🚀 go-pugleaf

From the keyboard of Josef Moellers <mollers.pad@sni.de>:
:In <AeRB8p600WB60JlPZc@andrew.cmu.edu> es2j+@andrew.cmu.edu (Edward John Sabol) writes:
:>Could someone post a awk, perl, or sed script which can center text. I'd
:>like to be able to pipe a file or region into it and get out the same
:>text only centered in 80 columns.
:
:>It shouldn't be hard to do, but I don't know awk or perl very well. I
:>can do it in C, of course, but I want a script for ease of portability
:>and the smaller size. Speed needn't be an issue.
:
:The following is a bourne shell script that does it, followed by a Korn
:shell script (caveat: both can't handle backslashes!):
		       ^^^^^^^^^^

In my techojargon brand of English, "both can't" is not the same
as "neither can".  This is "both can't":

    if (!a || !b)

Meaning it's true if EITHER A OR B OR BOTH CANNOT.

whereas this is "neither can":

    if (!(a && b))

Meaning is true only if neither A nor B can do something.

I realize that the poster is not a native speaker of English.
Perhaps these are equivalent in German.  Perhaps I'm no longer
a native speaker of English. :-)

:-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-sh-
:#!/bin/sh
:while read line
:do len=`expr "$line" : '.*'`
:   prefix=`expr \( 80 - $len \) / 2`
:   while [ $prefix -ge 8 ]
:   do echo '\t\c'
:      prefix=`expr $prefix - 8`
:   done
:   while [ $prefix -ge 0 ]
:   do echo ' \c'
:      prefix=`expr $prefix - 1`
:   done
:   echo $line
:done

[ksh solution omitted]

I think we have another candidate for the annual Dead Parrot Society award;
you know, those lethargic programs that remind you of the Monty Python skit.
Yes, folks, the Efficiency Cops are back!

Consider, if you would, the following file:

    $ wc /etc/motd
      74     404    4119 /etc/motd

First off, here are some timings:

A modestly sized file, I might add.

    $ time sh ctr.sh </etc/motd > /dev/null
    non-numeric argument
    test: argument expected
    test: argument expected
    16.377u 76.475s 0:55.71 50.0% 0+0k 0+19io 12275pf+0w

Well, maybe we're from the Correctness Cops, too.    Let's look at
the first line of /etc/motd:

    $ sed 1q /etc/motd
    * Fri Aug 16 19:13:53 CDT 1991:

Let's just run the program on that line:

    $ sed 1q /etc/motd | sh ctr.sh
    \t\c
    \t\c
    \t\c
     \c
    ftpget ftplib.pl ftpls getdate n0 n1 n2 n3 n4 n5 n6 n7 n8 n9 statmon tree Fri Aug 16 19:13:53 CDT 1991:

Oh my.  That's wrong twice.  First of all, you can't rely on echo 
doing backslash evalutation.  (From my perpective, the SysV crew
screwed up for trying to make it such.  They should have invented
another command, like print, to do that.)  Second of all, some place
in there you're the shell take a swipe and globbing your raw data.

While both these problems can be solved by enough quoting and mucking with
the echo args enough, you still have the problem of it running much as a
dead parrot flies.  Actually, I took a shot at fixing it, and kept
getting whinage like this,

    non-numeric argument
    test: argument expected

which just goes to show you that it's easier to port a shell 
than a shell script. :-(

On the other hand, this solution

    perl -pe 'print " " x int((80 - length())/2)' 

not rely on echo behavior, isn't a process pig, and doesn't
run afoul of my /bin/test command.  It also runs a tad faster:

    $ time perl -pe 'print " " x int((80 - length())/2)' </etc/motd >/dev/null
    0.030u 0.052s 0:00.05 50.0% 0+0k 0+5io 48pf+0w

where we'll define a tad as some 1,132 times faster on this dataset.  
It also inflicts some 256 fewer page faults on the flogged system.

'Nuff said.

Please direct followups appropriately:

    TOPIC                       NEWSGROUP

    shell programming           comp.unix.shell
				    (and maybe alt.sources.d)
    "both can't"                alt.usage.english      
				    (and maybe alt.folklore.computers)
    the evils of perl           alt.religion.computers
				    (and maybe alt.flame)
    

--tom
--