Consider two implementations of the same function:

Implementation 1 (the old way):

function num_listing (n : word) : ansistring;
var i : word;  num : string; result : ansistring;
begin
  result := '';
  for i := 0 to n do begin
    str (n, num);
    result := result + num; end;
  num_listing := result;
end;

Implementation 2 (possible because fpc permits nonrecursive use of the function identifer):

function num_listing (n : word) : ansistring;
var i : word;  num : string;
begin
  num_listing := '';
  for i := 0 to n do begin
    str (n, num);
    num_listing := num_listing + num; end;
end;

Which performs better under fpc? In other words, do we incur a performance penalty in #2 for the nonrecursive use of the function identifier, particularly in a tight loop as shown? Or does the additional variable and additional assignment at the end of #1 make it the loser? I know I could run a test, but I want to hear the wisdom of the compiler writers. Thanks.

Mark

The #2 is faster. There is no penalty for the use of the function identifier.

If i know good, the result variable exists in function without declare it. Use result instead function name.

#Tsch : Balázs Csaba

If i know good, the result variable exists in function without declare it. Use result instead function name.

That is correct. But the result keyword is only supported in Delphi and ObjFpc modes