NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/comp/lang/perl/5476

Internet Message Format | 1992-08-25 | 3.1 KB

Path: sparky!uunet!zaphod.mps.ohio-state.edu!usc!news!netlabs!lwall
From: [e-mail] (Larry Wall)
Newsgroups: comp.lang.perl
Subject: Re: Using Gotos?
Message-ID: <[e-mail]>
Date: 25 Aug 92 20:40:23 GMT
References: <[e-mail]>
Sender: [e-mail]
Distribution: comp
Organization: NetLabs, Inc.
Lines: 55
Nntp-Posting-Host: scalpel.netlabs.com

In article <[e-mail]> [e-mail] (David Wuertele) writes:
: P.S. I have been equating "next LABEL;" with GOTO, perhaps mistakenly.
: Please flame if I'm being naive.

OK, prepare to be crispy crittered.  :-)

The generic "goto" has been rightly condemned as obfuscatory.  You really
have no idea where the label is, or whether you're jumping into the middle
of some other control flow that would be better off without your presence.
It's trivial using goto to write a control flow that is too complicated
to think about except when high on Jolt Cola (and maybe not even then).

By contrast, certain customary uses of goto (and goto-like constructs)
are easy to understand because they don't violate the shape of the
control flow, but merely short-circuit it.

In the early days of structured programming, one stated goal was that
every block should have only one entry and one exit.  This is half
right.  Or perhaps twice right.  Every block *should* have only one
entry.  But the one-exit idea is bogus, because most logical branch
structures are in fact tree structures, and not linear.  A block needs
to categorize things, and once it has categorized things, it's more or
less done.  It shouldn't be forced to uncategorize things again.

There is implicit recognition of this in a language like C.  Notice
that the "return" statement can be used anywhere in a subroutine.
There is also the notion of "returning" from the current iteration of a
loop (via "continue") to start the next iteration, and "returning" from
the loop as a whole (via "break") for some reason that wouldn't have
been apparent at the top of the loop.  In real life, many control
structures are tree-shaped, so many loops have multiple (valid!) exits
or short-circuit conditions.

Perl just takes this notion one step further than C, by allowing you to
exit or short-curcuit a named loop.  Note however that, unlike with
goto, you KNOW that you're passing control out to an enclosing
construct.  That fact, together with the fact that Perl encourages you
to use meaningful loop labels, lets you write near-English statements
such as:

    next LINE if /^$/;        # skip blank lines
    next LINE if /^#/;        # skip comment lines

    last FILE if /^end$/;    # reached logical end of file

This is much better even than some proposed C extensions that would let
you specify how many levels you're breaking (as in "break 2").  The
Perl construct maps better onto your mental constuct of what you're
iterating over.  You don't have to count enclosing blocks, or even find
the block label.  You already know it, presuming you named it sensibly.

So there's a good reason you can't rewrite the "findword" example easily:
I'm such a lazy critter that I already did it the easy way!  :-)

Larry