Path: utzoo!utgpu!water!watmath!clyde!rutgers!ukma!psuvm.bitnet!uh2
From: UH2@PSUVM.BITNET (Lee Sailer)
Newsgroups: comp.software-eng
Subject: Re: program complexity metrics
Message-ID: <28001UH2@PSUVM>
Date: 18 Dec 87 14:26:39 GMT
References: <561@ndsuvax.UUCP> <3850002@wdl1.UUCP> <242@pedsga.UUCP> <2252@dasys1.UUCP>
Organization: The Pennsylvania State University - Computation Center
Lines: 22

This is an interesting topic.  How would YOU "measure" the complexity
of a module?  How can we "validate" the measures?
     
The <number of paths> thru a module doesn't impress me at all.  A very simple
module constructed from a sequence of big CASE or SWITCH statements will have
many paths, but is easy to understand.
     
How 'bout a measure based on the *hierarchy* of Data Flow Diagrams?  For
example, some modules might have a simple top level, with increasing
complexity as you break down each level.  Another module might have a
complex top level, but the lower levels might be simple.  Which is more
complex overall, the first or the second?
     
My *guess* is that the first is simpler, overall, on the principle that
it hides the complexity.  Yet, it probably has more paths, more tokens, more
ifs, ands, and whiles (fewer gotos?).
     
The main point is that the COMPLEXITY of a module depends to some extent
on how detailed an understanding of the module you want or need.
     
                                                                lee