All entries for Wednesday 22 January 2020

January 22, 2020

Magic numbers

We've mentioned magic number before but they are one of those things that are always worth mentioning again. Magic numbers are literal numbers that you write into your code that impact the clarity of the code. The alternative to magic numbers is to store the values into variables or preprocessor directives or other ways of associating a number with a name. Exactly what impacts the clarity of your code is often a bit subjective and is certainly research field specific but there are a few rules of thumb

  1. Is the number only an approximation? You don't want to risk using different approximations in different places in your code. Imagine what would happen if you defined Pi to have different values when using it in different contexts?
  2. Is the number immediately recognizable? As a physicist I'd recognize "0.5*m*v**2" when I see it in code as being the kinetic energy of an object despite the 0.5 but I'd have more trouble with "4.0e-7 * pi". It's probably the pre-2018 definition of the vacuum permeability but it isn't entirely clear.
  3. Is the number arbitrary? It's quite common to use a number to specify things like which problem to run or which optional package to use. You might remember what problem 3 is right now but you'll have to look if you ever come back to this code. Even worse, if you lose discipline then you might change what the numbers do if one of the test cases becomes redundant. Replacing your arbitrary numbers with named constants severely reduces these problems. If you give your problem number a sensible name then you'll have a much better chance of remembering what it does, and similarly if you ever remove a problem then if you remember to remove the named constant as well then you won't be able to compile the code when trying to run the old test. A lot of languages provide "enumerated types" or "enums" that allow you to automatically map numbers to names and can also prevent a user from chosing to supply a simple integer instead.

It can feel like this is unnecessary and slows you down when you are just writing a quick code, but one of the major problems with academic software is that it tends to grow. You write a code to solve a problem that you encounter during your research and you aren't terribly careful because you aren't going to keep it. But it is quite likely that you won't put it to one side and never touch it again. You might encounter a similar problem later and modify the code. At that point elements like magic numbers are annoying but you can usually work out what your own code is doing. The major problem comes if you move on and your code is inherited by someone else. In this case they might work everything out perfectly (which is good), they might find that they can't understand it and have to start again (which is annoying but not too troublesome) but worst of all is that they might misunderstand what your code is doing and make changes that are incorrect.

January 2020

Mo Tu We Th Fr Sa Su
Dec |  Today  | Feb
      1 2 3 4 5
6 7 8 9 10 11 12
13 14 15 16 17 18 19
20 21 22 23 24 25 26
27 28 29 30 31      

Search this blog



Blog archive

Not signed in
Sign in

Powered by BlogBuilder