|
|
 | | From: | Bob Bridges | | Subject: | Efficiency or lack thereof | | Date: | 18 Jan 2005 11:40:33 -0800 |
|
|
 | Robin, this probably isn't what you had in mind but I'll add this one
point: I was hired on this contract to automate a lot of random things
and also a couple specific processes. The latter are pretty long and are
in production so it'll take a lot of planning, but I just had occasion to
replace five or six production steps with a single SORT and a REXX. The
original five steps went something like this:
a) Check to be sure a certain FTP arrived as promised. If it did,
complete the following steps:
b) QUIKJOB to convert all lower-case letters to upper case.
c) SORT to INCLUDE a portion of the records from step b), sort on SSN and
write out a new format to file C.
d) SORT to OMIT some records from step b), sort on SSN and write out a new
format to file D.
e) QUICKJOB to convert a location code in file d) to a description.
I think that's it. You can see that the original author likes SORT, and
with good reason. But the replacement consisted of a single SORT on SSN
and a REXX that did all the other stuff in a single pass.
I like the simplicity, but a SORT-loving colleague checked the stats on
the new JCL and believes my two steps took about a CPU second longer to
run than the five original steps. I like interpreted languages, I really
do, and REXX is better than many for speed as well as being truly
excellent at string processing. But I infer it can't stand up to SORT for
efficiency.
---
Bob Bridges, robertbridges@discoverfinancial.com, 224 405-0811
rhbridg@attglobal.net, 847 520-1684 xt 243
/* Science is a differential equation. Religion is a boundary condition.
-Alan Turing, quoted in J D Barrow's _Theories of Everything_ */
"Ryerse, Robin"
2005-01-18 13:14
To: TSO-REXX@VM.MARIST.EDU
cc:
Subject: Re: massive boilerplate inclusions (was : functions returning bo olean
.... or failing ...)
The topic of this thread lead me to research some stats about the execs I
have written. Currently, I have exactlty 100 execs that are available for
general use within the lpar. The largest exec is 724 lines of code. The
total lines of code for the 100 execs is 8443; makes the average around 85
lines of code. Obviously no "massive boilerplate"s. I should point out
that
the stats do not include the code for the function package which
exclusively
Assembler.
I know there are those who prefer using the boiler plate strategy to
ensure
completeness of error checking. But I do not subscribe to that approach. I
get no complaints about completeness or accuracy of the delivered
functionality.
As a discuusion point, I think the alternative to "boilerplate" should be
considered.
----------------------------------------------------------------------
For TSO-REXX subscribe / signoff / archive access instructions,
send email to LISTSERV@VM.MARIST.EDU with the message: INFO TSO-REXX
|
|
| | From: | Robert Lawrence | | Subject: | Re: Efficiency or lack thereof | | Date: | 18 Jan 2005 12:02:42 -0800 |
|
|
| Bob,
Since you are using sort have you looked at the options available for converting from
lower to upper case on the INREC control statement FIELDS=(p,m,TRAN=LTOU) and other
variations and also for OUTFIL processing for splitting out to different files and code
translations. What you are trying may be all attainable by DFSORT especially if you are
current. I wish we were.
HTH
Bob Lawrence
DBA
Boscov's Dept Stores LLc
> -----Original Message-----
> From: TSO REXX Discussion List [mailto:TSO-REXX@VM.MARIST.EDU]On Behalf
> Of Bob Bridges
> Sent: Tuesday, January 18, 2005 2:36 PM
> To: TSO-REXX@VM.MARIST.EDU
> Subject: Efficiency or lack thereof
>
>
> Robin, this probably isn't what you had in mind but I'll add this one
> point: I was hired on this contract to automate a lot of random things
> and also a couple specific processes. The latter are pretty long and are
> in production so it'll take a lot of planning, but I just had occasion to
> replace five or six production steps with a single SORT and a REXX. The
> original five steps went something like this:
>
> a) Check to be sure a certain FTP arrived as promised. If it did,
> complete the following steps:
> b) QUIKJOB to convert all lower-case letters to upper case.
> c) SORT to INCLUDE a portion of the records from step b), sort on SSN and
> write out a new format to file C.
> d) SORT to OMIT some records from step b), sort on SSN and write out a new
> format to file D.
> e) QUICKJOB to convert a location code in file d) to a description.
>
> I think that's it. You can see that the original author likes SORT, and
> with good reason. But the replacement consisted of a single SORT on SSN
> and a REXX that did all the other stuff in a single pass.
>
> I like the simplicity, but a SORT-loving colleague checked the stats on
> the new JCL and believes my two steps took about a CPU second longer to
> run than the five original steps. I like interpreted languages, I really
> do, and REXX is better than many for speed as well as being truly
> excellent at string processing. But I infer it can't stand up to SORT for
> efficiency.
>
> ---
> Bob Bridges, robertbridges@discoverfinancial.com, 224 405-0811
> rhbridg@attglobal.net, 847 520-1684 xt 243
>
> /* Science is a differential equation. Religion is a boundary condition.
> -Alan Turing, quoted in J D Barrow's _Theories of Everything_ */
>
>
>
>
> "Ryerse, Robin"
> 2005-01-18 13:14
>
>
> To: TSO-REXX@VM.MARIST.EDU
> cc:
> Subject: Re: massive boilerplate inclusions (was : functions
> returning bo olean
> ... or failing ...)
>
> The topic of this thread lead me to research some stats about the execs I
> have written. Currently, I have exactlty 100 execs that are available for
> general use within the lpar. The largest exec is 724 lines of code. The
> total lines of code for the 100 execs is 8443; makes the average around 85
> lines of code. Obviously no "massive boilerplate"s. I should point out
> that
> the stats do not include the code for the function package which
> exclusively
> Assembler.
>
> I know there are those who prefer using the boiler plate strategy to
> ensure
> completeness of error checking. But I do not subscribe to that approach. I
> get no complaints about completeness or accuracy of the delivered
> functionality.
>
> As a discuusion point, I think the alternative to "boilerplate" should be
> considered.
>
> ----------------------------------------------------------------------
> For TSO-REXX subscribe / signoff / archive access instructions,
> send email to LISTSERV@VM.MARIST.EDU with the message: INFO TSO-REXX
>
----------------------------------------------------------------------
For TSO-REXX subscribe / signoff / archive access instructions,
send email to LISTSERV@VM.MARIST.EDU with the message: INFO TSO-REXX
|
|
|
