1 |
706 |
jeremybenn |
------------------------------------------------------------------------------
|
2 |
|
|
-- --
|
3 |
|
|
-- GNAT RUN-TIME COMPONENTS --
|
4 |
|
|
-- --
|
5 |
|
|
-- ADA.STRINGS.UTF_ENCODING.WIDE_WIDE_STRINGS --
|
6 |
|
|
-- --
|
7 |
|
|
-- B o d y --
|
8 |
|
|
-- --
|
9 |
|
|
-- Copyright (C) 2010-2011, Free Software Foundation, Inc. --
|
10 |
|
|
-- --
|
11 |
|
|
-- GNAT is free software; you can redistribute it and/or modify it under --
|
12 |
|
|
-- terms of the GNU General Public License as published by the Free Soft- --
|
13 |
|
|
-- ware Foundation; either version 3, or (at your option) any later ver- --
|
14 |
|
|
-- sion. GNAT is distributed in the hope that it will be useful, but WITH- --
|
15 |
|
|
-- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY --
|
16 |
|
|
-- or FITNESS FOR A PARTICULAR PURPOSE. --
|
17 |
|
|
-- --
|
18 |
|
|
-- As a special exception under Section 7 of GPL version 3, you are granted --
|
19 |
|
|
-- additional permissions described in the GCC Runtime Library Exception, --
|
20 |
|
|
-- version 3.1, as published by the Free Software Foundation. --
|
21 |
|
|
-- --
|
22 |
|
|
-- You should have received a copy of the GNU General Public License and --
|
23 |
|
|
-- a copy of the GCC Runtime Library Exception along with this program; --
|
24 |
|
|
-- see the files COPYING3 and COPYING.RUNTIME respectively. If not, see --
|
25 |
|
|
-- <http://www.gnu.org/licenses/>. --
|
26 |
|
|
-- --
|
27 |
|
|
-- GNAT was originally developed by the GNAT team at New York University. --
|
28 |
|
|
-- Extensive contributions were provided by Ada Core Technologies Inc. --
|
29 |
|
|
-- --
|
30 |
|
|
------------------------------------------------------------------------------
|
31 |
|
|
|
32 |
|
|
package body Ada.Strings.UTF_Encoding.Wide_Wide_Strings is
|
33 |
|
|
use Interfaces;
|
34 |
|
|
|
35 |
|
|
------------
|
36 |
|
|
-- Decode --
|
37 |
|
|
------------
|
38 |
|
|
|
39 |
|
|
-- Decode UTF-8/UTF-16BE/UTF-16LE input to Wide_Wide_String
|
40 |
|
|
|
41 |
|
|
function Decode
|
42 |
|
|
(Item : UTF_String;
|
43 |
|
|
Input_Scheme : Encoding_Scheme) return Wide_Wide_String
|
44 |
|
|
is
|
45 |
|
|
begin
|
46 |
|
|
if Input_Scheme = UTF_8 then
|
47 |
|
|
return Decode (Item);
|
48 |
|
|
else
|
49 |
|
|
return Decode (To_UTF_16 (Item, Input_Scheme));
|
50 |
|
|
end if;
|
51 |
|
|
end Decode;
|
52 |
|
|
|
53 |
|
|
-- Decode UTF-8 input to Wide_Wide_String
|
54 |
|
|
|
55 |
|
|
function Decode (Item : UTF_8_String) return Wide_Wide_String is
|
56 |
|
|
Result : Wide_Wide_String (1 .. Item'Length);
|
57 |
|
|
-- Result string (worst case is same length as input)
|
58 |
|
|
|
59 |
|
|
Len : Natural := 0;
|
60 |
|
|
-- Length of result stored so far
|
61 |
|
|
|
62 |
|
|
Iptr : Natural;
|
63 |
|
|
-- Input string pointer
|
64 |
|
|
|
65 |
|
|
C : Unsigned_8;
|
66 |
|
|
R : Unsigned_32;
|
67 |
|
|
|
68 |
|
|
procedure Get_Continuation;
|
69 |
|
|
-- Reads a continuation byte of the form 10xxxxxx, shifts R left by 6
|
70 |
|
|
-- bits, and or's in the xxxxxx to the low order 6 bits. On return Ptr
|
71 |
|
|
-- is incremented. Raises exception if continuation byte does not exist
|
72 |
|
|
-- or is invalid.
|
73 |
|
|
|
74 |
|
|
----------------------
|
75 |
|
|
-- Get_Continuation --
|
76 |
|
|
----------------------
|
77 |
|
|
|
78 |
|
|
procedure Get_Continuation is
|
79 |
|
|
begin
|
80 |
|
|
if Iptr > Item'Last then
|
81 |
|
|
Raise_Encoding_Error (Iptr - 1);
|
82 |
|
|
|
83 |
|
|
else
|
84 |
|
|
C := To_Unsigned_8 (Item (Iptr));
|
85 |
|
|
Iptr := Iptr + 1;
|
86 |
|
|
|
87 |
|
|
if C not in 2#10_000000# .. 2#10_111111# then
|
88 |
|
|
Raise_Encoding_Error (Iptr - 1);
|
89 |
|
|
else
|
90 |
|
|
R := Shift_Left (R, 6) or Unsigned_32 (C and 2#00_111111#);
|
91 |
|
|
end if;
|
92 |
|
|
end if;
|
93 |
|
|
end Get_Continuation;
|
94 |
|
|
|
95 |
|
|
-- Start of processing for Decode
|
96 |
|
|
|
97 |
|
|
begin
|
98 |
|
|
Iptr := Item'First;
|
99 |
|
|
|
100 |
|
|
-- Skip BOM at start
|
101 |
|
|
|
102 |
|
|
if Item'Length >= 3
|
103 |
|
|
and then Item (Iptr .. Iptr + 2) = BOM_8
|
104 |
|
|
then
|
105 |
|
|
Iptr := Iptr + 3;
|
106 |
|
|
|
107 |
|
|
-- Error if bad BOM
|
108 |
|
|
|
109 |
|
|
elsif Item'Length >= 2
|
110 |
|
|
and then (Item (Iptr .. Iptr + 1) = BOM_16BE
|
111 |
|
|
or else
|
112 |
|
|
Item (Iptr .. Iptr + 1) = BOM_16LE)
|
113 |
|
|
then
|
114 |
|
|
Raise_Encoding_Error (Iptr);
|
115 |
|
|
end if;
|
116 |
|
|
|
117 |
|
|
-- Loop through input characters
|
118 |
|
|
|
119 |
|
|
while Iptr <= Item'Last loop
|
120 |
|
|
C := To_Unsigned_8 (Item (Iptr));
|
121 |
|
|
Iptr := Iptr + 1;
|
122 |
|
|
|
123 |
|
|
-- Codes in the range 16#00# - 16#7F# are represented as
|
124 |
|
|
-- 0xxxxxxx
|
125 |
|
|
|
126 |
|
|
if C <= 16#7F# then
|
127 |
|
|
R := Unsigned_32 (C);
|
128 |
|
|
|
129 |
|
|
-- No initial code can be of the form 10xxxxxx. Such codes are used
|
130 |
|
|
-- only for continuations.
|
131 |
|
|
|
132 |
|
|
elsif C <= 2#10_111111# then
|
133 |
|
|
Raise_Encoding_Error (Iptr - 1);
|
134 |
|
|
|
135 |
|
|
-- Codes in the range 16#80# - 16#7FF# are represented as
|
136 |
|
|
-- 110yyyxx 10xxxxxx
|
137 |
|
|
|
138 |
|
|
elsif C <= 2#110_11111# then
|
139 |
|
|
R := Unsigned_32 (C and 2#000_11111#);
|
140 |
|
|
Get_Continuation;
|
141 |
|
|
|
142 |
|
|
-- Codes in the range 16#800# - 16#FFFF# are represented as
|
143 |
|
|
-- 1110yyyy 10yyyyxx 10xxxxxx
|
144 |
|
|
|
145 |
|
|
elsif C <= 2#1110_1111# then
|
146 |
|
|
R := Unsigned_32 (C and 2#0000_1111#);
|
147 |
|
|
Get_Continuation;
|
148 |
|
|
Get_Continuation;
|
149 |
|
|
|
150 |
|
|
-- Codes in the range 16#10000# - 16#10FFFF# are represented as
|
151 |
|
|
-- 11110zzz 10zzyyyy 10yyyyxx 10xxxxxx
|
152 |
|
|
|
153 |
|
|
elsif C <= 2#11110_111# then
|
154 |
|
|
R := Unsigned_32 (C and 2#00000_111#);
|
155 |
|
|
Get_Continuation;
|
156 |
|
|
Get_Continuation;
|
157 |
|
|
Get_Continuation;
|
158 |
|
|
|
159 |
|
|
-- Any other code is an error
|
160 |
|
|
|
161 |
|
|
else
|
162 |
|
|
Raise_Encoding_Error (Iptr - 1);
|
163 |
|
|
end if;
|
164 |
|
|
|
165 |
|
|
Len := Len + 1;
|
166 |
|
|
Result (Len) := Wide_Wide_Character'Val (R);
|
167 |
|
|
end loop;
|
168 |
|
|
|
169 |
|
|
return Result (1 .. Len);
|
170 |
|
|
end Decode;
|
171 |
|
|
|
172 |
|
|
-- Decode UTF-16 input to Wide_Wide_String
|
173 |
|
|
|
174 |
|
|
function Decode (Item : UTF_16_Wide_String) return Wide_Wide_String is
|
175 |
|
|
Result : Wide_Wide_String (1 .. Item'Length);
|
176 |
|
|
-- Result cannot be longer than the input string
|
177 |
|
|
|
178 |
|
|
Len : Natural := 0;
|
179 |
|
|
-- Length of result
|
180 |
|
|
|
181 |
|
|
Iptr : Natural;
|
182 |
|
|
-- Pointer to next element in Item
|
183 |
|
|
|
184 |
|
|
C : Unsigned_16;
|
185 |
|
|
R : Unsigned_32;
|
186 |
|
|
|
187 |
|
|
begin
|
188 |
|
|
-- Skip UTF-16 BOM at start
|
189 |
|
|
|
190 |
|
|
Iptr := Item'First;
|
191 |
|
|
|
192 |
|
|
if Iptr <= Item'Last and then Item (Iptr) = BOM_16 (1) then
|
193 |
|
|
Iptr := Iptr + 1;
|
194 |
|
|
end if;
|
195 |
|
|
|
196 |
|
|
-- Loop through input characters
|
197 |
|
|
|
198 |
|
|
while Iptr <= Item'Last loop
|
199 |
|
|
C := To_Unsigned_16 (Item (Iptr));
|
200 |
|
|
Iptr := Iptr + 1;
|
201 |
|
|
|
202 |
|
|
-- Codes in the range 16#0000#..16#D7FF# or 16#E000#..16#FFFD#
|
203 |
|
|
-- represent their own value.
|
204 |
|
|
|
205 |
|
|
if C <= 16#D7FF# or else C in 16#E000# .. 16#FFFD# then
|
206 |
|
|
Len := Len + 1;
|
207 |
|
|
Result (Len) := Wide_Wide_Character'Val (C);
|
208 |
|
|
|
209 |
|
|
-- Codes in the range 16#D800#..16#DBFF# represent the first of the
|
210 |
|
|
-- two surrogates used to encode the range 16#01_000#..16#10_FFFF".
|
211 |
|
|
-- The first surrogate provides 10 high order bits of the result.
|
212 |
|
|
|
213 |
|
|
elsif C <= 16#DBFF# then
|
214 |
|
|
R := Shift_Left ((Unsigned_32 (C) - 16#D800#), 10);
|
215 |
|
|
|
216 |
|
|
-- Error if at end of string
|
217 |
|
|
|
218 |
|
|
if Iptr > Item'Last then
|
219 |
|
|
Raise_Encoding_Error (Iptr - 1);
|
220 |
|
|
|
221 |
|
|
-- Otherwise next character must be valid low order surrogate
|
222 |
|
|
-- which provides the low 10 order bits of the result.
|
223 |
|
|
|
224 |
|
|
else
|
225 |
|
|
C := To_Unsigned_16 (Item (Iptr));
|
226 |
|
|
Iptr := Iptr + 1;
|
227 |
|
|
|
228 |
|
|
if C not in 16#DC00# .. 16#DFFF# then
|
229 |
|
|
Raise_Encoding_Error (Iptr - 1);
|
230 |
|
|
|
231 |
|
|
else
|
232 |
|
|
R := R or (Unsigned_32 (C) mod 2 ** 10);
|
233 |
|
|
|
234 |
|
|
-- The final adjustment is to add 16#01_0000 to get the
|
235 |
|
|
-- result back in the required 21 bit range.
|
236 |
|
|
|
237 |
|
|
R := R + 16#01_0000#;
|
238 |
|
|
Len := Len + 1;
|
239 |
|
|
Result (Len) := Wide_Wide_Character'Val (R);
|
240 |
|
|
end if;
|
241 |
|
|
end if;
|
242 |
|
|
|
243 |
|
|
-- Remaining codes are invalid
|
244 |
|
|
|
245 |
|
|
else
|
246 |
|
|
Raise_Encoding_Error (Iptr - 1);
|
247 |
|
|
end if;
|
248 |
|
|
end loop;
|
249 |
|
|
|
250 |
|
|
return Result (1 .. Len);
|
251 |
|
|
end Decode;
|
252 |
|
|
|
253 |
|
|
------------
|
254 |
|
|
-- Encode --
|
255 |
|
|
------------
|
256 |
|
|
|
257 |
|
|
-- Encode Wide_Wide_String in UTF-8, UTF-16BE or UTF-16LE
|
258 |
|
|
|
259 |
|
|
function Encode
|
260 |
|
|
(Item : Wide_Wide_String;
|
261 |
|
|
Output_Scheme : Encoding_Scheme;
|
262 |
|
|
Output_BOM : Boolean := False) return UTF_String
|
263 |
|
|
is
|
264 |
|
|
begin
|
265 |
|
|
if Output_Scheme = UTF_8 then
|
266 |
|
|
return Encode (Item, Output_BOM);
|
267 |
|
|
else
|
268 |
|
|
return From_UTF_16 (Encode (Item), Output_Scheme, Output_BOM);
|
269 |
|
|
end if;
|
270 |
|
|
end Encode;
|
271 |
|
|
|
272 |
|
|
-- Encode Wide_Wide_String in UTF-8
|
273 |
|
|
|
274 |
|
|
function Encode
|
275 |
|
|
(Item : Wide_Wide_String;
|
276 |
|
|
Output_BOM : Boolean := False) return UTF_8_String
|
277 |
|
|
is
|
278 |
|
|
Result : String (1 .. 4 * Item'Length + 3);
|
279 |
|
|
-- Worst case is four bytes per input byte + space for BOM
|
280 |
|
|
|
281 |
|
|
Len : Natural;
|
282 |
|
|
-- Number of output codes stored in Result
|
283 |
|
|
|
284 |
|
|
C : Unsigned_32;
|
285 |
|
|
-- Single input character
|
286 |
|
|
|
287 |
|
|
procedure Store (C : Unsigned_32);
|
288 |
|
|
pragma Inline (Store);
|
289 |
|
|
-- Store one output code (input is in range 0 .. 255)
|
290 |
|
|
|
291 |
|
|
-----------
|
292 |
|
|
-- Store --
|
293 |
|
|
-----------
|
294 |
|
|
|
295 |
|
|
procedure Store (C : Unsigned_32) is
|
296 |
|
|
begin
|
297 |
|
|
Len := Len + 1;
|
298 |
|
|
Result (Len) := Character'Val (C);
|
299 |
|
|
end Store;
|
300 |
|
|
|
301 |
|
|
-- Start of processing for Encode
|
302 |
|
|
|
303 |
|
|
begin
|
304 |
|
|
-- Output BOM if required
|
305 |
|
|
|
306 |
|
|
if Output_BOM then
|
307 |
|
|
Result (1 .. 3) := BOM_8;
|
308 |
|
|
Len := 3;
|
309 |
|
|
else
|
310 |
|
|
Len := 0;
|
311 |
|
|
end if;
|
312 |
|
|
|
313 |
|
|
-- Loop through characters of input
|
314 |
|
|
|
315 |
|
|
for Iptr in Item'Range loop
|
316 |
|
|
C := To_Unsigned_32 (Item (Iptr));
|
317 |
|
|
|
318 |
|
|
-- Codes in the range 16#00#..16#7F# are represented as
|
319 |
|
|
-- 0xxxxxxx
|
320 |
|
|
|
321 |
|
|
if C <= 16#7F# then
|
322 |
|
|
Store (C);
|
323 |
|
|
|
324 |
|
|
-- Codes in the range 16#80#..16#7FF# are represented as
|
325 |
|
|
-- 110yyyxx 10xxxxxx
|
326 |
|
|
|
327 |
|
|
elsif C <= 16#7FF# then
|
328 |
|
|
Store (2#110_00000# or Shift_Right (C, 6));
|
329 |
|
|
Store (2#10_000000# or (C and 2#00_111111#));
|
330 |
|
|
|
331 |
|
|
-- Codes in the range 16#800#..16#D7FF# or 16#E000#..16#FFFD# are
|
332 |
|
|
-- represented as
|
333 |
|
|
-- 1110yyyy 10yyyyxx 10xxxxxx
|
334 |
|
|
|
335 |
|
|
elsif C <= 16#D7FF# or else C in 16#E000# .. 16#FFFD# then
|
336 |
|
|
Store (2#1110_0000# or Shift_Right (C, 12));
|
337 |
|
|
Store (2#10_000000# or
|
338 |
|
|
Shift_Right (C and 2#111111_000000#, 6));
|
339 |
|
|
Store (2#10_000000# or (C and 2#00_111111#));
|
340 |
|
|
|
341 |
|
|
-- Codes in the range 16#10000# - 16#10FFFF# are represented as
|
342 |
|
|
-- 11110zzz 10zzyyyy 10yyyyxx 10xxxxxx
|
343 |
|
|
|
344 |
|
|
elsif C in 16#1_0000# .. 16#10_FFFF# then
|
345 |
|
|
Store (2#11110_000# or
|
346 |
|
|
Shift_Right (C, 18));
|
347 |
|
|
Store (2#10_000000# or
|
348 |
|
|
Shift_Right (C and 2#111111_000000_000000#, 12));
|
349 |
|
|
Store (2#10_000000# or
|
350 |
|
|
Shift_Right (C and 2#111111_000000#, 6));
|
351 |
|
|
Store (2#10_000000# or
|
352 |
|
|
(C and 2#00_111111#));
|
353 |
|
|
|
354 |
|
|
-- All other codes are invalid
|
355 |
|
|
|
356 |
|
|
else
|
357 |
|
|
Raise_Encoding_Error (Iptr);
|
358 |
|
|
end if;
|
359 |
|
|
end loop;
|
360 |
|
|
|
361 |
|
|
return Result (1 .. Len);
|
362 |
|
|
end Encode;
|
363 |
|
|
|
364 |
|
|
-- Encode Wide_Wide_String in UTF-16
|
365 |
|
|
|
366 |
|
|
function Encode
|
367 |
|
|
(Item : Wide_Wide_String;
|
368 |
|
|
Output_BOM : Boolean := False) return UTF_16_Wide_String
|
369 |
|
|
is
|
370 |
|
|
Result : UTF_16_Wide_String (1 .. 2 * Item'Length + 1);
|
371 |
|
|
-- Worst case is each input character generates two output characters
|
372 |
|
|
-- plus one for possible BOM.
|
373 |
|
|
|
374 |
|
|
Len : Integer;
|
375 |
|
|
-- Length of output string
|
376 |
|
|
|
377 |
|
|
C : Unsigned_32;
|
378 |
|
|
|
379 |
|
|
begin
|
380 |
|
|
-- Output BOM if needed
|
381 |
|
|
|
382 |
|
|
if Output_BOM then
|
383 |
|
|
Result (1) := BOM_16 (1);
|
384 |
|
|
Len := 1;
|
385 |
|
|
else
|
386 |
|
|
Len := 0;
|
387 |
|
|
end if;
|
388 |
|
|
|
389 |
|
|
-- Loop through input characters encoding them
|
390 |
|
|
|
391 |
|
|
for Iptr in Item'Range loop
|
392 |
|
|
C := To_Unsigned_32 (Item (Iptr));
|
393 |
|
|
|
394 |
|
|
-- Codes in the range 16#00_0000#..16#00_D7FF# or 16#E000#..16#FFFD#
|
395 |
|
|
-- are output unchanged
|
396 |
|
|
|
397 |
|
|
if C <= 16#00_D7FF# or else C in 16#E000# .. 16#FFFD# then
|
398 |
|
|
Len := Len + 1;
|
399 |
|
|
Result (Len) := Wide_Character'Val (C);
|
400 |
|
|
|
401 |
|
|
-- Codes in the range 16#01_0000#..16#10_FFFF# are output using two
|
402 |
|
|
-- surrogate characters. First 16#1_0000# is subtracted from the code
|
403 |
|
|
-- point to give a 20-bit value. This is then split into two separate
|
404 |
|
|
-- 10-bit values each of which is represented as a surrogate with the
|
405 |
|
|
-- most significant half placed in the first surrogate. The ranges of
|
406 |
|
|
-- values used for the two surrogates are 16#D800#-16#DBFF# for the
|
407 |
|
|
-- first, most significant surrogate and 16#DC00#-16#DFFF# for the
|
408 |
|
|
-- second, least significant surrogate.
|
409 |
|
|
|
410 |
|
|
elsif C in 16#1_0000# .. 16#10_FFFF# then
|
411 |
|
|
C := C - 16#1_0000#;
|
412 |
|
|
|
413 |
|
|
Len := Len + 1;
|
414 |
|
|
Result (Len) := Wide_Character'Val (16#D800# + C / 2 ** 10);
|
415 |
|
|
|
416 |
|
|
Len := Len + 1;
|
417 |
|
|
Result (Len) := Wide_Character'Val (16#DC00# + C mod 2 ** 10);
|
418 |
|
|
|
419 |
|
|
-- All other codes are invalid
|
420 |
|
|
|
421 |
|
|
else
|
422 |
|
|
Raise_Encoding_Error (Iptr);
|
423 |
|
|
end if;
|
424 |
|
|
end loop;
|
425 |
|
|
|
426 |
|
|
return Result (1 .. Len);
|
427 |
|
|
end Encode;
|
428 |
|
|
|
429 |
|
|
end Ada.Strings.UTF_Encoding.Wide_Wide_Strings;
|