Re: Re: Labeled Breaks (not the G-word)

Joao Cruz Morais wrote:

I'm sorry if I'm saying something that was already proposed, but why not use the Java approach to labeled breaks or something like this (dumb) example: while(true) as outer_cycle { $i = 0; while(true)

    if($i++ == 10) break outer_cycle;

}

well, this is only handling one of the three GOTO use cases that we seem to agree on as being 'ok' * escaping deeply nested loops or if statements this one it would handle, just as "break _number_;" does right now, so it would be a viable option even in parallel to GOTO or as a fallback position if we can't reach agreement on GOTO at all * jumping to a central error handling/cleanup point this it can't handle unless you use it in combination with an extra $error flag while this is a viable approach i prefere

    for (...) {
      for (...) {
        for (...) {
          ...
          if (something_failed()) goto error_handling;
          ...
        }
      }
    }
    return true;

    error_handling:
    ...
    return false;

to

    $error_flag = false;
    for (...) as outer_loop {
      for (...) {
        for (...) {
          ...
          if (something_failed()) {
             $error_flag = true;
             break outer_loop;
          }
          ...
        }
      }
    }

    if ($error_flag) {
      ...
      return false;
    }

    return true;

as now there are all those extra $error_flag lines and to see where the error condition is actually processed i first have to search up from the 'break' to the outer_loop definition, then down to the closing { of that loops body and than further on to the place where $error_flag is actually checked (if at all) with "goto error_handling;" on the other hand i just have to search for "error_handling:" and i'm at the error handling code in one step (and the compiler can even guarantee that the error_handling: label exists whereas a flag might actually end up being ignored and an exception might get all the way up the call stack without being called) * state machine implementations states get labels, state transitions goto labels -> pretty clean right now you have to implement stuff like this as switch() statements with an extra state variable being managed this has the advantage of being able to tell which state you are currently in and to have state transition logging in front of the switch, but i experience it as less readable and it doesn't perform as well as a GOTO based solution would as GOTOs can be resolved at compile time while switch requires run time comparisons -- Hartmut Holzgraefe, Senior Support Engineer . MySQL AB, www.mysql.com