Subroutines should be short—50 lines is often quoted as an upper
limit. This rule is widely known, widely accepted, and widely
Long subroutines have many technical problems
The technical problems are real, but it has always seemed to me that
they are somehow secondary: that there is something more fundamentally
wrong with long
subroutines. Something that gets to the very core of programming.
- They are hard to read and understand
- They invite dependencies between unrelated sections of code
- They are prone to sprawling control structures
- They impede code reuse
It finally occurred to me.
Long subroutines are run-on sentences.
Children who write run-on sentences haven't yet learned that there is
more to writing than transcribing their words. Good writing has
grammar, sentences, paragraphs, chapters: it has structure.
Programmers who write long subroutines haven't yet learned that there
is more to programming than transcribing their thoughts. Good software
has syntax, statements, subroutines, modules: it has structure.
Programmers who write long subroutines aren't doing the fundamental
work of programming: they aren't creating structure by which code can
be partitioned and organized. They are just transcribing their
thoughts, one after another.
Code vs. Prose
A 100 line paragraph would be solid block of single-spaced text, 2
pages long. Upon seeing a 100 line paragraph,
most readers would intuitively feel that something was wrong, either
with the ideas, or the writing, or the typesetting. Few would have the
persistence or the patience to read it.
As it happens, all software has syntax, because the compilers enforce
it. And programs above a certain size usually have some kind of module
structure, perhaps because management requires it, or perhaps because
they just become intractable otherwise.
But there doesn't seem to be anything that enforces structure at the
subroutine level. Programmers write 100, 200, 500, even 1000 line
subroutines, and as long as the code compiles and runs, no one
objects. Few people ever see the code, and fewer still seem to think
that it is a problem.
- This experience, nearly unknown in the print world, has become
rather common on the internet, as existing ASCII documents are posted
on the web without the necessary HTML markups.
Steven W. McDougall /
1999 Jun 27