Remove non-numeric characters from string in Delphi-CodePudding

I have these three functions that successfully remove all non-numeric characters from a given string:

The first function loops through the characters of the input string, and if the current character is a number, it adds it to a new string that is returned as the result of the function.

  function RemoveNonNumericChars(const s: string): string;
  begin
    Result := '';
    for var i := 1 to Length(s) do
    begin
      if s[i] in ['0'..'9'] then
        Result := Result   s[i];
    end;
  end;

The second function loops through the characters of the input string from right to left, and if the current character is not a number, it uses the Delete function to remove it from the string

  function RemoveNonNumericChars(const s: string): string;
  begin
    Result := s;
    for var i := Length(Result) downto 1 do
    begin
      if not(Result[i] in ['0'..'9']) then
        Delete(Result, i, 1);
    end;
  end;

The third function uses a regular expression to replace all non-numeric characters with nothing, thus removing them. TRegEx is from the System.RegularExpressions unit.

  function RemoveNonNumericChars(const s: string): string;
  begin
    var RegEx := TRegEx.Create('[^0-9]');
    Result := RegEx.Replace(s, '');
  end;

All three of them do what I need, but I want to know if there is maybe a built-in function in Delphi for this... Or maybe even a better way to do it than the way I'm doing it. What's the best and/or fastest way to remove non-numeric characters from a string in Delphi?

CodePudding user response：

Both your approaches are slow because you constantly change the length of the string. Also, they only recognise Arabic digits.

To solve the performance issue, preallocate the maximum result length:

function RemoveNonDigits(const S: string): string;
begin
  SetLength(Result, S.Length);
  var LActualLength := 0;
  for var i := 1 to S.Length do
    if CharInSet(S[i],  ['0'..'9']) then
    begin
      Inc(LActualLength);
      Result[LActualLength] := S[i];
    end;
  SetLength(Result, LActualLength);
end;

To support non-Arabic digits, use the TCharacter.IsDigit function:

function RemoveNonDigits(const S: string): string;
begin
  SetLength(Result, S.Length);
  var LActualLength := 0;
  for var i := 1 to S.Length do
    if S[i].IsDigit then
    begin
      Inc(LActualLength);
      Result[LActualLength] := S[i];
    end;
  SetLength(Result, LActualLength);
end;

To optimise even further, as suggested by Stefan Glienke, you can bypass the RTL's string handling machinery and write each character directly with some loss of code readability:

function RemoveNonDigits(const S: string): string;
begin
  SetLength(Result, S.Length);
  var ResChr := PChar(Result);
  var LActualLength := 0;
  for var i := 1 to S.Length do
    if CharInSet(S[i],  ['0'..'9']) then
    begin
      Inc(LActualLength);
      ResChr^ := S[i];
      Inc(ResChr);
    end;
  SetLength(Result, LActualLength);
end;