Tutorial about character handling in Ada

Using GNAT

Save your souce code as Latin-1.

   pragma Wide_Character_Encoding (UPPER);
   Latin_1_Character_3 : constant Character := 'À';

Save your souce code as UTF-8.

   pragma Wide_Character_Encoding (UTF8);
   Latin_1_Character_3 : constant Character := 'À';

Using GNAT

Save your souce code as UTF-8.

   pragma Wide_Character_Encoding (UTF8);
   UTF_16_Character_1 : constant Wide_Character := 'Ǡ';

Using GNAT

Save your souce code as UTF-8.

   pragma Wide_Character_Encoding (UTF8);
   UTF_32_Character_1 : constant Wide_Wide_Character := 'Ǡ';

Note about JIS code literals in GNAT

Do NOT use pragma Wide_Character_Encoding (SHIFT_JIS); or pragma Wide_Character_Encoding (EUC);.
These options are useless because string literals with these pragmas are encoded to Wide_String contains raw values of JIS code. But raw values of JIS code is not used anywhere in fact. For example, A-version of Windows API requires Shift-JIS, not raw values of JIS code. Mail applications requires ISO-2022-JP (that is another encoding), not raw values of JIS code.
So you have to save source code as UTF-8 and convert strings at run-time. (Or save source code just as Shift-JIS or EUC-JP and write these encoded literals into String directly without pragma.)

Save your source code as UTF-8.

   pragma Wide_Character_Encoding (UTF8);
   UTF_16_Character_1 : constant Wide_Character := 'あ';

And convert it at runtime with libiconv, wcstombs or WideCharToMultiByte API.

(Or save your source code as Shift-JIS.

   Shift_JIS_Character_1 : constant String := "あ";

Please attention, the standard library (or GNAT runtime) can not handle multi-byte encoded String like Shift-JIS/EUC-JP(or UTF-8). You have to write almost string operations yourself.)

Iterate each multi-byte characters of Interfaces.C.char_array containing local encoding

Import mblen to get length of one multi-byte character.

   function mblen (
      s : access constant C.char;
      n : Interfaces.C.size_t;
      return Interfaces.C.int;
   pragma Import (C, mblen);

Then, iterate.

declare
   Text : Interfaces.C.char_array := ...string containing local encoding...;
   I : Interfaces.C.size_t := Text'First;
begin
   while I <= Text'Last loop
      declare
         Length : Interfaces.C.size_t := Interfaces.C.size_t (
            mblen (Text (I)'Access, Text'Last - I + 1));
         One_Multi_Byte_Character : Interfaces.C.char_array
            renames Text (I .. I + Length - 1);
      begin
         ...
         ... -- use One_Multi_Byte_Character
         ...
         I := I + Length;
      end;
   end loop;
end;

convert between local encoding and string types of Ada

TBD...

Iterate each multi-byte characters of Interfaces.C.char_array containing local encoding

Import IsDBCSLeadByte to get length of one multi-byte character.

   function IsDBCSLeadByte (
      TestChar : C.char) -- BYTE
      return Interfaces.C.int; -- BOOL
   pragma Import (C, IsDBCSLeadByte);

Then, iterate.

declare
   Text : Interfaces.C.char_array := ...string containing local encoding...;
   I : Interfaces.C.size_t := Text'First;
begin
   while I <= Text'Last loop
      declare
         Length : Interfaces.C.size_t :=
            Boolean'Pos (IsDBCSLeadByte (Text (I)) /= 0) + 1; -- 1 or 2
         One_Multi_Byte_Character : Interfaces.C.char_array
            renames Text (I .. I + Length - 1);
      begin
         ...
         ... -- use One_Multi_Byte_Character
         ...
         I := I + Length;
      end;
   end loop;
end;

convert between local encoding and string types of Ada

TBD...

But, using GNAT

But GNAT runtime does not convert command-line and environment variables from local encoding to Latin-1.
(This behavior seems bug, but useful.)
So we can get raw command-line and environment variables with Ada.Command_Line/Ada.Environment_Varaibles.
And, it's necessary to convert from local encoding to UTF-32 (Wide_Wide_String) or any encoding what you need.

Using GNAT

Use "ENCODING=UTF8".

   Open (
      File,
      Name => Ada.Strings.UTF_Encoding.Wide_Wide_Strings.Encode (Wide_Wide_File_Name),
      Form => "ENCODING=UTF8");

Refer http://gcc.gnu.org/onlinedocs/gnat_rm/FORM-Strings.html.

Name_Case_Equivalence

Ada.Directories.Name_Case_Equivalence are defined in Ada 2012.
Refer http://ada-auth.org/standards/12rm/html/RM-A-16.html.

But, Actually this function can not help you.
Because it may return Case_Preserving on NTFS (file system of Windows). It's case-insensitive for accessing an existing file but keeping cases of each letter. And, it may return Case_Preserving on HFS+ (file system of Mac OS X). It's case-insensitive for accessing an existing file but keeping cases of each letter. Then, are behaviors of these two file systems same?
No. Case-insensitive rules of these file systems are different. (See below.)
So you have to write function to compare two file names for each file systems.

wchar_t of GNAT

Interfaces.C.wchar_t of GNAT's implementation is more bad.

   type wchar_t is new Wide_Character;
   for wchar_t'Size use Standard'Wchar_T_Size;

On UNIX platforms, it can not hold Unicode characters over 16-bit. Constraint_Error will be raised.

Ada.Wide_Text_IO of GNAT

GNAT's implementation of Wide_Text_IO outputs UTF-8 encoded from UCS-2 in Wide_String that given to Put/Put_Line, irrespective of Windows.
As a result, the output will be garbled to read.
I think it's bug of GNAT.