To LUGNET HomepageTo LUGNET News HomepageTo LUGNET Guide Homepage
 Help on Searching
 
Post new message to lugnet.roboticsOpen lugnet.robotics in your NNTP NewsreaderTo LUGNET News Traffic PageSign In (Members)
 Robotics / 21655
21654  |  21656
Subject: 
Re: Are there GOTO statements in NQC?
Newsgroups: 
lugnet.robotics
Date: 
Thu, 20 Nov 2003 23:35:51 GMT
Original-From: 
T. Alexander Popiel <POPIEL@antispamWOLFSKEEP.COM>
Viewed: 
1114 times
  
In message:  <3FBD39FE.3020300@airmail.net>
             Steve Baker <sjbaker1@airmail.net> writes:

"A function should have one and only one exit point" :-)

The thing everyone has forgotton about the "one exit point" rule
is that it applies to MACHINE CODE PROGRAMMING - not high level
languages.

This single misunderstanding has resulted in more convoluted and
illegible code than I care to remember!

When writing in machine code, you allocate yourself some stack space
for local variables at the top of the function and save whatever
hardware registers you are using onto the stack - and it's crucially
important that you free up that space and restore the saved registers
before you return from the function.

If the function has many exit points - and someone comes along and
adds another locals - or saves an additional register - then it's
crucially important to find each and every exit point and change the
code to reflect this new stack usage.

It's ALARMINGLY easy to forget one or more exit points - resulting
in flakey code.  Hence the very good rule to have just one exit point.

This exact same argument can be applied to high-level languages
where you dynamically allocate memory or other resources that need
to be released before leaving the function.  While I agree that
forcing people to have only one exit point at all times is overzealous,
there are times when it's quite appropriate... and amusingly enough, at
those times, I also find goto quite helpful.

Consider:

int my_function ()
{
  int result = FALSE;

  allocate_buffer;
  open_file;
  save_visual;
  create_gfx_context;

  if ( problem_exists ) goto exit_sequence;

  do_some_work;

  if ( some_other_problem_exists ) goto exit_sequence;

  do_more_work;

  result = TRUE;

exit_sequence:
  close_file;
  free_buffer;
  restore_visual;
  free_gfx_context;

  return result;
}

Some would argue that this just means that my_function() should be
broken into another_function_which_allocates_and_frees(), calling
my_new_function(with_everything_passed_in_as_args)... but I tend to
think that's silly and leads to horribly deep stackframes.

It's alarming that some teachers of programming are still teaching
the "only one exit" rule for high level languages!

I think that it's merely another aspect of overgeneralization.
It's useful sometimes, but not at other times.

- Alex



Message has 1 Reply:
  Re: Are there GOTO statements in NQC?
 
(...) Which is why we have C++. With memory constructors and destructors, quite complex resources can be automatically and cleanly free'd up on exit from a function. ---...--- Steve Baker ---...--- HomeEmail: <sjbaker1@airmail.net> WorkEmail: (...) (21 years ago, 20-Nov-03, to lugnet.robotics)

Message is in Reply To:
  Re: Are there GOTO statements in NQC?
 
(...) The thing everyone has forgotton about the "one exit point" rule is that it applies to MACHINE CODE PROGRAMMING - not high level languages. This single misunderstanding has resulted in more convoluted and illegible code than I care to (...) (21 years ago, 20-Nov-03, to lugnet.robotics)

19 Messages in This Thread:









Entire Thread on One Page:
Nested:  All | Brief | Compact | Dots
Linear:  All | Brief | Compact

This Message and its Replies on One Page:
Nested:  All | Brief | Compact | Dots
Linear:  All | Brief | Compact
    

Custom Search

©2005 LUGNET. All rights reserved. - hosted by steinbruch.info GbR