SCOUG-Programming Mailing List Archives

<< Previous Message <<

Date:	Sun, 16 Feb 2003 22:42:00 PST8
From:	Benedict G. Archer <bgarcher@gte.net >
Reply-To:	scoug-programming@scoug.com
To:	scoug-programming@scoug.com
Subject:	SCOUG-Programming: In retrospect

In , on 02/16/2003
at 07:52 PM, "Steven Levine" said:

>In <0HAF0094ZJU7FW@mta6.snfc21.pbi.net>, on 02/16/03
> at 05:50 PM, "Lynn H. Maxson" said:

>>The purpose of test data lies in testing all possible paths in a program
>>with data that insures it does what it is supposed to do and other data
>>to insure that it does not what it is not supposed to do. Probably as

>The test data is only as good as the understanding of the process to be
>tested.

Much of this interesting exchange is beyond what I know but I can relate
to this point. One role I've been in is algorithm development which I
regard as mostly distinct from programming. The general problem is
reducing a long time series of n-point fluorescence spectra to a DNA
sequence or genotype--another chemical data problem. The possible
variations in the raw output are so great that a regression set comprising
several hundred cases isn't complete--new types of failures keep showing
up. All the algorithm development is done in an interpreted environment,
matlab (sometimes octave, with some prototyping in the past in APL). The
development process has evolved so that the entire data processing task is
implemented in matlab with the released code a near literal translation of
the matlab into C++. The programmers writing the actual code don't always
understand the algorithms, but test against matlab results. Even DSP code
is developed this way. When processing failures show up in released code,
the matlab code is fixed first, then the tested fixes applied to the C++.
This double implementation may seem inefficient, maybe it is? It has the
merit, however, (some think weakness) that the algorithm developer needn't
be a skilled programmer and the programmer needn't know all the
mathematics. On the other hand someone with both capabilities will
probably be more productive. This issue is relevant only for apps
implementing math not in the typical programmer's repertoire.

>Steven

Don't know when I'll get to it, but still intend to attempt building the
latest octave sources (GNU GPL) for OS/2, a bigger app than HPcalc, but no
PM GUI as it runs from a command line window.

Ben A

--
-----------------------------------------------------------
Benedict G. Archer
-----------------------------------------------------------

=====================================================

To unsubscribe from this list, send an email message
to "steward@scoug.com". In the body of the message,
put the command "unsubscribe scoug-programming".

For problems, contact the list owner at
"rollin@scoug.com".

=====================================================

Return to [ 16 | February | 2003 ]

SCOUG, Warp Expo West, and Warpfest are trademarks of the Southern California OS/2 User Group. OS/2, Workplace Shell, and IBM are registered trademarks of International Business Machines Corporation. All other trademarks remain the property of their respective owners.