Finding commented out code

1

A comment line can explain what a code is doing, which is fine, or it can be used to remove code, which causes clutter and the possibility for error – did the programmer really intend to comment out that line permanently? It would be nice to have a tool that distinguishes between explanatory comments and commented out code. It would test if a comment line, after removing the initial ! (or C or * in the first column for fixed format code) is a valid line of code.

A tool for Python code is

It would be nice to have something similar for Fortran, and it would not have to be 100% accurate to be useful – the programmer would ultimately decide which comment lines to remove. To start, a heuristic would be that commented lines where the first token after the !
is one of

function , subroutine , program , call , if , else , do, real, integer, double, logical, complex

or lines that contain an = or :: are possibly commented out code.

1 Like
2

I often deal with @Beliavsky’s problem by using ! to explain what my code is doing
and !! to remove code, in either case at the beginning of a line or after some code on the same line.

1 Like
3

At one time some compilers would treat anything with a D in column 1 in fixed format as special debug related code. I presume they still support them for backwards compatibility. It would be nice if there was a more modern free format version of this to save having to surround code with C preprocessor #ifdef DEBUG #endif blocks. That would allow you to keep ! or !! for comments only. Probably doesn’t save you anything other than not having to deal with the preprocessor but for some of us that would be a plus.

4

A partial solution would be to categorize separators in the line:

  • If words (a block of letters) are most of the time separated by one space, the probability that it is a text is rather high.
  • If words are often separated by symbols like = ( ) " / * - + : ; the probability that it is code is higher.

By using several such (imperfect) criteria yielding a probability we could improve the detection efficiency of the algorithm. Of course, lines beginning by things like end do, end if, end function… would first be filtered.

1 Like