Re: pkg_alternatives, Darwin and locale

[Christian Biere <christianbiere%gmx.de@localhost>]

Joerg Sonnenberger wrote:
> On Sun, Jul 20, 2008 at 04:26:47AM +0200, Christian Biere wrote:
> > +cat "${userfile}" __CONF_FILE__ __DB_FILE__ 2>/dev/null | \
> > +{
> > +    while read line
> > +    do
> > +        line="${line##\#*}"
> > +        prog="${line%%\ *}" 
> > +        if [ X != "X${prog}" ] && [ -x "${prog}" ]
> > +        then exec ${line} "${@}"
> > +        fi
> > +    done
> 
> Can you make this awk please? read is horrible slow for larger files...

Do these files list more than some dozen lines at most?

While I do know that 'read' is rather slow due to single byte reads,
NetBSD's /bin/sh is exceptionally S-L-O-W and for some reason also burns
a lot of more system time than other shells at this. ksh takes only a
third of the time.

I tried the following with a file of about 400k lines, actually the
output of "find /", redirecting standard out to /dev/null:

while read line
do
        echo "$line"
done

I tested this with 3 shells repeatedly and got these results:
sh: 30 seconds (20 seconds system time)
bash: 18 seconds (2 seconds system time)
ksh: 9 seconds (less than 2 seconds system time)

Replacing "echo $line" with ":" gains only 3 seconds for sh and doesn't
change the rankings.

-- 
Christian