LCL Unicode Support

From Lazarus wiki
Revision as of 15:12, 22 June 2012 by Mschnell (talk | contribs) (Searching a substring)

Deutsch (de) English (en) español (es) français (fr) 日本語 (ja) 한국어 (ko) русский (ru) 中文(中国大陆)‎ (zh_CN) 中文(台灣)‎ (zh_TW)


As of 0.9.25, Lazarus has full Unicode support in all platforms except Gtk 1. In this page one can find instructions for Lazarus users, roadmaps, descriptions of basic concepts and implementation details.

Instructions for users

Even though Lazarus has Unicode widgetsets, it's important to note that not everything is Unicode. It's the responsibility of the developer to know what is the encoding of their strings and do the proper conversion between libraries which expect different encodings.

Usually the encoding is per library (e.g. a dynamic library DLL or a Lazarus package). Each library will uniformly expect 1 kind of encoding, which will usually either be Unicode (UTF-8 for Lazarus) or ANSI (which means the system encoding, and may be UTF-8 or not). The RTL and the FCL of FPC 2.4-2.6 expect ANSI strings.

You can convert between Unicode and ANSI using the UTF8ToAnsi and AnsiToUTF8 functions from the System unit or the UTF8ToSys and SysToUTF8 from the FileUtil unit. The latter two are smarter (faster) but pull more code into your program.

FPC is not Unicode aware

The Free Pascal Runtime Library (RTL), and the Free Pascal Free Component Library (FCL) in current FPC versions (up to 2.6.1) are ANSI, so you will need to convert strings coming from Unicode libraries or going to Unicode libraries (e.g. the LCL).

Converting between ANSI and Unicode


Say you get a string from a TEdit and you want to give it to some RTL file routine:

  MyString: string; // utf-8 encoded
  MyString := MyTEdit.Text;

And for the opposite direction:

  MyString: string; // ANSI encoded
  MyString := SomeRTLRoutine;
  MyTEdit.Text := AnsiToUTF8(MyString);

Important: UTF8ToAnsi will return an empty string if the UTF8 string contains invalid characters.

Important: AnsiToUTF8 and UTF8ToAnsi require a widestring manager under Linux, BSD and Mac OS X. You can use the SysToUTF8 and UTF8ToSys functions (unit FileUtil) or add the widestring manager by adding cwstring as one of the first units to your program's uses section.

Widestrings and Ansistrings

A widestring is a string type whose basic data holding elements have a size of 2 bytes. Widestrings almost always hold data in the UTF-16 encoding.

Note that while each data point accessible as an array of a widestring has 2 bytes, in UTF-16 a character may have 1 or 2 data points, which would then occupy 2 or 4 bytes. This means that accessing a Widestring as an array and expecting to obtain UTF-16 characters this way is completely wrong and will fail when a 4 byte character is present in the string. Note also that UTF-16, like UTF-8, may have decomposed characters. The character "Á" for example might be encoded as a single character or as 2 characters: "A" + a modifying accent. Thus Unicode is ambiguous: the same printed Text can be coded in multiple ways. Lazarus and FPC do not handle this in any automated way.

When passing Ansistrings to Widestrings you have to convert the encoding.

  w: widestring;
  w:='Über'; // wrong, because FPC will convert system codepage to UTF16
  w:=UTF8ToUTF16('Über'); // correct

Dealing with UTF8 strings and characters

Until Lazarus 0.9.30 the UTF-8 handling routines were in the LCL in the unit LCLProc. In Lazarus 0.9.31+ the routines in LCLProc are still available for backwards compatibility but the real code to deal with UTF-8 is located in the lazutils package in the unit lazutf8.

To execute operations on UTF-8 strings please use routines from the unit lazutf8 instead of routines from the SysUtils routine from Free Pascal, because SysUtils is not yet prepared to deal with Unicode, while lazutf8 is. Simply substitute the routines from SysUtils with their lazutf8 equivalent, which always has the same name except for an added "UTF8" prefix.

Also note that simply iterating over chars as if the string was an array does not work in Unicode. This is not something specific to UTF-8: one simply cannot suppose that a character will have a fixed size in Unicode. If you want to iterate over the characters of an UTF-8 string, there are basically two ways:

  • iterate over the bytes - useful for searching a substring or when looking only at the ASCII characters of the UTF8 string. For example when parsing XML files.
  • iterate over the characters - useful for graphical components like synedit. For example when you want to know the third printed character on the screen.

Searching a substring

Due to the special nature of UTF8 you can simply use the normal string functions for searching a sub-string. UTF-8 strings are valid ISO latin 1 strings if you reinterpret them that way (although the contents differ). So many routines made for ISO latin 1 and friends work in UTF-8, but not all, of course. For searching in a substring they do work perfectly:

uses lazutf8; // LCLProc for Lazarus 0.9.30 or inferior
procedure Where(SearchFor, aText: string);
  BytePos: LongInt;
  CharacterPos: LongInt;
  writeln('The substring "',SearchFor,'" is in the text "',aText,'"',
    ' at byte position ',BytePos,' and at character position ',CharacterPos);

Due to the ambiguity of Unicode, Pos() (just like any compare) might show unexpected behavior, when e.g. one of the string contains decomposed characters, while the other uses the direct codes for the same letter.

Accessing UTF8 characters

Unicode characters can vary in length, so the best solution for accessing them is to use an iteration when one intends to access the characters in the sequence in which they are. For iterating through the characters use this code:

uses lazutf8; // LCLProc for Lazarus 0.9.30 or lower
procedure DoSomethingWithString(AnUTF8String: string);
  p: PChar;
  CharLen: integer;
  FirstByte, SecondByte, ThirdByte: Char;
    CharLen := UTF8CharacterLength(p);

    // Here you have a pointer to the char and its length
    // You can access the bytes of the UTF-8 Char like this:
    if CharLen >= 1 then FirstByte := P[0];
    if CharLen >= 2 then SecondByte := P[1];
    if CharLen >= 3 then ThirdByte := P[2];

  until (CharLen=0) or (p^ = #0);

Accessing the Nth UTF8 character

Besides iterating one might also want to have random access to UTF-8 Characters.

uses lazutf8; // LCLProc for Lazarus 0.9.30 or inferior
  AnUTF8String, NthChar: string;
  NthChar := UTF8Copy(AnUTF8String, N, 1);

Showing character codepoints with UTF8CharacterToUnicode

The following demonstrates how to show the 32bit code point value of each character in an UTF8 string:

uses lazutf8; // LCLProc for Lazarus 0.9.30 or inferior
procedure IterateUTF8Characters(const AnUTF8String: string);
  p: PChar;
  unicode: Cardinal;
  CharLen: integer;
  until (CharLen=0) or (unicode=0);

UTF-8 String Copy, Length, LowerCase, etc

Nearly all operations which one might want to execute with UTF-8 strings are covered by the routines in the unit lazutf8 (unit LCLProc for Lazarus 0.9.30 or lower). See the following list of routines taken from lazutf8.pas:

function UTF8CharacterLength(p: PChar): integer;
function UTF8Length(const s: string): PtrInt;
function UTF8Length(p: PChar; ByteCount: PtrInt): PtrInt;
function UTF8CharacterToUnicode(p: PChar; out CharLen: integer): Cardinal;
function UnicodeToUTF8(u: cardinal; Buf: PChar): integer; inline;
function UnicodeToUTF8SkipErrors(u: cardinal; Buf: PChar): integer;
function UnicodeToUTF8(u: cardinal): shortstring; inline;
function UTF8ToDoubleByteString(const s: string): string;
function UTF8ToDoubleByte(UTF8Str: PChar; Len: PtrInt; DBStr: PByte): PtrInt;
function UTF8FindNearestCharStart(UTF8Str: PChar; Len: integer;
                                  BytePos: integer): integer;
// find the n-th UTF8 character, ignoring BIDI
function UTF8CharStart(UTF8Str: PChar; Len, CharIndex: PtrInt): PChar;
// find the byte index of the n-th UTF8 character, ignoring BIDI (byte len of substr)
function UTF8CharToByteIndex(UTF8Str: PChar; Len, CharIndex: PtrInt): PtrInt;
procedure UTF8FixBroken(P: PChar);
function UTF8CharacterStrictLength(P: PChar): integer;
function UTF8CStringToUTF8String(SourceStart: PChar; SourceLen: PtrInt) : string;
function UTF8Pos(const SearchForText, SearchInText: string): PtrInt;
function UTF8Copy(const s: string; StartCharIndex, CharCount: PtrInt): string;
procedure UTF8Delete(var s: String; StartCharIndex, CharCount: PtrInt);
procedure UTF8Insert(const source: String; var s: string; StartCharIndex: PtrInt);

function UTF8LowerCase(const AInStr: string; ALanguage: string=''): string;
function UTF8UpperCase(const AInStr: string; ALanguage: string=''): string;
function FindInvalidUTF8Character(p: PChar; Count: PtrInt;
                                  StopOnNonASCII: Boolean = false): PtrInt;
function ValidUTF8String(const s: String): String;

procedure AssignUTF8ListToAnsi(UTF8List, AnsiList: TStrings);

//compare functions

function UTF8CompareStr(const S1, S2: string): Integer;
function UTF8CompareText(const S1, S2: string): Integer;

Dealing with directory and filenames

Lazarus controls and functions expect filenames and directory names in UTF-8 encoding, but the RTL uses ANSI strings for directories and filenames.

For example, consider a button which sets the Directory property of the TFileListBox to the current directory. The RTL Function GetCurrentDir is ANSI, not Unicode, so conversion is needed:

procedure TForm1.Button1Click(Sender: TObject);
  // or use the functions from the FileUtil unit

The unit FileUtil defines common file functions with UTF-8 strings:

// basic functions similar to the RTL but working with UTF-8 instead of the
// system encoding

// AnsiToUTF8 and UTF8ToAnsi need a widestring manager under Linux, BSD, Mac OS X
// but normally these OS use UTF-8 as system encoding so the widestringmanager
// is not needed.
function NeedRTLAnsi: boolean;// true if system encoding is not UTF-8
procedure SetNeedRTLAnsi(NewValue: boolean);
function UTF8ToSys(const s: string): string;// as UTF8ToAnsi but more independent of widestringmanager
function SysToUTF8(const s: string): string;// as AnsiToUTF8 but more independent of widestringmanager

// file operations
function FileExistsUTF8(const Filename: string): boolean;
function FileAgeUTF8(const FileName: string): Longint;
function DirectoryExistsUTF8(const Directory: string): Boolean;
function ExpandFileNameUTF8(const FileName: string): string;
function ExpandUNCFileNameUTF8(const FileName: string): string;
{$IFNDEF VER2_2_0}
function ExtractShortPathNameUTF8(Const FileName : String) : String;
function FindFirstUTF8(const Path: string; Attr: Longint; out Rslt: TSearchRec): Longint;
function FindNextUTF8(var Rslt: TSearchRec): Longint;
procedure FindCloseUTF8(var F: TSearchrec);
function FileSetDateUTF8(const FileName: String; Age: Longint): Longint;
function FileGetAttrUTF8(const FileName: String): Longint;
function FileSetAttrUTF8(const Filename: String; Attr: longint): Longint;
function DeleteFileUTF8(const FileName: String): Boolean;
function RenameFileUTF8(const OldName, NewName: String): Boolean;
function FileSearchUTF8(const Name, DirList : String): String;
function FileIsReadOnlyUTF8(const FileName: String): Boolean;
function GetCurrentDirUTF8: String;
function SetCurrentDirUTF8(const NewDir: String): Boolean;
function CreateDirUTF8(const NewDir: String): Boolean;
function RemoveDirUTF8(const Dir: String): Boolean;
function ForceDirectoriesUTF8(const Dir: string): Boolean;

// environment
function ParamStrUTF8(Param: Integer): string;
function GetEnvironmentStringUTF8(Index : Integer): String;
function GetEnvironmentVariableUTF8(const EnvVar: String): String;
function GetAppConfigDirUTF8(Global: Boolean): string;

Mac OS X

The file functions of the FileUtil unit also take care of Mac OS X specific behaviour: OS X normalizes filenames. For example the filename 'ä.txt' can be encoded in Unicode with two different sequences (#$C3#$A4 and 'a'#$CC#$88). Under Linux and BSD you can create a filename with both encodings. OS X automatically converts the a umlaut to the three byte sequence. This means:

if Filename1 = Filename2 then ... // is not sufficient under OS X
if AnsiCompareFileName(Filename1, Filename2) = 0 then ... // not sufficient under fpc 2.2.2, not even with cwstring
if CompareFilenames(Filename1, Filename2) = 0 then ... // this always works (unit FileUtil or FileProcs

East Asian languages on Windows

The default font (Tahoma) for user interface controls under Windows XP is capable of correctly displaying several scripts/alphabets/languages, including Arabic, Russian (Cyrillic alphabet) and Western languages (Latin/Greek alphabets), but not East Asian languages, like Chinese, Japanese and Korean.

Simply by going to the Control Panel, choosing Regional Settings, clicking on the Languages Tab and installing the East Asia Language Pack, the standard user interface font will start showing those languages correctly. Obviously Windows XP versions localized for those languages will already have this language pack installed. Extended instructions here.

Later Windows versions presumably have support for these languages out of the box.

Free Pascal Particularities

UTF8 and source files - the missing BOM

When you create source files with Lazarus and type some non-ASCII characters the file is saved in UTF8. It does not use a BOM (Byte Order Mark). You can change the encoding via right click on source editor / File Settings / Encoding. Apart from the fact that UTF-8 files are not supposed to have BOMs, the reason for the lacking BOM is how FPC treats Ansistrings. For compatibility the LCL uses Ansistrings and for portability the LCL uses UTF8.

Note: Some MS Windows text editors might treat the files as encoded with the system codepage (OEM codepage) and show them as invalid characters. Do not add the BOM. If you add the BOM you have to change all string assignments.

For example:

Button1.Caption := 'Über';

When no BOM is given (and no codepage parameter was passed) the compiler treats the string as system encoding and copies each byte unconverted to the string. This is how the LCL expects strings.

// source file saved as UTF without BOM
if FileExists('Über.txt') then ; // wrong, because FileExists expects system encoding
if FileExistsUTF8('Über.txt') then ; // correct

Unicode essentials

The Unicode standard maps integers from 0 to 10FFFF(h) to characters. Each such mapping is called a code point. In other words, Unicode characters are in principle defined for code points from U+000000 to U+10FFFF (0 to 1 114 111).

There are three major schemes for representing Unicode code points as unique byte sequences. These schemes are called Unicode transformation formats: UTF-8, UTF-16 and UTF-32. Conversions between all of them are possible. Here are their basic properties:

                           UTF-8 UTF-16 UTF-32
Smallest code point [hex] 000000 000000 000000
Largest code point  [hex] 10FFFF 10FFFF 10FFFF
Code unit size [bits]          8     16     32
Minimal bytes/character        1      2      4
Maximal bytes/character        4      4      4

UTF-8 has several important and useful properties: It is interpreted as a sequence of bytes, so that the concept of lo- and hi-order byte does not exist. Unicode characters U+0000 to U+007F (ASCII) are encoded simply as bytes 00h to 7Fh (ASCII compatibility). This means that files and strings which contain only 7-bit ASCII characters have the same encoding under both ASCII and UTF-8. All characters >U+007F are encoded as a sequence of several bytes, each of which has the two most significant bits set. No byte sequence of one character is contained within a longer byte sequence of another character. This allows easy searching for substrings. The first byte of a multibyte sequence (representing a non-ASCII character) is always in the range C0h to FDh and it indicates how many bytes follow for this character. All further bytes in a multibyte sequence are in the range 80h to BFh. This allows easy resynchronization and robustness.

UTF-16 has the following most important properties: It uses a single 16-bit word to encode characters from U+0000 to U+d7ff, and a pair of 16-bit words to encode any of the remaining Unicode characters.

Finally, any Unicode character can be represented as a single 4 byte/32-bit unit in UTF-32.

For more, see: Unicode FAQ - Basic questions, Unicode FAQ - UTF-8, UTF-16, UTF-32 & BOM, Wikipedia: UTF-8 [1]

Implementation Details

Lazarus/LCL generally uses only UTF-8

Since the GTK1 interface was declared obsolete in Lazarus 0.9.31, all LCL interfaces are Unicode capable and Lazarus and the LCL use and accept only UTF-8 encoded strings, unless in routines explicitly marked as accepting other encodings.

Unicode-enabling the win32 interface


First, and most importantly, all Unicode patches for the Win32 interface must be enclosed by IFDEF WindowsUnicodeSupport, to avoid breaking the existing ANSI interface. After this stabilizes, all ifdefs will be removed and only the Unicode part will remain. At his moment all existing programs that use ANSI characters will need migration to Unicode.

No Unicode support on Win9x

Windows platforms <=Win9x are based on ISO code page standards and only partially support Unicode. Windows platforms starting with WinNT and Windows CE fully support Unicode.

Win 9x and NT offer two parallel sets of API functions: the old ANSI enabled *A and the new, Unicode enabled *W. *W functions accept wide strings, i.e. UTF-16 encoded strings, as parameters. Windows 9x has all *W functions but they have empty implementations. So they do nothing. Except for some of them which are fully implemented even in 9x and listed below in the section "Wide functions present on Windows 9x". This property is relevant as it allows to have one single application for both Win9x and WinNT and detect at runtime which set of APIs to use.

Windows CE only uses Wide API functions.

Wide functions present on Windows 9x

Some Wide API functions are present on Windows 9x. Here is a list of such functions:

Conversion example:

GetTextExtentPoint32(hdcNewBitmap, LPSTR(ButtonCaption),
Length(ButtonCaption), TextSize);


{$ifdef WindowsUnicodeSupport}
  GetTextExtentPoint32W(hdcNewBitmap, PWideChar(Utf8Decode(ButtonCaption)), Length(WideCaption), TextSize);
  GetTextExtentPoint32(hdcNewBitmap, LPSTR(ButtonCaption), Length(ButtonCaption), TextSize);

Functions that need Ansi and Wide versions

First Conversion example:

function TGDIWindow.GetTitle: String;
 l: Integer;
   l := Windows.GetWindowTextLength(Handle);
   SetLength(Result, l);
   Windows.GetWindowText(Handle, @Result[1], l);


function TGDIWindow.GetTitle: String;
  l: Integer;
  AnsiBuffer: string;
  WideBuffer: WideString;

{$ifdef WindowsUnicodeSupport}

if UnicodeEnabledOS then
  l := Windows.GetWindowTextLengthW(Handle);
  SetLength(WideBuffer, l);
  l := Windows.GetWindowTextW(Handle, @WideBuffer[1], l);
  SetLength(WideBuffer, l);
  Result := Utf8Encode(WideBuffer);
  l := Windows.GetWindowTextLength(Handle);
  SetLength(AnsiBuffer, l);
  l := Windows.GetWindowText(Handle, @AnsiBuffer[1], l);
  SetLength(AnsiBuffer, l);
  Result := AnsiToUtf8(AnsiBuffer);


   l := Windows.GetWindowTextLength(Handle);
   SetLength(Result, l);
   Windows.GetWindowText(Handle, @Result[1], l);




Lazarus Unicode Test.png

FPC codepages

Why does the LCL not use codepage for sources?

The -Fcutf8 and {$codepage utf8} exist since ages.

The following programs require FPC 2.7.1. The described problems exist on older compilers too.

There are some traps with -Fcutf8 and {$codepage utf8}. It only works if the RTL DefaultSystemCodePage is CP_UTF8. Otherwise your strings are converted by the compiler. For example under Linux the RTL default is CP_ACP, which defaults to ISO_8859-1. The RTL does not read your environment language on its own. So the the default is ISO_8859-1. This means your UTF-8 string constants are converted by the compiler:

Compile this with -Fcutf8 and run it on a Linux with LANG set to utf-8:

program project1;
{$mode objfpc}{$H+}
  writeln(DefaultSystemCodePage,' ',CP_UTF8);

This results in

0 65001

The LCL uses a widestringmanager (at the moment cwstring), which sets the DefaultSystemCodePage. You can do the same in your non LCL programs:

program project1;
{$mode objfpc}{$H+}
uses cwstring;
  writeln(DefaultSystemCodePage,' ',CP_UTF8);

This results in

65001 65001

The above is a simplification though, a lie for children. See the program below:

program project1;
{$mode objfpc}{$H+}
uses cwstring;
  a,b,c: string;
  writeln(DefaultSystemCodePage,' ',CP_UTF8);
  a:='ä'; b:='='#$C3#$A4; // #$C3#$A4 is UTF-8 for ä
  c:=       'ä='#$C3#$A4;
  writeln(a,b); // writes ä=ä
  writeln(c);   // writes ä=ä
65001 65001

You can see that an UTF-8 string constant works, a string constant with UTF-8 codes works too, but the combination does not work. The above was compiled with -Fcutf8 and uses cwstring to set the DefaultSystemCodePage to CP_UTF8.

So what went wrong?

The compiler treats any non ASCII string constant (here: the ä) as widestring (UCS2, not UTF-16).

You can not fool the compiler with 'ä='+#$C3#$A4. You must define two separate string constants.

Using any character outside the UCS-2 range results in

Fatal: illegal character "'�'" ($F0)

You can specify them with UTF-16 codes: #$D834#$DD1E. Yes, you read right. Specifying the codepage with -Fcutf8 or {$codepage utf8} actually defines a mix of UTF-8 and UTF-16.

Now compile the above without -Fcutf8:

65001 65001

Wow, everything looks as expected. You can even mix ASCII and non-ASCII string constants. Without the codepage the compiler stores string constants as byte sequences. That's what UTF-8 is.

That's one of the reasons why LCL applications do not use the codepage flags.

msegui has implemented an ecosystem of widestrings, so it works better with a codepage.

See Also

  • UTF-8 - Description of UTF-8 strings