Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rochester!ken
From: ken@rochester.ARPA (Ken Yap)
Newsgroups: comp.sources.d
Subject: Re: shell printf - a sensible solution
Message-ID: <26094@rochester.ARPA>
Date: Wed, 18-Mar-87 20:43:21 EST
Article-I.D.: rocheste.26094
Posted: Wed Mar 18 20:43:21 1987
Date-Received: Fri, 20-Mar-87 03:17:52 EST
References: <5008@brl-adm.ARPA> <210@quacky.mips.UUCP>
Reply-To: ken@rochester.UUCP (Ken Yap)
Organization: U of Rochester, CS Dept, Rochester, NY
Lines: 22
Summary: not for a trivial program, it ain't

|With machine speed increasing all of the time, this argument is
|becoming obsolete. Until we all have 20% of a 5 mips machine, it's

Jon Bentley, in one of the programming pearls columns asks the reader
to consider how much CPU time is saved compared to programmer time.
I would guess the posted printf took more time to write than all
the CPU time it will save throughout its entire lifetime.

|On the other hand, there is another type of waste involved, and there
|is an issue of portability.

Which is why I won't use getopt in shell scripts even though it is
tempting - I can't be sure it exists on every machine the shell script
will be run on.

Yes, there is a tradeoff, but the need has to be well documented before
turning a shell script into C code. I think it was Rob Pike who noted
that "ls | mc" where mc is a multi-column filter actually runs faster
than the Berkeley "ls". I hate it everytime I do a "rsh ls" and forget
to append -C!

	Ken