18. The knowledge you have, working with one language isn't always applicable to another language

The fragment is taken from Putty project. Ineffective code is detected by the following PVS-Studio diagnostic: V814 Decreased performance. Calls to the 'strlen' function have being made multiple times when a condition for the loop's continuation was calculated.

static void tell_str(FILE * stream, char *str)
  unsigned int i;
  for (i = 0; i < strlen(str); ++i)
    tell_char(stream, str[i]);


There's no actual error here, but such code can be extremely inefficient when we deal with long strings, as the strlen() function is called in every loop iteration. So the error, if there is one here, is one of inefficiency.

As a rule, this kind of thing is typically found in code written by those that have previously worked with the Pascal language (or Delphi). In Pascal, the evaluation of the terminating condition of the loop is computed just once, thus this code is suitable and quite commonly used.

Let's have a look at an example of code written in Pascal. The word called will be printed only once, because the pstrlen() is called only once.

program test;
  i   : integer;
  str : string;

function pstrlen(str : string): integer;
  strlen := Length(str);

  str := 'a pascal string';
  for i:= 1 to pstrlen(str) do 

Effective code:

static void tell_str(FILE * stream, char *str)
  size_t i;
  const size_t len = strlen(str);
  for (i = 0; i < len; ++i)
    tell_char(stream, str[i]);


Don't forget that in C/C++, loop termination conditions are re-computed at the end of each and every iteration. Therefore it's not a good idea to call inefficient slow functions as part of this evaluation, especially if you can compute it just the once, before the loop is entered.

In some cases the compiler might be able to optimize the code with strlen(). For instance, if the pointer always refers to the same string literal, but we shouldn't rely on that in any way.

results matching ""

    No results matching ""