Code Density

http://www.brandonbloom.name/blog/2013/06/24/code-density/

by llambda • 13 years ago

45 9 13 years ago

spullara• 13 years ago

This is a pretty good article and cuts at the heart of the "readability" vs "understandability" debates I see. If we decide reading code is an O(N) problem, where N = number of lines, then his argument makes sense and you might just argue about the constant factor (he states 5, I bet it is closer to 20) between understanding a Java code base and understanding a Clojure code base. But the real issue might be that understanding a Clojure codebase as is O(N log N) or possibly O(N^2) if there are lots of interactions and lots of new syntactical constructs.

Has anyone ever measured a Reading Comprehension score for languages?

snprbob86• 13 years ago

This is a pretty good article

Thanks!

he states 5, I bet it is closer to 20

I think it varies greatly with many factors, but agree that it's probably closer to 20. I'm just trying to avoid selling silver bullets.

Has anyone ever measured a Reading Comprehension score for languages?

While language is a huuuge component of that constant factor, there are certainly other components. For example, if you want to write a sudoku solver, you should start by writing a backtracking constraint solver and then implement the constraint solver on top of it. You'll write less code than trying to solve sudoku directly, but that code will certainly be slower to read. This applies within a language just as much as it does across languages.

seanmcdirmid4• 13 years ago

There has been a lot of research in the past in trying to come up with metrics for cognitive complexity; e.g.

http://en.wikipedia.org/wiki/Cyclomatic_complexity

Much of this work, however, predates the heavy use of indirection and relatively opaque abstractions. In my opinion, functional programming suck on the "understand every bit of the program" metric given the amount of indirection in code (e.g. via higher-order functions, OO languages can suffer also via virtual dispatch and Hollywood principle architectures). However, they potentially excel at "understand the program in the context of these understood abstractions" metric.

Unfortunately, humans are very diverse in their abilities to learn and understand abstractions, especially ones that are based on math. So some people get FP very well, while many others seem to do better with the more naturalistic (i.e. based on natural language) abstractions such as objects. There are good reasons why OOP is more popular than FP beyond the vacuous "they simply don't get it" arguments.

gizmo686• 13 years ago

There are good reasons why OOP is more popular than FP beyond the vacuous "they simply don't get it" arguments.

I think the main reason is simply historic. OOP is a lot easier to implement (in terms of reasonable runtime efficiency) than FP. Because of this, it had a huge head start to get itself into common usage. FP has only recently gotten to the point where it is actually competative on performace.

In terms of 'getting it', computers and programing were originally the domain of people who were much more procient in math than now. I suspect many of them would have been more comftorable with FP, but implenting it was way beyond their time.

Code Density

Related Stories