It's not a mistake to ban goto... it's not a mistake to keep it... [LWN.net]

It's not a mistake to ban goto... it's not a mistake to keep it...

Posted Feb 10, 2008 7:53 UTC (Sun) by khim (subscriber, #9252)
In reply to: More than one way to do it... by im14u2c
Parent article: Interview: Mark "Markey" Kretschmann (Not the Gentoo Weekly News)

If there are two (or more) ways to do something (like for and while cycles in your samples) but they are used often in real programs - it's not a big deal: reader will remember both and it's easy to understood both. The constructs usable once per a blue moon must be rejected unless they are doing something which will be impossible or very-very hard to do otherwise.

Now about goto: while it's not very useful it's still quite simple to understand because we have "goto in disguise": return. But of course plain old goto is so rare that ban makes sense. But any choice must be justified from reader's position, not from writer's one. This is not how perl chooses the constructs...

(Log in to post comments)

It's not a mistake to ban goto... it's not a mistake to keep it...

Posted Feb 10, 2008 11:57 UTC (Sun) by filipjoelsson (subscriber, #2622) [Link]

The last time I saw goto in serious use (where last is defined as most recent instance of
coding - not last time I viewed ancient code), was in some DirectX example code (shipped with
DirectX in ca 1999 - so it is a half way ancient, and it is MS being responsible).

Anyway, the way they used it was to skip to a dtor-like section of a function. Not
particularly elegant, not particularly object oriented, but not outright heresy either -
which, since I thought it _was_ outright heresy confused me.

Ever since, I have been trying to solve it in more elegant ways (using C++). Nested ifs,
functors, exceptions, you name it - it's not very elegant either. Anybody know a better way to
free resources in a C++ function when failing? (I tend to think the nested ifs are the least
evil - but they do eat the line width at an alarming rate.)

EORant ;)

It's not a mistake to ban goto... it's not a mistake to keep it...

Posted Feb 10, 2008 13:48 UTC (Sun) by nix (subscriber, #2304) [Link]

The Linux kernel uses goto in this manner all the time. It's useful if you 
can't do it by splitting the non-ctor-dtor stuff into a separate function.

freeing resources

Posted Feb 11, 2008 3:26 UTC (Mon) by pflugstad (subscriber, #224) [Link]

Are you looking for RAII?

I use it quite often for things like mutexes, but it can be used in many other situations - any time you hold a resource in a function that must be cleaned up when the function exits, including when throwing an exception.

It's a fairly common design pattern for C++ these days - I think I started seeing it a year or two ago...

freeing resources

Posted Feb 11, 2008 8:19 UTC (Mon) by nix (subscriber, #2304) [Link]

In C++ it's probably two decades old: I know it was landing in published 
textbooks ten years ago.

I didn't see it in C outside the Linux kernel before about 2004.