Tuesday, May 28, 2019

Bad naming?

Example C++ code

The above fragment is from [1]. I never write loops like this. I use \(n\) for limits or counts, but never for a loop index.

Looking at this, I realized I have many of these "rules". Such as:

  1. \(x\), \(y\), \(z\), and \(v\), \(w\) are always double precision variables. (I used to subtract points if a student would write
        for (int x=0; ... ).
  2. \(i\), \(j\), \(k\) and \(m\), \(n\) are always integer variables.
  3. Never use \(l\) (i.e. \(\ell\)) as variable name, it is too close to the digit 1 (one).
  4. Don't use short integers (unless for a specific reason) or single precision variables.
  5. Use \(i\),\(j\),\(k\) as loop indices in a predictable way (e.g. for a (sparse) matrix: \(i\) for rows, \(j\) for columns, \(k\) for things like nonzero elements).
  6. The previous rule also applies to AMPL which uses local index names. E.g. after declaring
       param f_max {j in FOOD} >= f_min[j]; 
    I always use j for FOOD
  7. Use short names for items (variables) that are often used and for locally declared indices. Use long names for items that are sparsely used. I call this Huffman-code [2] naming. 
I am so used to this, that code that disobeys this in a flagrant way, just hurts my eyes. I find that, if I follow these simple rules, reading code is easier. It minimizes the surprise factor. Of course, writing code is for consumption by a compiler (or other tool), but more importantly: for consumption by a human reader.

So, that loop should look like:

const int n = 10;
for (int i = 0; i < n; ++i) {...}


  1. https://en.cppreference.com/w/cpp/numeric/random/uniform_int_distribution
  2. https://en.wikipedia.org/wiki/Huffman_coding

No comments:

Post a Comment