[IC/Prog] Gawk's lint complains about empty strings

[Jorge Stolfi]

Hello,

Gawk's "lint" checker complains when the result of a {substr}
call would be an empty string.  In particular
  
  bash$ gawk --lint --source \
    ' BEGIN {
      x="12345"; 
      y = substr(x,3,0);
      y = substr(x,6,0);
      y = substr(x,6);
      print y;
    } '

gives the warnings

  gawk: cmd. line:2: warning: substr: length 0 is not >= 1
  gawk: cmd. line:3: warning: substr: length 0 is not >= 1
  gawk: cmd. line:4: warning: substr: start index 6 is past end of string
  gawk: cmd. line:5: warning: reference to uninitialized variable `y'

Note that the variable {y} gets incorrectly marked as "undefined"
because of the previous warnings. If one uses the "--lint=invalid"
option, things do not improve very much:

  bash$ gawk --lint=invalid --source \
    ' BEGIN {
      x="12345"; 
      y = substr(x,3,0);
      y = substr(x,6,0);
      y = substr(x,6);
      print y;
    } '
    
the first two errors are supressed, but the last two remain:

  gawk: cmd. line:4: warning: substr: start index 6 is past end of string
  gawk: cmd. line:5: warning: reference to uninitialized variable `y'

However, all three {substr} calls are perfectly valid and are often used in
scripts.  E.g.

  # Extract the substring {y} of {x} between '<'and '>'
  if (match(x, /[<].*[>]/)) { y = substr(x, RSTART+1, RLENGTH-2); }
  # Note that the length of the substring may be 0.

  # Remove everything from {x} up to the first ',' inclusive:
  if (match(x, /[,]/)) { x = substr(x, RSTART+RLENGTH); }
  # Note that the starting index may be {length(x)+1}.
  
I suggest that you fix gawk's "--lint" option so that

  * the call {substr(x,i,k)} is considered OK as long as
  {i >= 1}, {k >= 0}, and {i+k-1 <= length(x)}
  
  * the call {substr(x,i)} is considered OK as long as 
  {i >= 1} and {i <= length(x)+1}. 
  
Thank you,

--stolfi

Jorge Stolfi
Institute of Computing
State University of Campinas (UNICAMP)