The warning is confusing! Please help

[EganSolo]

I have to process text that contains non-ASCII characters. I understand that the String type in Free Pascal is keyed to UTF-8 by default. I don’t require Delphi compatibility, so mode Delphi isn’t essential for me.

Consider this bit of code:

Code: Pascal  [Select][+][-]

1. program UniCode;
2. {$H+}{$codepage utf8}
3. uses Classes, SysUtils, LazUTF8;
4.  
5. function IsNonAscii(const S : String; const aPos: integer): Boolean;
6. const NonAsciiChar = 'é';
7. //comment the line above and uncomment the line below to get rid of the warning:
8. //const NonAsciiChar : String = 'é';
9.  
10. var c : String;
11. begin
12.   c := UTF8Copy(s,apos,1);
13.   Result := c = NonAsciiChar;
14. end;
15.  
16. begin
17.  
18. end.
19.  
20.  

When I compile this code, the compiler issues the following warning: UniCode.lpr(12,13) Warning: Implicit string type conversion from “AnsiString” to “WideString”. However, if I switch to the typed constant, the warning disappears.

I don’t get it: even though I’m specifying $codepage utf8, why is the compiler defaulting string constants to WideString instead of String? Is there a way to switch that?


When I compile this, I get

[Martin_fr]

The codepage directive afaik does not control what type your constant gets.

It controls, if the bytes in your sourcefile are converted, and what from.

See the example.

Code: Pascal  [Select][+][-]

1. program P1;
2.  {$Codepage cp1250}
3.  { $codepage utf8}
4. type U8 = type AnsiString(CP_UTF8);
5. var s: U8;
6. begin
7.   s := #$B1#$B1;
8.   writeln(s);
9. end.

The string is always utf8.

But your source is not, with that directive, so the $B1 (same if you actually had a (or two) single char with that ordinal value) is converted from the codepage CP1250.  And it prints

Code: Text  [Select][+][-]

1. ±±

But if you said that the source was utf8 already, then the $B1 is invalid utf8, and print ??



Btw

Code: Pascal  [Select][+][-]

1. const NonAsciiChar = string('é');