Cooking With Technology

GOTO Can’t Die

GOTO is a statement all programmers know. Most of us are too young to have ever actually used it, but anyone with a formal education has been provided with GOTO as an example of the worst enabler of ’spaghetti code.’ So after accepting that something called GOTO is bad, we return to thinking about what we’ll do after class. But why, WHY is GOTO gone?

The disappearance of GOTO can be attributed to multiple factors, most of them well presented by the famous Edsger Dijkstra back in freaking 1968 in a letter to the editor of “Communications of the ACM.” He made some humorous points, such as “the quality of programmers is a decreasing function of the density of ‘go to’ statements in the programs they produce.” Maybe I’m further gone than most (run while you can!), but it made me chuckle. The main points that he was trying to convey however are less humorous: GOTO makes it harder to ‘prove’ programs[1], removes the ability for a programmer to determine the progress of the program in execution[2], and that GOTO is just too primitive of a statement.

After some jokes among computer scientists (see COME FROM - yes, our sense of humor is in need of calibration), GOTO really did start to disappear. Most major language designs lacked official GOTO reserved words, and others like ADA included it using a syntax unique only to that statement so it may be easily found in programs. This disappearance is only surface deep, however.

It turns out that while an unbounded GOTO may result in obnoxious and nondeterministic code, the ability to jump to a segment of code outside the normal flow of execution is pretty handy. ‘Case’ statements are a popular implementation of a restricted GOTO, with one conditional evaluation resulting in a jump to one of many locations in code. More importantly, error handling would be near impossible to implement without the concepts on which GOTO was founded.

Unless the developer is an idiot, error handling is implemented as a catch in case something unexpected happens (I’ve seen it used as part of normal execution, ick). When something unexpected happens, you can no longer depend on the state of execution within the program. In this case, using a GOTO is the only way to break away from whatever caused the error and make some attempt to either rectify it or exit cleanly. Even the parody COME FROM statement has a hidden usage. In debugging, the concepts in which the COME FROM statement was founded are used to handle breakpoints.

So GOTO isn’t really dead, it is just hidden and stripped of its freedom. And more importantly, it can’t die. We rely on its concepts far too much in structured programming to do away with it. Now that you know, spread the word that GOTOs aren’t inherently bad! They are like kids: they just need boundaries.

[1] Loops and conditional statements, and of course functions, can all be reduced to formal mathematical language. Especially when converted to their recursive forms, loops are incredibly easy to prove as being valid with induction. GOTO statements broke this by making code execution able to change state at any time, greatly reducing the ease in which algorithms could be reduced to and proven with math.

[2] Some people might say “hey, loops are kind of like GOTOs. What is the difference?” Or maybe not, but here is the difference anyway. As pointed out by Dijkstra, loops are able to have some index applied to their current iteration or state in recursion (think for(int i=0; i<end; i++) - i is the index). This allows the programmer to have definite knowledge of the state of execution through using multiple indexes as “coordinates.” With a GOTO, this idea breaks because it is trivial to jump to anywhere in the program at any time, making the current indexes useless.