Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!ulysses!dgk
From: dgk@ulysses.homer.nj.att.com (David Korn[drew])
Newsgroups: comp.unix.questions
Subject: Re: Some csh how-to, please
Summary: while read; do ...; done performance
Keywords: csh C-shell shell programming unix read
Message-ID: <11407@ulysses.homer.nj.att.com>
Date: 5 Apr 89 02:03:27 GMT
References: <2127@pikes.Colorado.EDU> <7467@thorin.cs.unc.edu> <718@salgado.stan.UUCP>
Organization: AT&T Bell Laboratories, Murray Hill
Lines: 39

In article <718@salgado.stan.UUCP>, dce@stan.UUCP (David Elliott) writes:
> In article <4258@omepd.UUCP> merlyn@intelob.intel.com (Randal L. Schwartz @ Stonehenge) writes:
> >Why do people use 'cat' at every opportunity?  Maybe that's why
> >they sell so many multiprocessor machines? :-) :-)
> 
> >Simpler:
> >
> >	while read fie
> >	do
> >		something with $fie
> >	done <file
> >
> >no additional processes.
> 
> 
> In practice, you are unfortunately wrong.  The best that redirection
> gives you is one less exec.  Current "standard" versions of sh (in
> other words, most sh's other than ksh) fork when redirection is done
> on a loop.
> 

The big problem with
	cat file | while read ...

is that it runs about 1/10th the speed of the redirection loop.
The reason is that the shell (sh and/or ksh) have to read one
character at a time when reading from a pipe.  The reason for
reading one character at a time is to make scripts like
	cat file | ( read line;cat)
behead only a single line.

Unfortunately there is not way to seek back on a pipe.  Nor is
there an ioctl() to specify line buffering.  Thus, single characters
reads are used.  This is not the case when you specify redirection.

I hope that someone will add a stream discipline that will cause
pipes to use a line buffering protocol.  Until such time, scripts
would do better to use a temporary file rather than piping to
a while read loop.