From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.5-pre1 (2020-06-20) on
	ip-172-31-74-118.ec2.internal
X-Spam-Level: *
X-Spam-Status: No, score=1.8 required=3.0 tests=BAYES_50,FROM_ADDR_WS
	autolearn=no autolearn_force=no version=3.4.5-pre1
Date: 17 Jul 93 21:44:10 GMT
From: att!csn!news.den.mmc.com!possum.den.mmc.com!hagar@ucbvax.Berkeley.EDU  (J
on Duncan Hagar)
Subject: Re: McCabe's and other Measures
Message-ID: <1993Jul17.214410.8538@den.mmc.com>
List-Id: <comp.lang.ada>

Re Articles on McCabe's Cyclomatic Complexity and measures
<rest deleted>
>In all honesty, McCabe's Cyclomatic Complexity does have some problems, since
>it does not (in original form) address data sensitizing and its level of
>coverage has been argued.  I would be very interested in other forms that
>people are using to measure complexity.  We have used both McCabe,
>LOC (5 kinds) and Halstead in our Complexity Measures Tool (CMT) and have
>found McCabe to be the second most widely used (Lines of Code as
>always is still the primary measure).  We are looking into other metrics now
>and would appreciate any comments other people have had.  It would be nice to
>bring some technical discussions to comp.lang.ada that would be worthy of
>crossposting :-)
>
>Brandon Goldfedder
>Director of Training and Consulting
>EVB Software Engineering, Inc.

In general measurement of software seem a tricky issue.  Measurement is 
certainly topical, but what do they really tell us.  Some say measurement 
should predict or indicate something (like will the thing be measured work).  
McCabeUs does have some predication power by identifying segments of 
code that are maybe more complex, and therefore maybe more prone to error 
(because humans make more mistakes as complexity increases).  But this is 
not an absolute like the way an "out of tolerance" measurement on a piece of 
hardware is (the bolt will not fit if the hole is to small).  A McCabe's 
Cyclomatic Complexity factor over the "magic" number (10 on some 
projects) can be okay, e.g. if you are dealing with a CASE statement in Ada.  
So, in my mind, current software measures are mostly indicators only and lack 
absolute prediction power.  We should use them, but understand what they 
are, what their limits are, and when to blow them off.  My fear would be 
management by numbers that blindly treats them like hardware measures (out 
of tolerance => trash bin).  The articles so far have hit the standard measures
 
(I canUt name the others right now, I'll let that for some other time, if this 
thread continues or the measurement researchers out there can fill this in).

-- 
Jon Hagar,  Mail Stop-H0512                  hagar@den.mmc.com
Martin Marietta Astronautics                (303) 977-1625
P.O. Box 179,   Denver, CO  80201       SD: Opinions are only mine
In the electronic village no matter where you go there you virtually are