; and undoing these after the selection.
(This also allows a "new extra" in that the WS or FS can now be a
multi-character string, rather than be restricted to a single
character. I see a real use for this in FSs, for example with
PC's end-of-line CRLF pair.) For fields, it is doing the same,
but adding a character after the FS's new SPACE so there are no
empty fields, as multiple WSs are merged for WORDS, but not FSs.
<br>
</p>
<p>Like several other complex stages, the code here processes all
the specified options and on the fly creates a new special purpose
stage, in NetRexx, and substitutes it into the pipeline. A little
expensive in setup, but much faster for each record. Your time
tests on short input files reflects this setup cost. This also
makes debugging these stages a real (unsexed) bitch.<br>
</p>
<p>I think that that logic is solid, but at the moment am having
problems getting even the original specs.nrx to compile and run.
It has been a couple of years since I've done this stuff.</p>
<p>Again, thank you for all.<br>
</p>
<p>Jeff<br>
</p>
<div class="moz-cite-prefix">On 6/22/2022 5:33 PM, <a class="moz-txt-link-abbreviated" href="mailto:hp...@we...">hp...@we...</a>
wrote:<br>
</div>
<blockquote type="cite"
cite="mid:f77...@we...">Hi Jeff!
<br>
<br>
Am 22.06.2022 um 14:45 schrieb Jeff Hennick:
<br>
<blockquote type="cite">Mike, I am using, and expanding, your
idea/suggestion on changing the spaces to
<br>
something else that is not already in the record, etc. Thank
you!
<br>
</blockquote>
<br>
You're welcome.
<br>
But, before you go for it, may be you should make sure it will not
<br>
lead to a dead end. Regarding nested 'substring ... of substring
<br>
... of substring ... of ...', each optionally with its own fs, I'm
<br>
not sure if this could be done by masking blanks iteratively.
<br>
I don't know how you program stages, in assembler or C or REXX? In
<br>
the last case you could use "mighty parse" very well.
<br>
Do you have the chance to take a look at John's sources? Highly
<br>
likely not. Maybe Rob van der Heij could give you a hint.
<br>
<br>
Best,
<br>
M.
<br>
<br>
<br>
_______________________________________________
<br>
netrexx-pipelines mailing list
<br>
<a class="moz-txt-link-abbreviated" href="mailto:net...@li...">net...@li...</a>
<br>
<a class="moz-txt-link-freetext" href="https://lists.sourceforge.net/lists/listinfo/netrexx-pipelines">https://lists.sourceforge.net/lists/listinfo/netrexx-pipelines</a>
<br>
</blockquote>
</body>
</html>
|
