1 /*
2 **********************************************************************
3 * Copyright (C) 2000-2015, International Business Machines
4 * Corporation and others. All Rights Reserved.
5 **********************************************************************
6 * file name: ucnv_lmb.cpp
7 * encoding: US-ASCII
8 * tab size: 4 (not used)
9 * indentation:4
10 *
11 * created on: 2000feb09
12 * created by: Brendan Murray
13 * extensively hacked up by: Jim Snyder-Grant
14 *
15 * Modification History:
16 *
17 * Date Name Description
18 *
19 * 06/20/2000 helena OS/400 port changes; mostly typecast.
20 * 06/27/2000 Jim Snyder-Grant Deal with partial characters and small buffers.
21 * Add comments to document LMBCS format and implementation
22 * restructured order & breakdown of functions
23 * 06/28/2000 helena Major rewrite for the callback API changes.
24 */
25
26 #include "unicode/utypes.h"
27
28 #if !UCONFIG_NO_CONVERSION && !UCONFIG_NO_LEGACY_CONVERSION && !UCONFIG_ONLY_HTML_CONVERSION
29
30 #include "unicode/ucnv_err.h"
31 #include "unicode/ucnv.h"
32 #include "unicode/uset.h"
33 #include "cmemory.h"
34 #include "cstring.h"
35 #include "uassert.h"
36 #include "ucnv_imp.h"
37 #include "ucnv_bld.h"
38 #include "ucnv_cnv.h"
39
40 #ifdef EBCDIC_RTL
41 #include "ascii_a.h"
42 #endif
43
44 /*
45 LMBCS
46
47 (Lotus Multi-Byte Character Set)
48
49 LMBCS was invented in the late 1980's and is primarily used in Lotus Notes
50 databases and in Lotus 1-2-3 files. Programmers who work with the APIs
51 into these products will sometimes need to deal with strings in this format.
52
53 The code in this file provides an implementation for an ICU converter of
54 LMBCS to and from Unicode.
55
56 Since the LMBCS character set is only sparsely documented in existing
57 printed or online material, we have added extensive annotation to this
58 file to serve as a guide to understanding LMBCS.
59
60 LMBCS was originally designed with these four sometimes-competing design goals:
61
62 -Provide encodings for the characters in 12 existing national standards
63 (plus a few other characters)
64 -Minimal memory footprint
65 -Maximal speed of conversion into the existing national character sets
66 -No need to track a changing state as you interpret a string.
67
68
69 All of the national character sets LMBCS was trying to encode are 'ANSI'
70 based, in that the bytes from 0x20 - 0x7F are almost exactly the
71 same common Latin unaccented characters and symbols in all character sets.
72
73 So, in order to help meet the speed & memory design goals, the common ANSI
74 bytes from 0x20-0x7F are represented by the same single-byte values in LMBCS.
75
76 The general LMBCS code unit is from 1-3 bytes. We can describe the 3 bytes as
77 follows:
78
79 [G] D1 [D2]
80
81 That is, a sometimes-optional 'group' byte, followed by 1 and sometimes 2
82 data bytes. The maximum size of a LMBCS chjaracter is 3 bytes:
83 */
84 #define ULMBCS_CHARSIZE_MAX 3
85 /*
86 The single-byte values from 0x20 to 0x7F are examples of single D1 bytes.
87 We often have to figure out if byte values are below or above this, so we
88 use the ANSI nomenclature 'C0' and 'C1' to refer to the range of control
89 characters just above & below the common lower-ANSI range */
90 #define ULMBCS_C0END 0x1F
91 #define ULMBCS_C1START 0x80
92 /*
93 Since LMBCS is always dealing in byte units. we create a local type here for
94 dealing with these units of LMBCS code units:
95
96 */
97 typedef uint8_t ulmbcs_byte_t;
98
99 /*
100 Most of the values less than 0x20 are reserved in LMBCS to announce
101 which national character standard is being used for the 'D' bytes.
102 In the comments we show the common name and the IBM character-set ID
103 for these character-set announcers:
104 */
105
106 #define ULMBCS_GRP_L1 0x01 /* Latin-1 :ibm-850 */
107 #define ULMBCS_GRP_GR 0x02 /* Greek :ibm-851 */
108 #define ULMBCS_GRP_HE 0x03 /* Hebrew :ibm-1255 */
109 #define ULMBCS_GRP_AR 0x04 /* Arabic :ibm-1256 */
110 #define ULMBCS_GRP_RU 0x05 /* Cyrillic :ibm-1251 */
111 #define ULMBCS_GRP_L2 0x06 /* Latin-2 :ibm-852 */
112 #define ULMBCS_GRP_TR 0x08 /* Turkish :ibm-1254 */
113 #define ULMBCS_GRP_TH 0x0B /* Thai :ibm-874 */
114 #define ULMBCS_GRP_JA 0x10 /* Japanese :ibm-943 */
115 #define ULMBCS_GRP_KO 0x11 /* Korean :ibm-1261 */
116 #define ULMBCS_GRP_TW 0x12 /* Chinese SC :ibm-950 */
117 #define ULMBCS_GRP_CN 0x13 /* Chinese TC :ibm-1386 */
118
119 /*
120 So, the beginning of understanding LMBCS is that IF the first byte of a LMBCS
121 character is one of those 12 values, you can interpret the remaining bytes of
122 that character as coming from one of those character sets. Since the lower
123 ANSI bytes already are represented in single bytes, using one of the character
124 set announcers is used to announce a character that starts with a byte of
125 0x80 or greater.
126
127 The character sets are arranged so that the single byte sets all appear
128 before the multi-byte character sets. When we need to tell whether a
129 group byte is for a single byte char set or not we use this define: */
130
131 #define ULMBCS_DOUBLEOPTGROUP_START 0x10
132
133 /*
134 However, to fully understand LMBCS, you must also understand a series of
135 exceptions & optimizations made in service of the design goals.
136
137 First, those of you who are character set mavens may have noticed that
138 the 'double-byte' character sets are actually multi-byte character sets
139 that can have 1 or two bytes, even in the upper-ascii range. To force
140 each group byte to introduce a fixed-width encoding (to make it faster to
141 count characters), we use a convention of doubling up on the group byte
142 to introduce any single-byte character > 0x80 in an otherwise double-byte
143 character set. So, for example, the LMBCS sequence x10 x10 xAE is the
144 same as '0xAE' in the Japanese code page 943.
145
146 Next, you will notice that the list of group bytes has some gaps.
147 These are used in various ways.
148
149 We reserve a few special single byte values for common control
150 characters. These are in the same place as their ANSI eqivalents for speed.
151 */
152
153 #define ULMBCS_HT 0x09 /* Fixed control char - Horizontal Tab */
154 #define ULMBCS_LF 0x0A /* Fixed control char - Line Feed */
155 #define ULMBCS_CR 0x0D /* Fixed control char - Carriage Return */
156
157 /* Then, 1-2-3 reserved a special single-byte character to put at the
158 beginning of internal 'system' range names: */
159
160 #define ULMBCS_123SYSTEMRANGE 0x19
161
162 /* Then we needed a place to put all the other ansi control characters
163 that must be moved to different values because LMBCS reserves those
164 values for other purposes. To represent the control characters, we start
165 with a first byte of 0xF & add the control chaarcter value as the
166 second byte */
167 #define ULMBCS_GRP_CTRL 0x0F
168
169 /* For the C0 controls (less than 0x20), we add 0x20 to preserve the
170 useful doctrine that any byte less than 0x20 in a LMBCS char must be
171 the first byte of a character:*/
172 #define ULMBCS_CTRLOFFSET 0x20
173
174 /*
175 Where to put the characters that aren't part of any of the 12 national
176 character sets? The first thing that was done, in the earlier years of
177 LMBCS, was to use up the spaces of the form
178
179 [G] D1,
180
181 where 'G' was one of the single-byte character groups, and
182 D1 was less than 0x80. These sequences are gathered together
183 into a Lotus-invented doublebyte character set to represent a
184 lot of stray values. Internally, in this implementation, we track this
185 as group '0', as a place to tuck this exceptions list.*/
186
187 #define ULMBCS_GRP_EXCEPT 0x00
188 /*
189 Finally, as the durability and usefulness of UNICODE became clear,
190 LOTUS added a new group 0x14 to hold Unicode values not otherwise
191 represented in LMBCS: */
192 #define ULMBCS_GRP_UNICODE 0x14
193 /* The two bytes appearing after a 0x14 are intrepreted as UFT-16 BE
194 (Big-Endian) characters. The exception comes when the UTF16
195 representation would have a zero as the second byte. In that case,
196 'F6' is used in its place, and the bytes are swapped. (This prevents
197 LMBCS from encoding any Unicode values of the form U+F6xx, but that's OK:
198 0xF6xx is in the middle of the Private Use Area.)*/
199 #define ULMBCS_UNICOMPATZERO 0xF6
200
201 /* It is also useful in our code to have a constant for the size of
202 a LMBCS char that holds a literal Unicode value */
203 #define ULMBCS_UNICODE_SIZE 3
204
205 /*
206 To squish the LMBCS representations down even further, and to make
207 translations even faster,sometimes the optimization group byte can be dropped
208 from a LMBCS character. This is decided on a process-by-process basis. The
209 group byte that is dropped is called the 'optimization group'.
210
211 For Notes, the optimzation group is always 0x1.*/
212 #define ULMBCS_DEFAULTOPTGROUP 0x1
213 /* For 1-2-3 files, the optimzation group is stored in the header of the 1-2-3
214 file.
215
216 In any case, when using ICU, you either pass in the
217 optimization group as part of the name of the converter (LMBCS-1, LMBCS-2,
218 etc.). Using plain 'LMBCS' as the name of the converter will give you
219 LMBCS-1.
220
221
222 *** Implementation strategy ***
223
224
225 Because of the extensive use of other character sets, the LMBCS converter
226 keeps a mapping between optimization groups and IBM character sets, so that
227 ICU converters can be created and used as needed. */
228
229 /* As you can see, even though any byte below 0x20 could be an optimization
230 byte, only those at 0x13 or below can map to an actual converter. To limit
231 some loops and searches, we define a value for that last group converter:*/
232
233 #define ULMBCS_GRP_LAST 0x13 /* last LMBCS group that has a converter */
234
235 static const char * const OptGroupByteToCPName[ULMBCS_GRP_LAST + 1] = {
236 /* 0x0000 */ "lmb-excp", /* internal home for the LOTUS exceptions list */
237 /* 0x0001 */ "ibm-850",
238 /* 0x0002 */ "ibm-851",
239 /* 0x0003 */ "windows-1255",
240 /* 0x0004 */ "windows-1256",
241 /* 0x0005 */ "windows-1251",
242 /* 0x0006 */ "ibm-852",
243 /* 0x0007 */ NULL, /* Unused */
244 /* 0x0008 */ "windows-1254",
245 /* 0x0009 */ NULL, /* Control char HT */
246 /* 0x000A */ NULL, /* Control char LF */
247 /* 0x000B */ "windows-874",
248 /* 0x000C */ NULL, /* Unused */
249 /* 0x000D */ NULL, /* Control char CR */
250 /* 0x000E */ NULL, /* Unused */
251 /* 0x000F */ NULL, /* Control chars: 0x0F20 + C0/C1 character: algorithmic */
252 /* 0x0010 */ "windows-932",
253 /* 0x0011 */ "windows-949",
254 /* 0x0012 */ "windows-950",
255 /* 0x0013 */ "windows-936"
256
257 /* The rest are null, including the 0x0014 Unicode compatibility region
258 and 0x0019, the 1-2-3 system range control char */
259 };
260
261
262 /* That's approximately all the data that's needed for translating
263 LMBCS to Unicode.
264
265
266 However, to translate Unicode to LMBCS, we need some more support.
267
268 That's because there are often more than one possible mappings from a Unicode
269 code point back into LMBCS. The first thing we do is look up into a table
270 to figure out if there are more than one possible mappings. This table,
271 arranged by Unicode values (including ranges) either lists which group
272 to use, or says that it could go into one or more of the SBCS sets, or
273 into one or more of the DBCS sets. (If the character exists in both DBCS &
274 SBCS, the table will place it in the SBCS sets, to make the LMBCS code point
275 length as small as possible. Here's the two special markers we use to indicate
276 ambiguous mappings: */
277
278 #define ULMBCS_AMBIGUOUS_SBCS 0x80 /* could fit in more than one
279 LMBCS sbcs native encoding
280 (example: most accented latin) */
281 #define ULMBCS_AMBIGUOUS_MBCS 0x81 /* could fit in more than one
282 LMBCS mbcs native encoding
283 (example: Unihan) */
284 #define ULMBCS_AMBIGUOUS_ALL 0x82
285 /* And here's a simple way to see if a group falls in an appropriate range */
286 #define ULMBCS_AMBIGUOUS_MATCH(agroup, xgroup) \
287 ((((agroup) == ULMBCS_AMBIGUOUS_SBCS) && \
288 (xgroup) < ULMBCS_DOUBLEOPTGROUP_START) || \
289 (((agroup) == ULMBCS_AMBIGUOUS_MBCS) && \
290 (xgroup) >= ULMBCS_DOUBLEOPTGROUP_START)) || \
291 ((agroup) == ULMBCS_AMBIGUOUS_ALL)
292
293
294 /* The table & some code to use it: */
295
296
297 static const struct _UniLMBCSGrpMap
298 {
299 const UChar uniStartRange;
300 const UChar uniEndRange;
301 const ulmbcs_byte_t GrpType;
302 } UniLMBCSGrpMap[]
303 =
304 {
305
306 {0x0001, 0x001F, ULMBCS_GRP_CTRL},
307 {0x0080, 0x009F, ULMBCS_GRP_CTRL},
308 {0x00A0, 0x00A6, ULMBCS_AMBIGUOUS_SBCS},
309 {0x00A7, 0x00A8, ULMBCS_AMBIGUOUS_ALL},
310 {0x00A9, 0x00AF, ULMBCS_AMBIGUOUS_SBCS},
311 {0x00B0, 0x00B1, ULMBCS_AMBIGUOUS_ALL},
312 {0x00B2, 0x00B3, ULMBCS_AMBIGUOUS_SBCS},
313 {0x00B4, 0x00B4, ULMBCS_AMBIGUOUS_ALL},
314 {0x00B5, 0x00B5, ULMBCS_AMBIGUOUS_SBCS},
315 {0x00B6, 0x00B6, ULMBCS_AMBIGUOUS_ALL},
316 {0x00B7, 0x00D6, ULMBCS_AMBIGUOUS_SBCS},
317 {0x00D7, 0x00D7, ULMBCS_AMBIGUOUS_ALL},
318 {0x00D8, 0x00F6, ULMBCS_AMBIGUOUS_SBCS},
319 {0x00F7, 0x00F7, ULMBCS_AMBIGUOUS_ALL},
320 {0x00F8, 0x01CD, ULMBCS_AMBIGUOUS_SBCS},
321 {0x01CE, 0x01CE, ULMBCS_GRP_TW },
322 {0x01CF, 0x02B9, ULMBCS_AMBIGUOUS_SBCS},
323 {0x02BA, 0x02BA, ULMBCS_GRP_CN},
324 {0x02BC, 0x02C8, ULMBCS_AMBIGUOUS_SBCS},
325 {0x02C9, 0x02D0, ULMBCS_AMBIGUOUS_MBCS},
326 {0x02D8, 0x02DD, ULMBCS_AMBIGUOUS_SBCS},
327 {0x0384, 0x0390, ULMBCS_AMBIGUOUS_SBCS},
328 {0x0391, 0x03A9, ULMBCS_AMBIGUOUS_ALL},
329 {0x03AA, 0x03B0, ULMBCS_AMBIGUOUS_SBCS},
330 {0x03B1, 0x03C9, ULMBCS_AMBIGUOUS_ALL},
331 {0x03CA, 0x03CE, ULMBCS_AMBIGUOUS_SBCS},
332 {0x0400, 0x0400, ULMBCS_GRP_RU},
333 {0x0401, 0x0401, ULMBCS_AMBIGUOUS_ALL},
334 {0x0402, 0x040F, ULMBCS_GRP_RU},
335 {0x0410, 0x0431, ULMBCS_AMBIGUOUS_ALL},
336 {0x0432, 0x044E, ULMBCS_GRP_RU},
337 {0x044F, 0x044F, ULMBCS_AMBIGUOUS_ALL},
338 {0x0450, 0x0491, ULMBCS_GRP_RU},
339 {0x05B0, 0x05F2, ULMBCS_GRP_HE},
340 {0x060C, 0x06AF, ULMBCS_GRP_AR},
341 {0x0E01, 0x0E5B, ULMBCS_GRP_TH},
342 {0x200C, 0x200F, ULMBCS_AMBIGUOUS_SBCS},
343 {0x2010, 0x2010, ULMBCS_AMBIGUOUS_MBCS},
344 {0x2013, 0x2014, ULMBCS_AMBIGUOUS_SBCS},
345 {0x2015, 0x2015, ULMBCS_AMBIGUOUS_MBCS},
346 {0x2016, 0x2016, ULMBCS_AMBIGUOUS_MBCS},
347 {0x2017, 0x2017, ULMBCS_AMBIGUOUS_SBCS},
348 {0x2018, 0x2019, ULMBCS_AMBIGUOUS_ALL},
349 {0x201A, 0x201B, ULMBCS_AMBIGUOUS_SBCS},
350 {0x201C, 0x201D, ULMBCS_AMBIGUOUS_ALL},
351 {0x201E, 0x201F, ULMBCS_AMBIGUOUS_SBCS},
352 {0x2020, 0x2021, ULMBCS_AMBIGUOUS_ALL},
353 {0x2022, 0x2024, ULMBCS_AMBIGUOUS_SBCS},
354 {0x2025, 0x2025, ULMBCS_AMBIGUOUS_MBCS},
355 {0x2026, 0x2026, ULMBCS_AMBIGUOUS_ALL},
356 {0x2027, 0x2027, ULMBCS_GRP_TW},
357 {0x2030, 0x2030, ULMBCS_AMBIGUOUS_ALL},
358 {0x2031, 0x2031, ULMBCS_AMBIGUOUS_SBCS},
359 {0x2032, 0x2033, ULMBCS_AMBIGUOUS_MBCS},
360 {0x2035, 0x2035, ULMBCS_AMBIGUOUS_MBCS},
361 {0x2039, 0x203A, ULMBCS_AMBIGUOUS_SBCS},
362 {0x203B, 0x203B, ULMBCS_AMBIGUOUS_MBCS},
363 {0x203C, 0x203C, ULMBCS_GRP_EXCEPT},
364 {0x2074, 0x2074, ULMBCS_GRP_KO},
365 {0x207F, 0x207F, ULMBCS_GRP_EXCEPT},
366 {0x2081, 0x2084, ULMBCS_GRP_KO},
367 {0x20A4, 0x20AC, ULMBCS_AMBIGUOUS_SBCS},
368 {0x2103, 0x2109, ULMBCS_AMBIGUOUS_MBCS},
369 {0x2111, 0x2120, ULMBCS_AMBIGUOUS_SBCS},
370 /*zhujin: upgrade, for regressiont test, spr HKIA4YHTSU*/
371 {0x2121, 0x2121, ULMBCS_AMBIGUOUS_MBCS},
372 {0x2122, 0x2126, ULMBCS_AMBIGUOUS_SBCS},
373 {0x212B, 0x212B, ULMBCS_AMBIGUOUS_MBCS},
374 {0x2135, 0x2135, ULMBCS_AMBIGUOUS_SBCS},
375 {0x2153, 0x2154, ULMBCS_GRP_KO},
376 {0x215B, 0x215E, ULMBCS_GRP_EXCEPT},
377 {0x2160, 0x2179, ULMBCS_AMBIGUOUS_MBCS},
378 {0x2190, 0x2193, ULMBCS_AMBIGUOUS_ALL},
379 {0x2194, 0x2195, ULMBCS_GRP_EXCEPT},
380 {0x2196, 0x2199, ULMBCS_AMBIGUOUS_MBCS},
381 {0x21A8, 0x21A8, ULMBCS_GRP_EXCEPT},
382 {0x21B8, 0x21B9, ULMBCS_GRP_CN},
383 {0x21D0, 0x21D1, ULMBCS_GRP_EXCEPT},
384 {0x21D2, 0x21D2, ULMBCS_AMBIGUOUS_MBCS},
385 {0x21D3, 0x21D3, ULMBCS_GRP_EXCEPT},
386 {0x21D4, 0x21D4, ULMBCS_AMBIGUOUS_MBCS},
387 {0x21D5, 0x21D5, ULMBCS_GRP_EXCEPT},
388 {0x21E7, 0x21E7, ULMBCS_GRP_CN},
389 {0x2200, 0x2200, ULMBCS_AMBIGUOUS_MBCS},
390 {0x2201, 0x2201, ULMBCS_GRP_EXCEPT},
391 {0x2202, 0x2202, ULMBCS_AMBIGUOUS_MBCS},
392 {0x2203, 0x2203, ULMBCS_AMBIGUOUS_MBCS},
393 {0x2204, 0x2206, ULMBCS_GRP_EXCEPT},
394 {0x2207, 0x2208, ULMBCS_AMBIGUOUS_MBCS},
395 {0x2209, 0x220A, ULMBCS_GRP_EXCEPT},
396 {0x220B, 0x220B, ULMBCS_AMBIGUOUS_MBCS},
397 {0x220F, 0x2215, ULMBCS_AMBIGUOUS_MBCS},
398 {0x2219, 0x2219, ULMBCS_GRP_EXCEPT},
399 {0x221A, 0x221A, ULMBCS_AMBIGUOUS_MBCS},
400 {0x221B, 0x221C, ULMBCS_GRP_EXCEPT},
401 {0x221D, 0x221E, ULMBCS_AMBIGUOUS_MBCS},
402 {0x221F, 0x221F, ULMBCS_GRP_EXCEPT},
403 {0x2220, 0x2220, ULMBCS_AMBIGUOUS_MBCS},
404 {0x2223, 0x222A, ULMBCS_AMBIGUOUS_MBCS},
405 {0x222B, 0x223D, ULMBCS_AMBIGUOUS_MBCS},
406 {0x2245, 0x2248, ULMBCS_GRP_EXCEPT},
407 {0x224C, 0x224C, ULMBCS_GRP_TW},
408 {0x2252, 0x2252, ULMBCS_AMBIGUOUS_MBCS},
409 {0x2260, 0x2261, ULMBCS_AMBIGUOUS_MBCS},
410 {0x2262, 0x2265, ULMBCS_GRP_EXCEPT},
411 {0x2266, 0x226F, ULMBCS_AMBIGUOUS_MBCS},
412 {0x2282, 0x2283, ULMBCS_AMBIGUOUS_MBCS},
413 {0x2284, 0x2285, ULMBCS_GRP_EXCEPT},
414 {0x2286, 0x2287, ULMBCS_AMBIGUOUS_MBCS},
415 {0x2288, 0x2297, ULMBCS_GRP_EXCEPT},
416 {0x2299, 0x22BF, ULMBCS_AMBIGUOUS_MBCS},
417 {0x22C0, 0x22C0, ULMBCS_GRP_EXCEPT},
418 {0x2310, 0x2310, ULMBCS_GRP_EXCEPT},
419 {0x2312, 0x2312, ULMBCS_AMBIGUOUS_MBCS},
420 {0x2318, 0x2321, ULMBCS_GRP_EXCEPT},
421 {0x2318, 0x2321, ULMBCS_GRP_CN},
422 {0x2460, 0x24E9, ULMBCS_AMBIGUOUS_MBCS},
423 {0x2500, 0x2500, ULMBCS_AMBIGUOUS_SBCS},
424 {0x2501, 0x2501, ULMBCS_AMBIGUOUS_MBCS},
425 {0x2502, 0x2502, ULMBCS_AMBIGUOUS_ALL},
426 {0x2503, 0x2503, ULMBCS_AMBIGUOUS_MBCS},
427 {0x2504, 0x2505, ULMBCS_GRP_TW},
428 {0x2506, 0x2665, ULMBCS_AMBIGUOUS_ALL},
429 {0x2666, 0x2666, ULMBCS_GRP_EXCEPT},
430 {0x2667, 0x2669, ULMBCS_AMBIGUOUS_SBCS},
431 {0x266A, 0x266A, ULMBCS_AMBIGUOUS_ALL},
432 {0x266B, 0x266C, ULMBCS_AMBIGUOUS_SBCS},
433 {0x266D, 0x266D, ULMBCS_AMBIGUOUS_MBCS},
434 {0x266E, 0x266E, ULMBCS_AMBIGUOUS_SBCS},
435 {0x266F, 0x266F, ULMBCS_GRP_JA},
436 {0x2670, 0x2E7F, ULMBCS_AMBIGUOUS_SBCS},
437 {0x2E80, 0xF861, ULMBCS_AMBIGUOUS_MBCS},
438 {0xF862, 0xF8FF, ULMBCS_GRP_EXCEPT},
439 {0xF900, 0xFA2D, ULMBCS_AMBIGUOUS_MBCS},
440 {0xFB00, 0xFEFF, ULMBCS_AMBIGUOUS_SBCS},
441 {0xFF01, 0xFFEE, ULMBCS_AMBIGUOUS_MBCS},
442 {0xFFFF, 0xFFFF, ULMBCS_GRP_UNICODE}
443 };
444
445 static ulmbcs_byte_t
FindLMBCSUniRange(UChar uniChar)446 FindLMBCSUniRange(UChar uniChar)
447 {
448 const struct _UniLMBCSGrpMap * pTable = UniLMBCSGrpMap;
449
450 while (uniChar > pTable->uniEndRange)
451 {
452 pTable++;
453 }
454
455 if (uniChar >= pTable->uniStartRange)
456 {
457 return pTable->GrpType;
458 }
459 return ULMBCS_GRP_UNICODE;
460 }
461
462 /*
463 We also ask the creator of a converter to send in a preferred locale
464 that we can use in resolving ambiguous mappings. They send the locale
465 in as a string, and we map it, if possible, to one of the
466 LMBCS groups. We use this table, and the associated code, to
467 do the lookup: */
468
469 /**************************************************
470 This table maps locale ID's to LMBCS opt groups.
471 The default return is group 0x01. Note that for
472 performance reasons, the table is sorted in
473 increasing alphabetic order, with the notable
474 exception of zhTW. This is to force the check
475 for Traditonal Chinese before dropping back to
476 Simplified.
477
478 Note too that the Latin-1 groups have been
479 commented out because it's the default, and
480 this shortens the table, allowing a serial
481 search to go quickly.
482 *************************************************/
483
484 static const struct _LocaleLMBCSGrpMap
485 {
486 const char *LocaleID;
487 const ulmbcs_byte_t OptGroup;
488 } LocaleLMBCSGrpMap[] =
489 {
490 {"ar", ULMBCS_GRP_AR},
491 {"be", ULMBCS_GRP_RU},
492 {"bg", ULMBCS_GRP_L2},
493 /* {"ca", ULMBCS_GRP_L1}, */
494 {"cs", ULMBCS_GRP_L2},
495 /* {"da", ULMBCS_GRP_L1}, */
496 /* {"de", ULMBCS_GRP_L1}, */
497 {"el", ULMBCS_GRP_GR},
498 /* {"en", ULMBCS_GRP_L1}, */
499 /* {"es", ULMBCS_GRP_L1}, */
500 /* {"et", ULMBCS_GRP_L1}, */
501 /* {"fi", ULMBCS_GRP_L1}, */
502 /* {"fr", ULMBCS_GRP_L1}, */
503 {"he", ULMBCS_GRP_HE},
504 {"hu", ULMBCS_GRP_L2},
505 /* {"is", ULMBCS_GRP_L1}, */
506 /* {"it", ULMBCS_GRP_L1}, */
507 {"iw", ULMBCS_GRP_HE},
508 {"ja", ULMBCS_GRP_JA},
509 {"ko", ULMBCS_GRP_KO},
510 /* {"lt", ULMBCS_GRP_L1}, */
511 /* {"lv", ULMBCS_GRP_L1}, */
512 {"mk", ULMBCS_GRP_RU},
513 /* {"nl", ULMBCS_GRP_L1}, */
514 /* {"no", ULMBCS_GRP_L1}, */
515 {"pl", ULMBCS_GRP_L2},
516 /* {"pt", ULMBCS_GRP_L1}, */
517 {"ro", ULMBCS_GRP_L2},
518 {"ru", ULMBCS_GRP_RU},
519 {"sh", ULMBCS_GRP_L2},
520 {"sk", ULMBCS_GRP_L2},
521 {"sl", ULMBCS_GRP_L2},
522 {"sq", ULMBCS_GRP_L2},
523 {"sr", ULMBCS_GRP_RU},
524 /* {"sv", ULMBCS_GRP_L1}, */
525 {"th", ULMBCS_GRP_TH},
526 {"tr", ULMBCS_GRP_TR},
527 {"uk", ULMBCS_GRP_RU},
528 /* {"vi", ULMBCS_GRP_L1}, */
529 {"zhTW", ULMBCS_GRP_TW},
530 {"zh", ULMBCS_GRP_CN},
531 {NULL, ULMBCS_GRP_L1}
532 };
533
534
535 static ulmbcs_byte_t
FindLMBCSLocale(const char * LocaleID)536 FindLMBCSLocale(const char *LocaleID)
537 {
538 const struct _LocaleLMBCSGrpMap *pTable = LocaleLMBCSGrpMap;
539
540 if ((!LocaleID) || (!*LocaleID))
541 {
542 return 0;
543 }
544
545 while (pTable->LocaleID)
546 {
547 if (*pTable->LocaleID == *LocaleID) /* Check only first char for speed */
548 {
549 /* First char matches - check whole name, for entry-length */
550 if (uprv_strncmp(pTable->LocaleID, LocaleID, strlen(pTable->LocaleID)) == 0)
551 return pTable->OptGroup;
552 }
553 else
554 if (*pTable->LocaleID > *LocaleID) /* Sorted alphabetically - exit */
555 break;
556 pTable++;
557 }
558 return ULMBCS_GRP_L1;
559 }
560
561
562 /*
563 Before we get to the main body of code, here's how we hook up to the rest
564 of ICU. ICU converters are required to define a structure that includes
565 some function pointers, and some common data, in the style of a C++
566 vtable. There is also room in there for converter-specific data. LMBCS
567 uses that converter-specific data to keep track of the 12 subconverters
568 we use, the optimization group, and the group (if any) that matches the
569 locale. We have one structure instantiated for each of the 12 possible
570 optimization groups. To avoid typos & to avoid boring the reader, we
571 put the declarations of these structures and functions into macros. To see
572 the definitions of these structures, see unicode\ucnv_bld.h
573 */
574
575 typedef struct
576 {
577 UConverterSharedData *OptGrpConverter[ULMBCS_GRP_LAST+1]; /* Converter per Opt. grp. */
578 uint8_t OptGroup; /* default Opt. grp. for this LMBCS session */
579 uint8_t localeConverterIndex; /* reasonable locale match for index */
580 }
581 UConverterDataLMBCS;
582
583 static void _LMBCSClose(UConverter * _this);
584
585 #define DECLARE_LMBCS_DATA(n) \
586 static const UConverterImpl _LMBCSImpl##n={\
587 UCNV_LMBCS_##n,\
588 NULL,NULL,\
589 _LMBCSOpen##n,\
590 _LMBCSClose,\
591 NULL,\
592 _LMBCSToUnicodeWithOffsets,\
593 _LMBCSToUnicodeWithOffsets,\
594 _LMBCSFromUnicode,\
595 _LMBCSFromUnicode,\
596 NULL,\
597 NULL,\
598 NULL,\
599 NULL,\
600 _LMBCSSafeClone,\
601 ucnv_getCompleteUnicodeSet\
602 };\
603 static const UConverterStaticData _LMBCSStaticData##n={\
604 sizeof(UConverterStaticData),\
605 "LMBCS-" #n,\
606 0, UCNV_IBM, UCNV_LMBCS_##n, 1, 3,\
607 { 0x3f, 0, 0, 0 },1,FALSE,FALSE,0,0,{0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0} \
608 };\
609 const UConverterSharedData _LMBCSData##n={\
610 sizeof(UConverterSharedData), ~((uint32_t) 0),\
611 NULL, NULL, &_LMBCSStaticData##n, FALSE, &_LMBCSImpl##n, \
612 0 \
613 };
614
615 /* The only function we needed to duplicate 12 times was the 'open'
616 function, which will do basically the same thing except set a different
617 optimization group. So, we put the common stuff into a worker function,
618 and set up another macro to stamp out the 12 open functions:*/
619 #define DEFINE_LMBCS_OPEN(n) \
620 static void \
621 _LMBCSOpen##n(UConverter* _this, UConverterLoadArgs* pArgs, UErrorCode* err) \
622 { _LMBCSOpenWorker(_this, pArgs, err, n); }
623
624
625
626 /* Here's the open worker & the common close function */
627 static void
_LMBCSOpenWorker(UConverter * _this,UConverterLoadArgs * pArgs,UErrorCode * err,ulmbcs_byte_t OptGroup)628 _LMBCSOpenWorker(UConverter* _this,
629 UConverterLoadArgs *pArgs,
630 UErrorCode* err,
631 ulmbcs_byte_t OptGroup)
632 {
633 UConverterDataLMBCS * extraInfo = _this->extraInfo =
634 (UConverterDataLMBCS*)uprv_malloc (sizeof (UConverterDataLMBCS));
635 if(extraInfo != NULL)
636 {
637 UConverterNamePieces stackPieces;
638 UConverterLoadArgs stackArgs={ (int32_t)sizeof(UConverterLoadArgs) };
639 ulmbcs_byte_t i;
640
641 uprv_memset(extraInfo, 0, sizeof(UConverterDataLMBCS));
642
643 stackArgs.onlyTestIsLoadable = pArgs->onlyTestIsLoadable;
644
645 for (i=0; i <= ULMBCS_GRP_LAST && U_SUCCESS(*err); i++)
646 {
647 if(OptGroupByteToCPName[i] != NULL) {
648 extraInfo->OptGrpConverter[i] = ucnv_loadSharedData(OptGroupByteToCPName[i], &stackPieces, &stackArgs, err);
649 }
650 }
651
652 if(U_FAILURE(*err) || pArgs->onlyTestIsLoadable) {
653 _LMBCSClose(_this);
654 return;
655 }
656 extraInfo->OptGroup = OptGroup;
657 extraInfo->localeConverterIndex = FindLMBCSLocale(pArgs->locale);
658 }
659 else
660 {
661 *err = U_MEMORY_ALLOCATION_ERROR;
662 }
663 }
664
665 static void
_LMBCSClose(UConverter * _this)666 _LMBCSClose(UConverter * _this)
667 {
668 if (_this->extraInfo != NULL)
669 {
670 ulmbcs_byte_t Ix;
671 UConverterDataLMBCS * extraInfo = (UConverterDataLMBCS *) _this->extraInfo;
672
673 for (Ix=0; Ix <= ULMBCS_GRP_LAST; Ix++)
674 {
675 if (extraInfo->OptGrpConverter[Ix] != NULL)
676 ucnv_unloadSharedDataIfReady(extraInfo->OptGrpConverter[Ix]);
677 }
678 if (!_this->isExtraLocal) {
679 uprv_free (_this->extraInfo);
680 _this->extraInfo = NULL;
681 }
682 }
683 }
684
685 typedef struct LMBCSClone {
686 UConverter cnv;
687 UConverterDataLMBCS lmbcs;
688 } LMBCSClone;
689
690 static UConverter *
_LMBCSSafeClone(const UConverter * cnv,void * stackBuffer,int32_t * pBufferSize,UErrorCode * status)691 _LMBCSSafeClone(const UConverter *cnv,
692 void *stackBuffer,
693 int32_t *pBufferSize,
694 UErrorCode *status) {
695 LMBCSClone *newLMBCS;
696 UConverterDataLMBCS *extraInfo;
697 int32_t i;
698
699 if(*pBufferSize<=0) {
700 *pBufferSize=(int32_t)sizeof(LMBCSClone);
701 return NULL;
702 }
703
704 extraInfo=(UConverterDataLMBCS *)cnv->extraInfo;
705 newLMBCS=(LMBCSClone *)stackBuffer;
706
707 /* ucnv.c/ucnv_safeClone() copied the main UConverter already */
708
709 uprv_memcpy(&newLMBCS->lmbcs, extraInfo, sizeof(UConverterDataLMBCS));
710
711 /* share the subconverters */
712 for(i = 0; i <= ULMBCS_GRP_LAST; ++i) {
713 if(extraInfo->OptGrpConverter[i] != NULL) {
714 ucnv_incrementRefCount(extraInfo->OptGrpConverter[i]);
715 }
716 }
717
718 newLMBCS->cnv.extraInfo = &newLMBCS->lmbcs;
719 newLMBCS->cnv.isExtraLocal = TRUE;
720 return &newLMBCS->cnv;
721 }
722
723 /*
724 * There used to be a _LMBCSGetUnicodeSet() function here (up to svn revision 20117)
725 * which added all code points except for U+F6xx
726 * because those cannot be represented in the Unicode group.
727 * However, it turns out that windows-950 has roundtrips for all of U+F6xx
728 * which means that LMBCS can convert all Unicode code points after all.
729 * We now simply use ucnv_getCompleteUnicodeSet().
730 *
731 * This may need to be looked at again as Lotus uses _LMBCSGetUnicodeSet(). (091216)
732 */
733
734 /*
735 Here's the basic helper function that we use when converting from
736 Unicode to LMBCS, and we suspect that a Unicode character will fit into
737 one of the 12 groups. The return value is the number of bytes written
738 starting at pStartLMBCS (if any).
739 */
740
741 static size_t
LMBCSConversionWorker(UConverterDataLMBCS * extraInfo,ulmbcs_byte_t group,ulmbcs_byte_t * pStartLMBCS,UChar * pUniChar,ulmbcs_byte_t * lastConverterIndex,UBool * groups_tried)742 LMBCSConversionWorker (
743 UConverterDataLMBCS * extraInfo, /* subconverters, opt & locale groups */
744 ulmbcs_byte_t group, /* The group to try */
745 ulmbcs_byte_t * pStartLMBCS, /* where to put the results */
746 UChar * pUniChar, /* The input unicode character */
747 ulmbcs_byte_t * lastConverterIndex, /* output: track last successful group used */
748 UBool * groups_tried /* output: track any unsuccessful groups */
749 )
750 {
751 ulmbcs_byte_t * pLMBCS = pStartLMBCS;
752 UConverterSharedData * xcnv = extraInfo->OptGrpConverter[group];
753
754 int bytesConverted;
755 uint32_t value;
756 ulmbcs_byte_t firstByte;
757
758 U_ASSERT(xcnv);
759 U_ASSERT(group<ULMBCS_GRP_UNICODE);
760
761 bytesConverted = ucnv_MBCSFromUChar32(xcnv, *pUniChar, &value, FALSE);
762
763 /* get the first result byte */
764 if(bytesConverted > 0) {
765 firstByte = (ulmbcs_byte_t)(value >> ((bytesConverted - 1) * 8));
766 } else {
767 /* most common failure mode is an unassigned character */
768 groups_tried[group] = TRUE;
769 return 0;
770 }
771
772 *lastConverterIndex = group;
773
774 /* All initial byte values in lower ascii range should have been caught by now,
775 except with the exception group.
776 */
777 U_ASSERT((firstByte <= ULMBCS_C0END) || (firstByte >= ULMBCS_C1START) || (group == ULMBCS_GRP_EXCEPT));
778
779 /* use converted data: first write 0, 1 or two group bytes */
780 if (group != ULMBCS_GRP_EXCEPT && extraInfo->OptGroup != group)
781 {
782 *pLMBCS++ = group;
783 if (bytesConverted == 1 && group >= ULMBCS_DOUBLEOPTGROUP_START)
784 {
785 *pLMBCS++ = group;
786 }
787 }
788
789 /* don't emit control chars */
790 if ( bytesConverted == 1 && firstByte < 0x20 )
791 return 0;
792
793
794 /* then move over the converted data */
795 switch(bytesConverted)
796 {
797 case 4:
798 *pLMBCS++ = (ulmbcs_byte_t)(value >> 24);
799 case 3: /*fall through*/
800 *pLMBCS++ = (ulmbcs_byte_t)(value >> 16);
801 case 2: /*fall through*/
802 *pLMBCS++ = (ulmbcs_byte_t)(value >> 8);
803 case 1: /*fall through*/
804 *pLMBCS++ = (ulmbcs_byte_t)value;
805 default:
806 /* will never occur */
807 break;
808 }
809
810 return (pLMBCS - pStartLMBCS);
811 }
812
813
814 /* This is a much simpler version of above, when we
815 know we are writing LMBCS using the Unicode group
816 */
817 static size_t
LMBCSConvertUni(ulmbcs_byte_t * pLMBCS,UChar uniChar)818 LMBCSConvertUni(ulmbcs_byte_t * pLMBCS, UChar uniChar)
819 {
820 /* encode into LMBCS Unicode range */
821 uint8_t LowCh = (uint8_t)(uniChar & 0x00FF);
822 uint8_t HighCh = (uint8_t)(uniChar >> 8);
823
824 *pLMBCS++ = ULMBCS_GRP_UNICODE;
825
826 if (LowCh == 0)
827 {
828 *pLMBCS++ = ULMBCS_UNICOMPATZERO;
829 *pLMBCS++ = HighCh;
830 }
831 else
832 {
833 *pLMBCS++ = HighCh;
834 *pLMBCS++ = LowCh;
835 }
836 return ULMBCS_UNICODE_SIZE;
837 }
838
839
840
841 /* The main Unicode to LMBCS conversion function */
842 static void
_LMBCSFromUnicode(UConverterFromUnicodeArgs * args,UErrorCode * err)843 _LMBCSFromUnicode(UConverterFromUnicodeArgs* args,
844 UErrorCode* err)
845 {
846 ulmbcs_byte_t lastConverterIndex = 0;
847 UChar uniChar;
848 ulmbcs_byte_t LMBCS[ULMBCS_CHARSIZE_MAX];
849 ulmbcs_byte_t * pLMBCS;
850 int32_t bytes_written;
851 UBool groups_tried[ULMBCS_GRP_LAST+1];
852 UConverterDataLMBCS * extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo;
853 int sourceIndex = 0;
854
855 /* Basic strategy: attempt to fill in local LMBCS 1-char buffer.(LMBCS)
856 If that succeeds, see if it will all fit into the target & copy it over
857 if it does.
858
859 We try conversions in the following order:
860
861 1. Single-byte ascii & special fixed control chars (&null)
862 2. Look up group in table & try that (could be
863 A) Unicode group
864 B) control group,
865 C) national encoding,
866 or ambiguous SBCS or MBCS group (on to step 4...)
867
868 3. If its ambiguous, try this order:
869 A) The optimization group
870 B) The locale group
871 C) The last group that succeeded with this string.
872 D) every other group that's relevent (single or double)
873 E) If its single-byte ambiguous, try the exceptions group
874
875 4. And as a grand fallback: Unicode
876 */
877
878 /*Fix for SPR#DJOE66JFN3 (Lotus)*/
879 ulmbcs_byte_t OldConverterIndex = 0;
880
881 while (args->source < args->sourceLimit && !U_FAILURE(*err))
882 {
883 /*Fix for SPR#DJOE66JFN3 (Lotus)*/
884 OldConverterIndex = extraInfo->localeConverterIndex;
885
886 if (args->target >= args->targetLimit)
887 {
888 *err = U_BUFFER_OVERFLOW_ERROR;
889 break;
890 }
891 uniChar = *(args->source);
892 bytes_written = 0;
893 pLMBCS = LMBCS;
894
895 /* check cases in rough order of how common they are, for speed */
896
897 /* single byte matches: strategy 1 */
898 /*Fix for SPR#DJOE66JFN3 (Lotus)*/
899 if((uniChar>=0x80) && (uniChar<=0xff)
900 /*Fix for SPR#JUYA6XAERU and TSAO7GL5NK (Lotus)*/ &&(uniChar!=0xB1) &&(uniChar!=0xD7) &&(uniChar!=0xF7)
901 &&(uniChar!=0xB0) &&(uniChar!=0xB4) &&(uniChar!=0xB6) &&(uniChar!=0xA7) &&(uniChar!=0xA8))
902 {
903 extraInfo->localeConverterIndex = ULMBCS_GRP_L1;
904 }
905 if (((uniChar > ULMBCS_C0END) && (uniChar < ULMBCS_C1START)) ||
906 uniChar == 0 || uniChar == ULMBCS_HT || uniChar == ULMBCS_CR ||
907 uniChar == ULMBCS_LF || uniChar == ULMBCS_123SYSTEMRANGE
908 )
909 {
910 *pLMBCS++ = (ulmbcs_byte_t ) uniChar;
911 bytes_written = 1;
912 }
913
914
915 if (!bytes_written)
916 {
917 /* Check by UNICODE range (Strategy 2) */
918 ulmbcs_byte_t group = FindLMBCSUniRange(uniChar);
919
920 if (group == ULMBCS_GRP_UNICODE) /* (Strategy 2A) */
921 {
922 pLMBCS += LMBCSConvertUni(pLMBCS,uniChar);
923
924 bytes_written = (int32_t)(pLMBCS - LMBCS);
925 }
926 else if (group == ULMBCS_GRP_CTRL) /* (Strategy 2B) */
927 {
928 /* Handle control characters here */
929 if (uniChar <= ULMBCS_C0END)
930 {
931 *pLMBCS++ = ULMBCS_GRP_CTRL;
932 *pLMBCS++ = (ulmbcs_byte_t)(ULMBCS_CTRLOFFSET + uniChar);
933 }
934 else if (uniChar >= ULMBCS_C1START && uniChar <= ULMBCS_C1START + ULMBCS_CTRLOFFSET)
935 {
936 *pLMBCS++ = ULMBCS_GRP_CTRL;
937 *pLMBCS++ = (ulmbcs_byte_t ) (uniChar & 0x00FF);
938 }
939 bytes_written = (int32_t)(pLMBCS - LMBCS);
940 }
941 else if (group < ULMBCS_GRP_UNICODE) /* (Strategy 2C) */
942 {
943 /* a specific converter has been identified - use it */
944 bytes_written = (int32_t)LMBCSConversionWorker (
945 extraInfo, group, pLMBCS, &uniChar,
946 &lastConverterIndex, groups_tried);
947 }
948 if (!bytes_written) /* the ambiguous group cases (Strategy 3) */
949 {
950 uprv_memset(groups_tried, 0, sizeof(groups_tried));
951
952 /* check for non-default optimization group (Strategy 3A )*/
953 if ((extraInfo->OptGroup != 1) && (ULMBCS_AMBIGUOUS_MATCH(group, extraInfo->OptGroup)))
954 {
955 /*zhujin: upgrade, merge #39299 here (Lotus) */
956 /*To make R5 compatible translation, look for exceptional group first for non-DBCS*/
957
958 if(extraInfo->localeConverterIndex < ULMBCS_DOUBLEOPTGROUP_START)
959 {
960 bytes_written = LMBCSConversionWorker (extraInfo,
961 ULMBCS_GRP_L1, pLMBCS, &uniChar,
962 &lastConverterIndex, groups_tried);
963
964 if(!bytes_written)
965 {
966 bytes_written = LMBCSConversionWorker (extraInfo,
967 ULMBCS_GRP_EXCEPT, pLMBCS, &uniChar,
968 &lastConverterIndex, groups_tried);
969 }
970 if(!bytes_written)
971 {
972 bytes_written = LMBCSConversionWorker (extraInfo,
973 extraInfo->localeConverterIndex, pLMBCS, &uniChar,
974 &lastConverterIndex, groups_tried);
975 }
976 }
977 else
978 {
979 bytes_written = LMBCSConversionWorker (extraInfo,
980 extraInfo->localeConverterIndex, pLMBCS, &uniChar,
981 &lastConverterIndex, groups_tried);
982 }
983 }
984 /* check for locale optimization group (Strategy 3B) */
985 if (!bytes_written && (extraInfo->localeConverterIndex) && (ULMBCS_AMBIGUOUS_MATCH(group, extraInfo->localeConverterIndex)))
986 {
987 bytes_written = (int32_t)LMBCSConversionWorker (extraInfo,
988 extraInfo->localeConverterIndex, pLMBCS, &uniChar, &lastConverterIndex, groups_tried);
989 }
990 /* check for last optimization group used for this string (Strategy 3C) */
991 if (!bytes_written && (lastConverterIndex) && (ULMBCS_AMBIGUOUS_MATCH(group, lastConverterIndex)))
992 {
993 bytes_written = (int32_t)LMBCSConversionWorker (extraInfo,
994 lastConverterIndex, pLMBCS, &uniChar, &lastConverterIndex, groups_tried);
995 }
996 if (!bytes_written)
997 {
998 /* just check every possible matching converter (Strategy 3D) */
999 ulmbcs_byte_t grp_start;
1000 ulmbcs_byte_t grp_end;
1001 ulmbcs_byte_t grp_ix;
1002 grp_start = (ulmbcs_byte_t)((group == ULMBCS_AMBIGUOUS_MBCS)
1003 ? ULMBCS_DOUBLEOPTGROUP_START
1004 : ULMBCS_GRP_L1);
1005 grp_end = (ulmbcs_byte_t)((group == ULMBCS_AMBIGUOUS_MBCS)
1006 ? ULMBCS_GRP_LAST
1007 : ULMBCS_GRP_TH);
1008 if(group == ULMBCS_AMBIGUOUS_ALL)
1009 {
1010 grp_start = ULMBCS_GRP_L1;
1011 grp_end = ULMBCS_GRP_LAST;
1012 }
1013 for (grp_ix = grp_start;
1014 grp_ix <= grp_end && !bytes_written;
1015 grp_ix++)
1016 {
1017 if (extraInfo->OptGrpConverter [grp_ix] && !groups_tried [grp_ix])
1018 {
1019 bytes_written = (int32_t)LMBCSConversionWorker (extraInfo,
1020 grp_ix, pLMBCS, &uniChar,
1021 &lastConverterIndex, groups_tried);
1022 }
1023 }
1024 /* a final conversion fallback to the exceptions group if its likely
1025 to be single byte (Strategy 3E) */
1026 if (!bytes_written && grp_start == ULMBCS_GRP_L1)
1027 {
1028 bytes_written = (int32_t)LMBCSConversionWorker (extraInfo,
1029 ULMBCS_GRP_EXCEPT, pLMBCS, &uniChar,
1030 &lastConverterIndex, groups_tried);
1031 }
1032 }
1033 /* all of our other strategies failed. Fallback to Unicode. (Strategy 4)*/
1034 if (!bytes_written)
1035 {
1036
1037 pLMBCS += LMBCSConvertUni(pLMBCS, uniChar);
1038 bytes_written = (int32_t)(pLMBCS - LMBCS);
1039 }
1040 }
1041 }
1042
1043 /* we have a translation. increment source and write as much as posible to target */
1044 args->source++;
1045 pLMBCS = LMBCS;
1046 while (args->target < args->targetLimit && bytes_written--)
1047 {
1048 *(args->target)++ = *pLMBCS++;
1049 if (args->offsets)
1050 {
1051 *(args->offsets)++ = sourceIndex;
1052 }
1053 }
1054 sourceIndex++;
1055 if (bytes_written > 0)
1056 {
1057 /* write any bytes that didn't fit in target to the error buffer,
1058 common code will move this to target if we get called back with
1059 enough target room
1060 */
1061 uint8_t * pErrorBuffer = args->converter->charErrorBuffer;
1062 *err = U_BUFFER_OVERFLOW_ERROR;
1063 args->converter->charErrorBufferLength = (int8_t)bytes_written;
1064 while (bytes_written--)
1065 {
1066 *pErrorBuffer++ = *pLMBCS++;
1067 }
1068 }
1069 /*Fix for SPR#DJOE66JFN3 (Lotus)*/
1070 extraInfo->localeConverterIndex = OldConverterIndex;
1071 }
1072 }
1073
1074
1075 /* Now, the Unicode from LMBCS section */
1076
1077
1078 /* A function to call when we are looking at the Unicode group byte in LMBCS */
1079 static UChar
GetUniFromLMBCSUni(char const ** ppLMBCSin)1080 GetUniFromLMBCSUni(char const ** ppLMBCSin) /* Called with LMBCS-style Unicode byte stream */
1081 {
1082 uint8_t HighCh = *(*ppLMBCSin)++; /* Big-endian Unicode in LMBCS compatibility group*/
1083 uint8_t LowCh = *(*ppLMBCSin)++;
1084
1085 if (HighCh == ULMBCS_UNICOMPATZERO )
1086 {
1087 HighCh = LowCh;
1088 LowCh = 0; /* zero-byte in LSB special character */
1089 }
1090 return (UChar)((HighCh << 8) | LowCh);
1091 }
1092
1093
1094
1095 /* CHECK_SOURCE_LIMIT: Helper macro to verify that there are at least'index'
1096 bytes left in source up to sourceLimit.Errors appropriately if not.
1097 If we reach the limit, then update the source pointer to there to consume
1098 all input as required by ICU converter semantics.
1099 */
1100
1101 #define CHECK_SOURCE_LIMIT(index) \
1102 if (args->source+index > args->sourceLimit){\
1103 *err = U_TRUNCATED_CHAR_FOUND;\
1104 args->source = args->sourceLimit;\
1105 return 0xffff;}
1106
1107 /* Return the Unicode representation for the current LMBCS character */
1108
1109 static UChar32
_LMBCSGetNextUCharWorker(UConverterToUnicodeArgs * args,UErrorCode * err)1110 _LMBCSGetNextUCharWorker(UConverterToUnicodeArgs* args,
1111 UErrorCode* err)
1112 {
1113 UChar32 uniChar = 0; /* an output UNICODE char */
1114 ulmbcs_byte_t CurByte; /* A byte from the input stream */
1115
1116 /* error check */
1117 if (args->source >= args->sourceLimit)
1118 {
1119 *err = U_ILLEGAL_ARGUMENT_ERROR;
1120 return 0xffff;
1121 }
1122 /* Grab first byte & save address for error recovery */
1123 CurByte = *((ulmbcs_byte_t *) (args->source++));
1124
1125 /*
1126 * at entry of each if clause:
1127 * 1. 'CurByte' points at the first byte of a LMBCS character
1128 * 2. '*source'points to the next byte of the source stream after 'CurByte'
1129 *
1130 * the job of each if clause is:
1131 * 1. set '*source' to point at the beginning of next char (nop if LMBCS char is only 1 byte)
1132 * 2. set 'uniChar' up with the right Unicode value, or set 'err' appropriately
1133 */
1134
1135 /* First lets check the simple fixed values. */
1136
1137 if(((CurByte > ULMBCS_C0END) && (CurByte < ULMBCS_C1START)) /* ascii range */
1138 || (CurByte == 0)
1139 || CurByte == ULMBCS_HT || CurByte == ULMBCS_CR
1140 || CurByte == ULMBCS_LF || CurByte == ULMBCS_123SYSTEMRANGE)
1141 {
1142 uniChar = CurByte;
1143 }
1144 else
1145 {
1146 UConverterDataLMBCS * extraInfo;
1147 ulmbcs_byte_t group;
1148 UConverterSharedData *cnv;
1149
1150 if (CurByte == ULMBCS_GRP_CTRL) /* Control character group - no opt group update */
1151 {
1152 ulmbcs_byte_t C0C1byte;
1153 CHECK_SOURCE_LIMIT(1);
1154 C0C1byte = *(args->source)++;
1155 uniChar = (C0C1byte < ULMBCS_C1START) ? C0C1byte - ULMBCS_CTRLOFFSET : C0C1byte;
1156 }
1157 else
1158 if (CurByte == ULMBCS_GRP_UNICODE) /* Unicode compatibility group: BigEndian UTF16 */
1159 {
1160 CHECK_SOURCE_LIMIT(2);
1161
1162 /* don't check for error indicators fffe/ffff below */
1163 return GetUniFromLMBCSUni(&(args->source));
1164 }
1165 else if (CurByte <= ULMBCS_CTRLOFFSET)
1166 {
1167 group = CurByte; /* group byte is in the source */
1168 extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo;
1169 if (group > ULMBCS_GRP_LAST || (cnv = extraInfo->OptGrpConverter[group]) == NULL)
1170 {
1171 /* this is not a valid group byte - no converter*/
1172 *err = U_INVALID_CHAR_FOUND;
1173 }
1174 else if (group >= ULMBCS_DOUBLEOPTGROUP_START) /* double byte conversion */
1175 {
1176
1177 CHECK_SOURCE_LIMIT(2);
1178
1179 /* check for LMBCS doubled-group-byte case */
1180 if (*args->source == group) {
1181 /* single byte */
1182 ++args->source;
1183 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source, 1, FALSE);
1184 ++args->source;
1185 } else {
1186 /* double byte */
1187 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source, 2, FALSE);
1188 args->source += 2;
1189 }
1190 }
1191 else { /* single byte conversion */
1192 CHECK_SOURCE_LIMIT(1);
1193 CurByte = *(args->source)++;
1194
1195 if (CurByte >= ULMBCS_C1START)
1196 {
1197 uniChar = _MBCS_SINGLE_SIMPLE_GET_NEXT_BMP(cnv, CurByte);
1198 }
1199 else
1200 {
1201 /* The non-optimizable oddballs where there is an explicit byte
1202 * AND the second byte is not in the upper ascii range
1203 */
1204 char bytes[2];
1205
1206 extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo;
1207 cnv = extraInfo->OptGrpConverter [ULMBCS_GRP_EXCEPT];
1208
1209 /* Lookup value must include opt group */
1210 bytes[0] = group;
1211 bytes[1] = CurByte;
1212 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, bytes, 2, FALSE);
1213 }
1214 }
1215 }
1216 else if (CurByte >= ULMBCS_C1START) /* group byte is implicit */
1217 {
1218 extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo;
1219 group = extraInfo->OptGroup;
1220 cnv = extraInfo->OptGrpConverter[group];
1221 if (group >= ULMBCS_DOUBLEOPTGROUP_START) /* double byte conversion */
1222 {
1223 if (!ucnv_MBCSIsLeadByte(cnv, CurByte))
1224 {
1225 CHECK_SOURCE_LIMIT(0);
1226
1227 /* let the MBCS conversion consume CurByte again */
1228 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source - 1, 1, FALSE);
1229 }
1230 else
1231 {
1232 CHECK_SOURCE_LIMIT(1);
1233 /* let the MBCS conversion consume CurByte again */
1234 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source - 1, 2, FALSE);
1235 ++args->source;
1236 }
1237 }
1238 else /* single byte conversion */
1239 {
1240 uniChar = _MBCS_SINGLE_SIMPLE_GET_NEXT_BMP(cnv, CurByte);
1241 }
1242 }
1243 }
1244 return uniChar;
1245 }
1246
1247
1248 /* The exported function that converts lmbcs to one or more
1249 UChars - currently UTF-16
1250 */
1251 static void
_LMBCSToUnicodeWithOffsets(UConverterToUnicodeArgs * args,UErrorCode * err)1252 _LMBCSToUnicodeWithOffsets(UConverterToUnicodeArgs* args,
1253 UErrorCode* err)
1254 {
1255 char LMBCS [ULMBCS_CHARSIZE_MAX];
1256 UChar uniChar; /* one output UNICODE char */
1257 const char * saveSource; /* beginning of current code point */
1258 const char * pStartLMBCS = args->source; /* beginning of whole string */
1259 const char * errSource = NULL; /* pointer to actual input in case an error occurs */
1260 int8_t savebytes = 0;
1261
1262 /* Process from source to limit, or until error */
1263 while (U_SUCCESS(*err) && args->sourceLimit > args->source && args->targetLimit > args->target)
1264 {
1265 saveSource = args->source; /* beginning of current code point */
1266
1267 if (args->converter->toULength) /* reassemble char from previous call */
1268 {
1269 const char *saveSourceLimit;
1270 size_t size_old = args->converter->toULength;
1271
1272 /* limit from source is either remainder of temp buffer, or user limit on source */
1273 size_t size_new_maybe_1 = sizeof(LMBCS) - size_old;
1274 size_t size_new_maybe_2 = args->sourceLimit - args->source;
1275 size_t size_new = (size_new_maybe_1 < size_new_maybe_2) ? size_new_maybe_1 : size_new_maybe_2;
1276
1277
1278 uprv_memcpy(LMBCS, args->converter->toUBytes, size_old);
1279 uprv_memcpy(LMBCS + size_old, args->source, size_new);
1280 saveSourceLimit = args->sourceLimit;
1281 args->source = errSource = LMBCS;
1282 args->sourceLimit = LMBCS+size_old+size_new;
1283 savebytes = (int8_t)(size_old+size_new);
1284 uniChar = (UChar) _LMBCSGetNextUCharWorker(args, err);
1285 args->source = saveSource + ((args->source - LMBCS) - size_old);
1286 args->sourceLimit = saveSourceLimit;
1287
1288 if (*err == U_TRUNCATED_CHAR_FOUND)
1289 {
1290 /* evil special case: source buffers so small a char spans more than 2 buffers */
1291 args->converter->toULength = savebytes;
1292 uprv_memcpy(args->converter->toUBytes, LMBCS, savebytes);
1293 args->source = args->sourceLimit;
1294 *err = U_ZERO_ERROR;
1295 return;
1296 }
1297 else
1298 {
1299 /* clear the partial-char marker */
1300 args->converter->toULength = 0;
1301 }
1302 }
1303 else
1304 {
1305 errSource = saveSource;
1306 uniChar = (UChar) _LMBCSGetNextUCharWorker(args, err);
1307 savebytes = (int8_t)(args->source - saveSource);
1308 }
1309 if (U_SUCCESS(*err))
1310 {
1311 if (uniChar < 0xfffe)
1312 {
1313 *(args->target)++ = uniChar;
1314 if(args->offsets)
1315 {
1316 *(args->offsets)++ = (int32_t)(saveSource - pStartLMBCS);
1317 }
1318 }
1319 else if (uniChar == 0xfffe)
1320 {
1321 *err = U_INVALID_CHAR_FOUND;
1322 }
1323 else /* if (uniChar == 0xffff) */
1324 {
1325 *err = U_ILLEGAL_CHAR_FOUND;
1326 }
1327 }
1328 }
1329 /* if target ran out before source, return U_BUFFER_OVERFLOW_ERROR */
1330 if (U_SUCCESS(*err) && args->sourceLimit > args->source && args->targetLimit <= args->target)
1331 {
1332 *err = U_BUFFER_OVERFLOW_ERROR;
1333 }
1334 else if (U_FAILURE(*err))
1335 {
1336 /* If character incomplete or unmappable/illegal, store it in toUBytes[] */
1337 args->converter->toULength = savebytes;
1338 if (savebytes > 0) {
1339 uprv_memcpy(args->converter->toUBytes, errSource, savebytes);
1340 }
1341 if (*err == U_TRUNCATED_CHAR_FOUND) {
1342 *err = U_ZERO_ERROR;
1343 }
1344 }
1345 }
1346
1347 /* And now, the macroized declarations of data & functions: */
1348 DEFINE_LMBCS_OPEN(1)
1349 DEFINE_LMBCS_OPEN(2)
1350 DEFINE_LMBCS_OPEN(3)
1351 DEFINE_LMBCS_OPEN(4)
1352 DEFINE_LMBCS_OPEN(5)
1353 DEFINE_LMBCS_OPEN(6)
1354 DEFINE_LMBCS_OPEN(8)
1355 DEFINE_LMBCS_OPEN(11)
1356 DEFINE_LMBCS_OPEN(16)
1357 DEFINE_LMBCS_OPEN(17)
1358 DEFINE_LMBCS_OPEN(18)
1359 DEFINE_LMBCS_OPEN(19)
1360
1361
1362 DECLARE_LMBCS_DATA(1)
1363 DECLARE_LMBCS_DATA(2)
1364 DECLARE_LMBCS_DATA(3)
1365 DECLARE_LMBCS_DATA(4)
1366 DECLARE_LMBCS_DATA(5)
1367 DECLARE_LMBCS_DATA(6)
1368 DECLARE_LMBCS_DATA(8)
1369 DECLARE_LMBCS_DATA(11)
1370 DECLARE_LMBCS_DATA(16)
1371 DECLARE_LMBCS_DATA(17)
1372 DECLARE_LMBCS_DATA(18)
1373 DECLARE_LMBCS_DATA(19)
1374
1375 #endif /* #if !UCONFIG_NO_LEGACY_CONVERSION */
1376