2
**********************************************************************
3
* Copyright (C) 2000-2004, International Business Machines
4
* Corporation and others. All Rights Reserved.
5
**********************************************************************
6
* file name: ucnv_lmb.cpp
8
* tab size: 4 (not used)
11
* created on: 2000feb09
12
* created by: Brendan Murray
13
* extensively hacked up by: Jim Snyder-Grant
15
* Modification History:
17
* Date Name Description
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.
26
#include "unicode/utypes.h"
28
#if !UCONFIG_NO_CONVERSION && !UCONFIG_NO_LEGACY_CONVERSION
30
#include "unicode/ucnv_err.h"
31
#include "unicode/ucnv.h"
32
#include "unicode/uset.h"
40
#define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0]))
45
(Lotus Multi-Byte Character Set)
47
LMBCS was invented in the late 1980's and is primarily used in Lotus Notes
48
databases and in Lotus 1-2-3 files. Programmers who work with the APIs
49
into these products will sometimes need to deal with strings in this format.
51
The code in this file provides an implementation for an ICU converter of
52
LMBCS to and from Unicode.
54
Since the LMBCS character set is only sparsely documented in existing
55
printed or online material, we have added extensive annotation to this
56
file to serve as a guide to understanding LMBCS.
58
LMBCS was originally designed with these four sometimes-competing design goals:
60
-Provide encodings for the characters in 12 existing national standards
61
(plus a few other characters)
62
-Minimal memory footprint
63
-Maximal speed of conversion into the existing national character sets
64
-No need to track a changing state as you interpret a string.
67
All of the national character sets LMBCS was trying to encode are 'ANSI'
68
based, in that the bytes from 0x20 - 0x7F are almost exactly the
69
same common Latin unaccented characters and symbols in all character sets.
71
So, in order to help meet the speed & memory design goals, the common ANSI
72
bytes from 0x20-0x7F are represented by the same single-byte values in LMBCS.
74
The general LMBCS code unit is from 1-3 bytes. We can describe the 3 bytes as
79
That is, a sometimes-optional 'group' byte, followed by 1 and sometimes 2
80
data bytes. The maximum size of a LMBCS chjaracter is 3 bytes:
82
#define ULMBCS_CHARSIZE_MAX 3
84
The single-byte values from 0x20 to 0x7F are examples of single D1 bytes.
85
We often have to figure out if byte values are below or above this, so we
86
use the ANSI nomenclature 'C0' and 'C1' to refer to the range of control
87
characters just above & below the common lower-ANSI range */
88
#define ULMBCS_C0END 0x1F
89
#define ULMBCS_C1START 0x80
91
Since LMBCS is always dealing in byte units. we create a local type here for
92
dealing with these units of LMBCS code units:
95
typedef uint8_t ulmbcs_byte_t;
98
Most of the values less than 0x20 are reserved in LMBCS to announce
99
which national character standard is being used for the 'D' bytes.
100
In the comments we show the common name and the IBM character-set ID
101
for these character-set announcers:
104
#define ULMBCS_GRP_L1 0x01 /* Latin-1 :ibm-850 */
105
#define ULMBCS_GRP_GR 0x02 /* Greek :ibm-851 */
106
#define ULMBCS_GRP_HE 0x03 /* Hebrew :ibm-1255 */
107
#define ULMBCS_GRP_AR 0x04 /* Arabic :ibm-1256 */
108
#define ULMBCS_GRP_RU 0x05 /* Cyrillic :ibm-1251 */
109
#define ULMBCS_GRP_L2 0x06 /* Latin-2 :ibm-852 */
110
#define ULMBCS_GRP_TR 0x08 /* Turkish :ibm-1254 */
111
#define ULMBCS_GRP_TH 0x0B /* Thai :ibm-874 */
112
#define ULMBCS_GRP_JA 0x10 /* Japanese :ibm-943 */
113
#define ULMBCS_GRP_KO 0x11 /* Korean :ibm-1261 */
114
#define ULMBCS_GRP_TW 0x12 /* Chinese SC :ibm-950 */
115
#define ULMBCS_GRP_CN 0x13 /* Chinese TC :ibm-1386 */
118
So, the beginning of understanding LMBCS is that IF the first byte of a LMBCS
119
character is one of those 12 values, you can interpret the remaining bytes of
120
that character as coming from one of those character sets. Since the lower
121
ANSI bytes already are represented in single bytes, using one of the character
122
set announcers is used to announce a character that starts with a byte of
125
The character sets are arranged so that the single byte sets all appear
126
before the multi-byte character sets. When we need to tell whether a
127
group byte is for a single byte char set or not we use this define: */
129
#define ULMBCS_DOUBLEOPTGROUP_START 0x10
132
However, to fully understand LMBCS, you must also understand a series of
133
exceptions & optimizations made in service of the design goals.
135
First, those of you who are character set mavens may have noticed that
136
the 'double-byte' character sets are actually multi-byte character sets
137
that can have 1 or two bytes, even in the upper-ascii range. To force
138
each group byte to introduce a fixed-width encoding (to make it faster to
139
count characters), we use a convention of doubling up on the group byte
140
to introduce any single-byte character > 0x80 in an otherwise double-byte
141
character set. So, for example, the LMBCS sequence x10 x10 xAE is the
142
same as '0xAE' in the Japanese code page 943.
144
Next, you will notice that the list of group bytes has some gaps.
145
These are used in various ways.
147
We reserve a few special single byte values for common control
148
characters. These are in the same place as their ANSI eqivalents for speed.
151
#define ULMBCS_HT 0x09 /* Fixed control char - Horizontal Tab */
152
#define ULMBCS_LF 0x0A /* Fixed control char - Line Feed */
153
#define ULMBCS_CR 0x0D /* Fixed control char - Carriage Return */
155
/* Then, 1-2-3 reserved a special single-byte character to put at the
156
beginning of internal 'system' range names: */
158
#define ULMBCS_123SYSTEMRANGE 0x19
160
/* Then we needed a place to put all the other ansi control characters
161
that must be moved to different values because LMBCS reserves those
162
values for other purposes. To represent the control characters, we start
163
with a first byte of 0xF & add the control chaarcter value as the
165
#define ULMBCS_GRP_CTRL 0x0F
167
/* For the C0 controls (less than 0x20), we add 0x20 to preserve the
168
useful doctrine that any byte less than 0x20 in a LMBCS char must be
169
the first byte of a character:*/
170
#define ULMBCS_CTRLOFFSET 0x20
173
Where to put the characters that aren't part of any of the 12 national
174
character sets? The first thing that was done, in the earlier years of
175
LMBCS, was to use up the spaces of the form
179
where 'G' was one of the single-byte character groups, and
180
D1 was less than 0x80. These sequences are gathered together
181
into a Lotus-invented doublebyte character set to represent a
182
lot of stray values. Internally, in this implementation, we track this
183
as group '0', as a place to tuck this exceptions list.*/
185
#define ULMBCS_GRP_EXCEPT 0x00
187
Finally, as the durability and usefulness of UNICODE became clear,
188
LOTUS added a new group 0x14 to hold Unicode values not otherwise
189
represented in LMBCS: */
190
#define ULMBCS_GRP_UNICODE 0x14
191
/* The two bytes appearing after a 0x14 are intrepreted as UFT-16 BE
192
(Big-Endian) characters. The exception comes when the UTF16
193
representation would have a zero as the second byte. In that case,
194
'F6' is used in its place, and the bytes are swapped. (This prevents
195
LMBCS from encoding any Unicode values of the form U+F6xx, but that's OK:
196
0xF6xx is in the middle of the Private Use Area.)*/
197
#define ULMBCS_UNICOMPATZERO 0xF6
199
/* It is also useful in our code to have a constant for the size of
200
a LMBCS char that holds a literal Unicode value */
201
#define ULMBCS_UNICODE_SIZE 3
204
To squish the LMBCS representations down even further, and to make
205
translations even faster,sometimes the optimization group byte can be dropped
206
from a LMBCS character. This is decided on a process-by-process basis. The
207
group byte that is dropped is called the 'optimization group'.
209
For Notes, the optimzation group is always 0x1.*/
210
#define ULMBCS_DEFAULTOPTGROUP 0x1
211
/* For 1-2-3 files, the optimzation group is stored in the header of the 1-2-3
214
In any case, when using ICU, you either pass in the
215
optimization group as part of the name of the converter (LMBCS-1, LMBCS-2,
216
etc.). Using plain 'LMBCS' as the name of the converter will give you
220
*** Implementation strategy ***
223
Because of the extensive use of other character sets, the LMBCS converter
224
keeps a mapping between optimization groups and IBM character sets, so that
225
ICU converters can be created and used as needed. */
227
/* As you can see, even though any byte below 0x20 could be an optimization
228
byte, only those at 0x13 or below can map to an actual converter. To limit
229
some loops and searches, we define a value for that last group converter:*/
231
#define ULMBCS_GRP_LAST 0x13 /* last LMBCS group that has a converter */
233
static const char * const OptGroupByteToCPName[ULMBCS_GRP_LAST + 1] = {
234
/* 0x0000 */ "lmb-excp", /* internal home for the LOTUS exceptions list */
235
/* 0x0001 */ "ibm-850",
236
/* 0x0002 */ "ibm-851",
237
/* 0x0003 */ "windows-1255",
238
/* 0x0004 */ "windows-1256",
239
/* 0x0005 */ "windows-1251",
240
/* 0x0006 */ "ibm-852",
241
/* 0x0007 */ NULL, /* Unused */
242
/* 0x0008 */ "windows-1254",
243
/* 0x0009 */ NULL, /* Control char HT */
244
/* 0x000A */ NULL, /* Control char LF */
245
/* 0x000B */ "windows-874",
246
/* 0x000C */ NULL, /* Unused */
247
/* 0x000D */ NULL, /* Control char CR */
248
/* 0x000E */ NULL, /* Unused */
249
/* 0x000F */ NULL, /* Control chars: 0x0F20 + C0/C1 character: algorithmic */
250
/* 0x0010 */ "windows-932",
251
/* 0x0011 */ "windows-949",
252
/* 0x0012 */ "windows-950",
253
/* 0x0013 */ "windows-936"
255
/* The rest are null, including the 0x0014 Unicode compatibility region
256
and 0x0019, the 1-2-3 system range control char */
260
/* That's approximately all the data that's needed for translating
264
However, to translate Unicode to LMBCS, we need some more support.
266
That's because there are often more than one possible mappings from a Unicode
267
code point back into LMBCS. The first thing we do is look up into a table
268
to figure out if there are more than one possible mappings. This table,
269
arranged by Unicode values (including ranges) either lists which group
270
to use, or says that it could go into one or more of the SBCS sets, or
271
into one or more of the DBCS sets. (If the character exists in both DBCS &
272
SBCS, the table will place it in the SBCS sets, to make the LMBCS code point
273
length as small as possible. Here's the two special markers we use to indicate
274
ambiguous mappings: */
276
#define ULMBCS_AMBIGUOUS_SBCS 0x80 /* could fit in more than one
277
LMBCS sbcs native encoding
278
(example: most accented latin) */
279
#define ULMBCS_AMBIGUOUS_MBCS 0x81 /* could fit in more than one
280
LMBCS mbcs native encoding
283
/* And here's a simple way to see if a group falls in an appropriate range */
284
#define ULMBCS_AMBIGUOUS_MATCH(agroup, xgroup) \
285
((((agroup) == ULMBCS_AMBIGUOUS_SBCS) && \
286
(xgroup) < ULMBCS_DOUBLEOPTGROUP_START) || \
287
(((agroup) == ULMBCS_AMBIGUOUS_MBCS) && \
288
(xgroup) >= ULMBCS_DOUBLEOPTGROUP_START))
291
/* The table & some code to use it: */
294
static const struct _UniLMBCSGrpMap
296
const UChar uniStartRange;
297
const UChar uniEndRange;
298
const ulmbcs_byte_t GrpType;
303
{0x0001, 0x001F, ULMBCS_GRP_CTRL},
304
{0x0080, 0x009F, ULMBCS_GRP_CTRL},
305
{0x00A0, 0x01CD, ULMBCS_AMBIGUOUS_SBCS},
306
{0x01CE, 0x01CE, ULMBCS_GRP_TW },
307
{0x01CF, 0x02B9, ULMBCS_AMBIGUOUS_SBCS},
308
{0x02BA, 0x02BA, ULMBCS_GRP_CN},
309
{0x02BC, 0x02C8, ULMBCS_AMBIGUOUS_SBCS},
310
{0x02C9, 0x02D0, ULMBCS_AMBIGUOUS_MBCS},
311
{0x02D8, 0x02DD, ULMBCS_AMBIGUOUS_SBCS},
312
{0x0384, 0x03CE, ULMBCS_AMBIGUOUS_SBCS},
313
{0x0400, 0x044E, ULMBCS_GRP_RU},
314
{0x044F, 0x044F, ULMBCS_AMBIGUOUS_MBCS},
315
{0x0450, 0x0491, ULMBCS_GRP_RU},
316
{0x05B0, 0x05F2, ULMBCS_GRP_HE},
317
{0x060C, 0x06AF, ULMBCS_GRP_AR},
318
{0x0E01, 0x0E5B, ULMBCS_GRP_TH},
319
{0x200C, 0x200F, ULMBCS_AMBIGUOUS_SBCS},
320
{0x2010, 0x2010, ULMBCS_AMBIGUOUS_MBCS},
321
{0x2013, 0x2015, ULMBCS_AMBIGUOUS_SBCS},
322
{0x2016, 0x2016, ULMBCS_AMBIGUOUS_MBCS},
323
{0x2017, 0x2024, ULMBCS_AMBIGUOUS_SBCS},
324
{0x2025, 0x2025, ULMBCS_AMBIGUOUS_MBCS},
325
{0x2026, 0x2026, ULMBCS_AMBIGUOUS_SBCS},
326
{0x2027, 0x2027, ULMBCS_GRP_CN},
327
{0x2030, 0x2033, ULMBCS_AMBIGUOUS_SBCS},
328
{0x2035, 0x2035, ULMBCS_AMBIGUOUS_MBCS},
329
{0x2039, 0x203A, ULMBCS_AMBIGUOUS_SBCS},
330
{0x203B, 0x203B, ULMBCS_AMBIGUOUS_MBCS},
331
{0x2074, 0x2074, ULMBCS_GRP_KO},
332
{0x207F, 0x207F, ULMBCS_GRP_EXCEPT},
333
{0x2081, 0x2084, ULMBCS_GRP_KO},
334
{0x20A4, 0x20AC, ULMBCS_AMBIGUOUS_SBCS},
335
{0x2103, 0x2109, ULMBCS_AMBIGUOUS_MBCS},
336
{0x2111, 0x2126, ULMBCS_AMBIGUOUS_SBCS},
337
{0x212B, 0x212B, ULMBCS_AMBIGUOUS_MBCS},
338
{0x2135, 0x2135, ULMBCS_AMBIGUOUS_SBCS},
339
{0x2153, 0x2154, ULMBCS_GRP_KO},
340
{0x215B, 0x215E, ULMBCS_GRP_EXCEPT},
341
{0x2160, 0x2179, ULMBCS_AMBIGUOUS_MBCS},
342
{0x2190, 0x2195, ULMBCS_GRP_EXCEPT},
343
{0x2196, 0x2199, ULMBCS_AMBIGUOUS_MBCS},
344
{0x21A8, 0x21A8, ULMBCS_GRP_EXCEPT},
345
{0x21B8, 0x21B9, ULMBCS_GRP_CN},
346
{0x21D0, 0x21D5, ULMBCS_GRP_EXCEPT},
347
{0x21E7, 0x21E7, ULMBCS_GRP_CN},
348
{0x2200, 0x220B, ULMBCS_GRP_EXCEPT},
349
{0x220F, 0x2215, ULMBCS_AMBIGUOUS_MBCS},
350
{0x2219, 0x2220, ULMBCS_GRP_EXCEPT},
351
{0x2223, 0x2228, ULMBCS_AMBIGUOUS_MBCS},
352
{0x2229, 0x222B, ULMBCS_GRP_EXCEPT},
353
{0x222C, 0x223D, ULMBCS_AMBIGUOUS_MBCS},
354
{0x2245, 0x2248, ULMBCS_GRP_EXCEPT},
355
{0x224C, 0x224C, ULMBCS_GRP_TW},
356
{0x2252, 0x2252, ULMBCS_AMBIGUOUS_MBCS},
357
{0x2260, 0x2265, ULMBCS_GRP_EXCEPT},
358
{0x2266, 0x226F, ULMBCS_AMBIGUOUS_MBCS},
359
{0x2282, 0x2297, ULMBCS_GRP_EXCEPT},
360
{0x2299, 0x22BF, ULMBCS_AMBIGUOUS_MBCS},
361
{0x22C0, 0x22C0, ULMBCS_GRP_EXCEPT},
362
{0x2310, 0x2310, ULMBCS_GRP_EXCEPT},
363
{0x2312, 0x2312, ULMBCS_AMBIGUOUS_MBCS},
364
{0x2318, 0x2321, ULMBCS_GRP_EXCEPT},
365
{0x2318, 0x2321, ULMBCS_GRP_CN},
366
{0x2460, 0x24E9, ULMBCS_AMBIGUOUS_MBCS},
367
{0x2500, 0x2500, ULMBCS_AMBIGUOUS_SBCS},
368
{0x2501, 0x2501, ULMBCS_AMBIGUOUS_MBCS},
369
{0x2502, 0x2502, ULMBCS_AMBIGUOUS_SBCS},
370
{0x2503, 0x2503, ULMBCS_AMBIGUOUS_MBCS},
371
{0x2504, 0x2505, ULMBCS_GRP_TW},
372
{0x2506, 0x2665, ULMBCS_AMBIGUOUS_MBCS},
373
{0x2666, 0x2666, ULMBCS_GRP_EXCEPT},
374
{0x2666, 0x2666, ULMBCS_GRP_EXCEPT},
375
{0x2667, 0x2E7F, ULMBCS_AMBIGUOUS_SBCS},
376
{0x2E80, 0xF861, ULMBCS_AMBIGUOUS_MBCS},
377
{0xF862, 0xF8FF, ULMBCS_GRP_EXCEPT},
378
{0xF900, 0xFA2D, ULMBCS_AMBIGUOUS_MBCS},
379
{0xFB00, 0xFEFF, ULMBCS_AMBIGUOUS_SBCS},
380
{0xFF01, 0xFFEE, ULMBCS_AMBIGUOUS_MBCS},
381
{0xFFFF, 0xFFFF, ULMBCS_GRP_UNICODE}
385
FindLMBCSUniRange(UChar uniChar)
387
const struct _UniLMBCSGrpMap * pTable = UniLMBCSGrpMap;
389
while (uniChar > pTable->uniEndRange)
394
if (uniChar >= pTable->uniStartRange)
396
return pTable->GrpType;
398
return ULMBCS_GRP_UNICODE;
402
We also ask the creator of a converter to send in a preferred locale
403
that we can use in resolving ambiguous mappings. They send the locale
404
in as a string, and we map it, if possible, to one of the
405
LMBCS groups. We use this table, and the associated code, to
408
/**************************************************
409
This table maps locale ID's to LMBCS opt groups.
410
The default return is group 0x01. Note that for
411
performance reasons, the table is sorted in
412
increasing alphabetic order, with the notable
413
exception of zhTW. This is to force the check
414
for Traditonal Chinese before dropping back to
417
Note too that the Latin-1 groups have been
418
commented out because it's the default, and
419
this shortens the table, allowing a serial
420
search to go quickly.
421
*************************************************/
423
static const struct _LocaleLMBCSGrpMap
425
const char *LocaleID;
426
const ulmbcs_byte_t OptGroup;
427
} LocaleLMBCSGrpMap[] =
429
{"ar", ULMBCS_GRP_AR},
430
{"be", ULMBCS_GRP_RU},
431
{"bg", ULMBCS_GRP_L2},
432
/* {"ca", ULMBCS_GRP_L1}, */
433
{"cs", ULMBCS_GRP_L2},
434
/* {"da", ULMBCS_GRP_L1}, */
435
/* {"de", ULMBCS_GRP_L1}, */
436
{"el", ULMBCS_GRP_GR},
437
/* {"en", ULMBCS_GRP_L1}, */
438
/* {"es", ULMBCS_GRP_L1}, */
439
/* {"et", ULMBCS_GRP_L1}, */
440
/* {"fi", ULMBCS_GRP_L1}, */
441
/* {"fr", ULMBCS_GRP_L1}, */
442
{"he", ULMBCS_GRP_HE},
443
{"hu", ULMBCS_GRP_L2},
444
/* {"is", ULMBCS_GRP_L1}, */
445
/* {"it", ULMBCS_GRP_L1}, */
446
{"iw", ULMBCS_GRP_HE},
447
{"ja", ULMBCS_GRP_JA},
448
{"ko", ULMBCS_GRP_KO},
449
/* {"lt", ULMBCS_GRP_L1}, */
450
/* {"lv", ULMBCS_GRP_L1}, */
451
{"mk", ULMBCS_GRP_RU},
452
/* {"nl", ULMBCS_GRP_L1}, */
453
/* {"no", ULMBCS_GRP_L1}, */
454
{"pl", ULMBCS_GRP_L2},
455
/* {"pt", ULMBCS_GRP_L1}, */
456
{"ro", ULMBCS_GRP_L2},
457
{"ru", ULMBCS_GRP_RU},
458
{"sh", ULMBCS_GRP_L2},
459
{"sk", ULMBCS_GRP_L2},
460
{"sl", ULMBCS_GRP_L2},
461
{"sq", ULMBCS_GRP_L2},
462
{"sr", ULMBCS_GRP_RU},
463
/* {"sv", ULMBCS_GRP_L1}, */
464
{"th", ULMBCS_GRP_TH},
465
{"tr", ULMBCS_GRP_TR},
466
{"uk", ULMBCS_GRP_RU},
467
/* {"vi", ULMBCS_GRP_L1}, */
468
{"zhTW", ULMBCS_GRP_TW},
469
{"zh", ULMBCS_GRP_CN},
470
{NULL, ULMBCS_GRP_L1}
475
FindLMBCSLocale(const char *LocaleID)
477
const struct _LocaleLMBCSGrpMap *pTable = LocaleLMBCSGrpMap;
479
if ((!LocaleID) || (!*LocaleID))
484
while (pTable->LocaleID)
486
if (*pTable->LocaleID == *LocaleID) /* Check only first char for speed */
488
/* First char matches - check whole name, for entry-length */
489
if (uprv_strncmp(pTable->LocaleID, LocaleID, strlen(pTable->LocaleID)) == 0)
490
return pTable->OptGroup;
493
if (*pTable->LocaleID > *LocaleID) /* Sorted alphabetically - exit */
497
return ULMBCS_GRP_L1;
502
Before we get to the main body of code, here's how we hook up to the rest
503
of ICU. ICU converters are required to define a structure that includes
504
some function pointers, and some common data, in the style of a C++
505
vtable. There is also room in there for converter-specific data. LMBCS
506
uses that converter-specific data to keep track of the 12 subconverters
507
we use, the optimization group, and the group (if any) that matches the
508
locale. We have one structure instantiated for each of the 12 possible
509
optimization groups. To avoid typos & to avoid boring the reader, we
510
put the declarations of these structures and functions into macros. To see
511
the definitions of these structures, see unicode\ucnv_bld.h
516
UConverterSharedData *OptGrpConverter[ULMBCS_GRP_LAST+1]; /* Converter per Opt. grp. */
517
uint8_t OptGroup; /* default Opt. grp. for this LMBCS session */
518
uint8_t localeConverterIndex; /* reasonable locale match for index */
523
#define DECLARE_LMBCS_DATA(n) \
524
static const UConverterImpl _LMBCSImpl##n={\
530
_LMBCSToUnicodeWithOffsets,\
531
_LMBCSToUnicodeWithOffsets,\
541
static const UConverterStaticData _LMBCSStaticData##n={\
542
sizeof(UConverterStaticData),\
544
0, UCNV_IBM, UCNV_LMBCS_##n, 1, 3,\
545
{ 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} \
547
const UConverterSharedData _LMBCSData##n={\
548
sizeof(UConverterSharedData), ~((uint32_t) 0),\
549
NULL, NULL, &_LMBCSStaticData##n, FALSE, &_LMBCSImpl##n, \
553
/* The only function we needed to duplicate 12 times was the 'open'
554
function, which will do basically the same thing except set a different
555
optimization group. So, we put the common stuff into a worker function,
556
and set up another macro to stamp out the 12 open functions:*/
557
#define DEFINE_LMBCS_OPEN(n) \
559
_LMBCSOpen##n(UConverter* _this,const char* name,const char* locale,uint32_t options,UErrorCode* err) \
560
{ _LMBCSOpenWorker(_this, name,locale,options, err, n);}
564
/* Here's the open worker & the common close function */
566
_LMBCSOpenWorker(UConverter* _this,
571
ulmbcs_byte_t OptGroup
574
UConverterDataLMBCS * extraInfo = (UConverterDataLMBCS*)uprv_malloc (sizeof (UConverterDataLMBCS));
575
if(extraInfo != NULL)
579
uprv_memset(extraInfo, 0, sizeof(UConverterDataLMBCS));
581
for (i=0; i <= ULMBCS_GRP_LAST && U_SUCCESS(*err); i++)
583
if(OptGroupByteToCPName[i] != NULL) {
584
extraInfo->OptGrpConverter[i] = ucnv_loadSharedData(OptGroupByteToCPName[i], NULL, err);
588
if(U_SUCCESS(*err)) {
589
extraInfo->OptGroup = OptGroup;
590
extraInfo->localeConverterIndex = FindLMBCSLocale(locale);
592
/* one of the subconverters could not be loaded, unload the previous ones */
594
if(extraInfo->OptGrpConverter[--i] != NULL) {
595
ucnv_unloadSharedDataIfReady(extraInfo->OptGrpConverter[i]);
596
extraInfo->OptGrpConverter[i] = NULL;
603
*err = U_MEMORY_ALLOCATION_ERROR;
605
_this->extraInfo = extraInfo;
609
_LMBCSClose(UConverter * _this)
611
if (_this->extraInfo != NULL)
614
UConverterDataLMBCS * extraInfo = (UConverterDataLMBCS *) _this->extraInfo;
616
for (Ix=0; Ix <= ULMBCS_GRP_LAST; Ix++)
618
if (extraInfo->OptGrpConverter[Ix] != NULL)
619
ucnv_unloadSharedDataIfReady(extraInfo->OptGrpConverter[Ix]);
621
if (!_this->isExtraLocal) {
622
uprv_free (_this->extraInfo);
627
typedef struct LMBCSClone {
629
UConverterDataLMBCS lmbcs;
633
_LMBCSSafeClone(const UConverter *cnv,
635
int32_t *pBufferSize,
636
UErrorCode *status) {
637
LMBCSClone *newLMBCS;
638
UConverterDataLMBCS *extraInfo;
641
if(*pBufferSize<=0) {
642
*pBufferSize=(int32_t)sizeof(LMBCSClone);
646
extraInfo=(UConverterDataLMBCS *)cnv->extraInfo;
647
newLMBCS=(LMBCSClone *)stackBuffer;
649
/* ucnv.c/ucnv_safeClone() copied the main UConverter already */
651
uprv_memcpy(&newLMBCS->lmbcs, extraInfo, sizeof(UConverterDataLMBCS));
653
/* share the subconverters */
654
for(i = 0; i <= ULMBCS_GRP_LAST; ++i) {
655
if(extraInfo->OptGrpConverter[i] != NULL) {
656
ucnv_incrementRefCount(extraInfo->OptGrpConverter[i]);
660
newLMBCS->cnv.extraInfo = &newLMBCS->lmbcs;
661
newLMBCS->cnv.isExtraLocal = TRUE;
662
return &newLMBCS->cnv;
666
_LMBCSGetUnicodeSet(const UConverter *cnv,
668
UConverterUnicodeSet which,
669
UErrorCode *pErrorCode) {
670
/* all but U+F6xx, see LMBCS explanation above (search for F6xx) */
671
sa->addRange(sa->set, 0, 0xf5ff);
672
sa->addRange(sa->set, 0xf700, 0x10ffff);
676
Here's the basic helper function that we use when converting from
677
Unicode to LMBCS, and we suspect that a Unicode character will fit into
678
one of the 12 groups. The return value is the number of bytes written
679
starting at pStartLMBCS (if any).
683
LMBCSConversionWorker (
684
UConverterDataLMBCS * extraInfo, /* subconverters, opt & locale groups */
685
ulmbcs_byte_t group, /* The group to try */
686
ulmbcs_byte_t * pStartLMBCS, /* where to put the results */
687
UChar * pUniChar, /* The input unicode character */
688
ulmbcs_byte_t * lastConverterIndex, /* output: track last successful group used */
689
UBool * groups_tried /* output: track any unsuccessful groups */
692
ulmbcs_byte_t * pLMBCS = pStartLMBCS;
693
UConverterSharedData * xcnv = extraInfo->OptGrpConverter[group];
697
ulmbcs_byte_t firstByte;
700
U_ASSERT(group<ULMBCS_GRP_UNICODE);
702
bytesConverted = ucnv_MBCSFromUChar32(xcnv, *pUniChar, &value, FALSE);
704
/* get the first result byte */
705
if(bytesConverted > 0) {
706
firstByte = (ulmbcs_byte_t)(value >> ((bytesConverted - 1) * 8));
708
/* most common failure mode is an unassigned character */
709
groups_tried[group] = TRUE;
713
*lastConverterIndex = group;
715
/* All initial byte values in lower ascii range should have been caught by now,
716
except with the exception group.
718
U_ASSERT((firstByte <= ULMBCS_C0END) || (firstByte >= ULMBCS_C1START) || (group == ULMBCS_GRP_EXCEPT));
720
/* use converted data: first write 0, 1 or two group bytes */
721
if (group != ULMBCS_GRP_EXCEPT && extraInfo->OptGroup != group)
724
if (bytesConverted == 1 && group >= ULMBCS_DOUBLEOPTGROUP_START)
730
/* don't emit control chars */
731
if ( bytesConverted == 1 && firstByte < 0x20 )
735
/* then move over the converted data */
736
switch(bytesConverted)
739
*pLMBCS++ = (ulmbcs_byte_t)(value >> 24);
741
*pLMBCS++ = (ulmbcs_byte_t)(value >> 16);
743
*pLMBCS++ = (ulmbcs_byte_t)(value >> 8);
745
*pLMBCS++ = (ulmbcs_byte_t)value;
747
/* will never occur */
751
return (pLMBCS - pStartLMBCS);
755
/* This is a much simpler version of above, when we
756
know we are writing LMBCS using the Unicode group
759
LMBCSConvertUni(ulmbcs_byte_t * pLMBCS, UChar uniChar)
761
/* encode into LMBCS Unicode range */
762
uint8_t LowCh = (uint8_t)(uniChar & 0x00FF);
763
uint8_t HighCh = (uint8_t)(uniChar >> 8);
765
*pLMBCS++ = ULMBCS_GRP_UNICODE;
769
*pLMBCS++ = ULMBCS_UNICOMPATZERO;
777
return ULMBCS_UNICODE_SIZE;
782
/* The main Unicode to LMBCS conversion function */
784
_LMBCSFromUnicode(UConverterFromUnicodeArgs* args,
787
ulmbcs_byte_t lastConverterIndex = 0;
789
ulmbcs_byte_t LMBCS[ULMBCS_CHARSIZE_MAX];
790
ulmbcs_byte_t * pLMBCS;
792
UBool groups_tried[ULMBCS_GRP_LAST+1];
793
UConverterDataLMBCS * extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo;
797
/* Basic strategy: attempt to fill in local LMBCS 1-char buffer.(LMBCS)
798
If that succeeds, see if it will all fit into the target & copy it over
801
We try conversions in the following order:
803
1. Single-byte ascii & special fixed control chars (&null)
804
2. Look up group in table & try that (could be
807
C) national encoding,
808
or ambiguous SBCS or MBCS group (on to step 4...)
810
3. If its ambiguous, try this order:
811
A) The optimization group
813
C) The last group that succeeded with this string.
814
D) every other group that's relevent (single or double)
815
E) If its single-byte ambiguous, try the exceptions group
817
4. And as a grand fallback: Unicode
820
while (args->source < args->sourceLimit && !U_FAILURE(*err))
822
if (args->target >= args->targetLimit)
824
*err = U_BUFFER_OVERFLOW_ERROR;
827
uniChar = *(args->source);
831
/* check cases in rough order of how common they are, for speed */
833
/* single byte matches: strategy 1 */
835
if (((uniChar > ULMBCS_C0END) && (uniChar < ULMBCS_C1START)) ||
836
uniChar == 0 || uniChar == ULMBCS_HT || uniChar == ULMBCS_CR ||
837
uniChar == ULMBCS_LF || uniChar == ULMBCS_123SYSTEMRANGE
840
*pLMBCS++ = (ulmbcs_byte_t ) uniChar;
847
/* Check by UNICODE range (Strategy 2) */
848
ulmbcs_byte_t group = FindLMBCSUniRange(uniChar);
850
if (group == ULMBCS_GRP_UNICODE) /* (Strategy 2A) */
852
pLMBCS += LMBCSConvertUni(pLMBCS,uniChar);
854
bytes_written = pLMBCS - LMBCS;
856
else if (group == ULMBCS_GRP_CTRL) /* (Strategy 2B) */
858
/* Handle control characters here */
859
if (uniChar <= ULMBCS_C0END)
861
*pLMBCS++ = ULMBCS_GRP_CTRL;
862
*pLMBCS++ = (ulmbcs_byte_t)(ULMBCS_CTRLOFFSET + uniChar);
864
else if (uniChar >= ULMBCS_C1START && uniChar <= ULMBCS_C1START + ULMBCS_CTRLOFFSET)
866
*pLMBCS++ = ULMBCS_GRP_CTRL;
867
*pLMBCS++ = (ulmbcs_byte_t ) (uniChar & 0x00FF);
869
bytes_written = pLMBCS - LMBCS;
871
else if (group < ULMBCS_GRP_UNICODE) /* (Strategy 2C) */
873
/* a specific converter has been identified - use it */
874
bytes_written = LMBCSConversionWorker (
875
extraInfo, group, pLMBCS, &uniChar,
876
&lastConverterIndex, groups_tried);
878
if (!bytes_written) /* the ambiguous group cases (Strategy 3) */
880
uprv_memset(groups_tried, 0, sizeof(groups_tried));
882
/* check for non-default optimization group (Strategy 3A )*/
883
if (extraInfo->OptGroup != 1
884
&& ULMBCS_AMBIGUOUS_MATCH(group, extraInfo->OptGroup))
886
bytes_written = LMBCSConversionWorker (extraInfo,
887
extraInfo->OptGroup, pLMBCS, &uniChar,
888
&lastConverterIndex, groups_tried);
890
/* check for locale optimization group (Strategy 3B) */
892
&& (extraInfo->localeConverterIndex)
893
&& (ULMBCS_AMBIGUOUS_MATCH(group, extraInfo->localeConverterIndex)))
895
bytes_written = LMBCSConversionWorker (extraInfo,
896
extraInfo->localeConverterIndex, pLMBCS, &uniChar,
897
&lastConverterIndex, groups_tried);
899
/* check for last optimization group used for this string (Strategy 3C) */
901
&& (lastConverterIndex)
902
&& (ULMBCS_AMBIGUOUS_MATCH(group, lastConverterIndex)))
904
bytes_written = LMBCSConversionWorker (extraInfo,
905
lastConverterIndex, pLMBCS, &uniChar,
906
&lastConverterIndex, groups_tried);
911
/* just check every possible matching converter (Strategy 3D) */
912
ulmbcs_byte_t grp_start;
913
ulmbcs_byte_t grp_end;
914
ulmbcs_byte_t grp_ix;
915
grp_start = (ulmbcs_byte_t)((group == ULMBCS_AMBIGUOUS_MBCS)
916
? ULMBCS_DOUBLEOPTGROUP_START
918
grp_end = (ulmbcs_byte_t)((group == ULMBCS_AMBIGUOUS_MBCS)
921
for (grp_ix = grp_start;
922
grp_ix <= grp_end && !bytes_written;
925
if (extraInfo->OptGrpConverter [grp_ix] && !groups_tried [grp_ix])
927
bytes_written = LMBCSConversionWorker (extraInfo,
928
grp_ix, pLMBCS, &uniChar,
929
&lastConverterIndex, groups_tried);
932
/* a final conversion fallback to the exceptions group if its likely
933
to be single byte (Strategy 3E) */
934
if (!bytes_written && grp_start == ULMBCS_GRP_L1)
936
bytes_written = LMBCSConversionWorker (extraInfo,
937
ULMBCS_GRP_EXCEPT, pLMBCS, &uniChar,
938
&lastConverterIndex, groups_tried);
941
/* all of our other strategies failed. Fallback to Unicode. (Strategy 4)*/
945
pLMBCS += LMBCSConvertUni(pLMBCS, uniChar);
946
bytes_written = pLMBCS - LMBCS;
951
/* we have a translation. increment source and write as much as posible to target */
954
while (args->target < args->targetLimit && bytes_written--)
956
*(args->target)++ = *pLMBCS++;
959
*(args->offsets)++ = sourceIndex;
963
if (bytes_written > 0)
965
/* write any bytes that didn't fit in target to the error buffer,
966
common code will move this to target if we get called back with
969
uint8_t * pErrorBuffer = args->converter->charErrorBuffer;
970
*err = U_BUFFER_OVERFLOW_ERROR;
971
args->converter->charErrorBufferLength = (int8_t)bytes_written;
972
while (bytes_written--)
974
*pErrorBuffer++ = *pLMBCS++;
981
/* Now, the Unicode from LMBCS section */
984
/* A function to call when we are looking at the Unicode group byte in LMBCS */
986
GetUniFromLMBCSUni(char const ** ppLMBCSin) /* Called with LMBCS-style Unicode byte stream */
988
uint8_t HighCh = *(*ppLMBCSin)++; /* Big-endian Unicode in LMBCS compatibility group*/
989
uint8_t LowCh = *(*ppLMBCSin)++;
991
if (HighCh == ULMBCS_UNICOMPATZERO )
994
LowCh = 0; /* zero-byte in LSB special character */
996
return (UChar)((HighCh << 8) | LowCh);
1001
/* CHECK_SOURCE_LIMIT: Helper macro to verify that there are at least'index'
1002
bytes left in source up to sourceLimit.Errors appropriately if not.
1003
If we reach the limit, then update the source pointer to there to consume
1004
all input as required by ICU converter semantics.
1007
#define CHECK_SOURCE_LIMIT(index) \
1008
if (args->source+index > args->sourceLimit){\
1009
*err = U_TRUNCATED_CHAR_FOUND;\
1010
args->source = args->sourceLimit;\
1013
/* Return the Unicode representation for the current LMBCS character */
1016
_LMBCSGetNextUCharWorker(UConverterToUnicodeArgs* args,
1019
UChar32 uniChar = 0; /* an output UNICODE char */
1020
ulmbcs_byte_t CurByte; /* A byte from the input stream */
1021
const char * saveSource;
1024
if (args->source >= args->sourceLimit)
1026
*err = U_ILLEGAL_ARGUMENT_ERROR;
1029
/* Grab first byte & save address for error recovery */
1030
CurByte = *((ulmbcs_byte_t *) (saveSource = args->source++));
1033
* at entry of each if clause:
1034
* 1. 'CurByte' points at the first byte of a LMBCS character
1035
* 2. '*source'points to the next byte of the source stream after 'CurByte'
1037
* the job of each if clause is:
1038
* 1. set '*source' to point at the beginning of next char (nop if LMBCS char is only 1 byte)
1039
* 2. set 'uniChar' up with the right Unicode value, or set 'err' appropriately
1042
/* First lets check the simple fixed values. */
1044
if(((CurByte > ULMBCS_C0END) && (CurByte < ULMBCS_C1START)) /* ascii range */
1046
|| CurByte == ULMBCS_HT || CurByte == ULMBCS_CR
1047
|| CurByte == ULMBCS_LF || CurByte == ULMBCS_123SYSTEMRANGE)
1053
UConverterDataLMBCS * extraInfo;
1054
ulmbcs_byte_t group;
1055
UConverterSharedData *cnv;
1057
if (CurByte == ULMBCS_GRP_CTRL) /* Control character group - no opt group update */
1059
ulmbcs_byte_t C0C1byte;
1060
CHECK_SOURCE_LIMIT(1);
1061
C0C1byte = *(args->source)++;
1062
uniChar = (C0C1byte < ULMBCS_C1START) ? C0C1byte - ULMBCS_CTRLOFFSET : C0C1byte;
1065
if (CurByte == ULMBCS_GRP_UNICODE) /* Unicode compatibility group: BigEndian UTF16 */
1067
CHECK_SOURCE_LIMIT(2);
1069
/* don't check for error indicators fffe/ffff below */
1070
return GetUniFromLMBCSUni(&(args->source));
1072
else if (CurByte <= ULMBCS_CTRLOFFSET)
1074
group = CurByte; /* group byte is in the source */
1075
extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo;
1076
if (group > ULMBCS_GRP_LAST || (cnv = extraInfo->OptGrpConverter[group]) == NULL)
1078
/* this is not a valid group byte - no converter*/
1079
*err = U_INVALID_CHAR_FOUND;
1081
else if (group >= ULMBCS_DOUBLEOPTGROUP_START) /* double byte conversion */
1084
CHECK_SOURCE_LIMIT(2);
1086
/* check for LMBCS doubled-group-byte case */
1087
if (*args->source == group) {
1090
uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source, 1, FALSE);
1094
uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source, 2, FALSE);
1098
else { /* single byte conversion */
1099
CHECK_SOURCE_LIMIT(1);
1100
CurByte = *(args->source)++;
1102
if (CurByte >= ULMBCS_C1START)
1104
uniChar = _MBCS_SINGLE_SIMPLE_GET_NEXT_BMP(cnv, CurByte);
1108
/* The non-optimizable oddballs where there is an explicit byte
1109
* AND the second byte is not in the upper ascii range
1113
extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo;
1114
cnv = extraInfo->OptGrpConverter [ULMBCS_GRP_EXCEPT];
1116
/* Lookup value must include opt group */
1119
uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, bytes, 2, FALSE);
1123
else if (CurByte >= ULMBCS_C1START) /* group byte is implicit */
1125
extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo;
1126
group = extraInfo->OptGroup;
1127
cnv = extraInfo->OptGrpConverter[group];
1128
if (group >= ULMBCS_DOUBLEOPTGROUP_START) /* double byte conversion */
1130
if (!ucnv_MBCSIsLeadByte(cnv, CurByte))
1132
CHECK_SOURCE_LIMIT(0);
1134
/* let the MBCS conversion consume CurByte again */
1135
uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source - 1, 1, FALSE);
1139
CHECK_SOURCE_LIMIT(1);
1140
/* let the MBCS conversion consume CurByte again */
1141
uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source - 1, 2, FALSE);
1145
else /* single byte conversion */
1147
uniChar = _MBCS_SINGLE_SIMPLE_GET_NEXT_BMP(cnv, CurByte);
1155
/* The exported function that converts lmbcs to one or more
1156
UChars - currently UTF-16
1159
_LMBCSToUnicodeWithOffsets(UConverterToUnicodeArgs* args,
1162
char LMBCS [ULMBCS_CHARSIZE_MAX];
1163
UChar uniChar; /* one output UNICODE char */
1164
const char * saveSource; /* beginning of current code point */
1165
const char * pStartLMBCS = args->source; /* beginning of whole string */
1166
const char * errSource = NULL; /* pointer to actual input in case an error occurs */
1167
int8_t savebytes = 0;
1169
/* Process from source to limit, or until error */
1170
while (U_SUCCESS(*err) && args->sourceLimit > args->source && args->targetLimit > args->target)
1172
saveSource = args->source; /* beginning of current code point */
1174
if (args->converter->toULength) /* reassemble char from previous call */
1176
const char *saveSourceLimit;
1177
size_t size_old = args->converter->toULength;
1179
/* limit from source is either remainder of temp buffer, or user limit on source */
1180
size_t size_new_maybe_1 = sizeof(LMBCS) - size_old;
1181
size_t size_new_maybe_2 = args->sourceLimit - args->source;
1182
size_t size_new = (size_new_maybe_1 < size_new_maybe_2) ? size_new_maybe_1 : size_new_maybe_2;
1185
uprv_memcpy(LMBCS, args->converter->toUBytes, size_old);
1186
uprv_memcpy(LMBCS + size_old, args->source, size_new);
1187
saveSourceLimit = args->sourceLimit;
1188
args->source = errSource = LMBCS;
1189
args->sourceLimit = LMBCS+size_old+size_new;
1190
savebytes = (int8_t)(size_old+size_new);
1191
uniChar = (UChar) _LMBCSGetNextUCharWorker(args, err);
1192
args->source = saveSource + ((args->source - LMBCS) - size_old);
1193
args->sourceLimit = saveSourceLimit;
1195
if (*err == U_TRUNCATED_CHAR_FOUND)
1197
/* evil special case: source buffers so small a char spans more than 2 buffers */
1198
args->converter->toULength = savebytes;
1199
uprv_memcpy(args->converter->toUBytes, LMBCS, savebytes);
1200
args->source = args->sourceLimit;
1201
*err = U_ZERO_ERROR;
1206
/* clear the partial-char marker */
1207
args->converter->toULength = 0;
1212
errSource = saveSource;
1213
uniChar = (UChar) _LMBCSGetNextUCharWorker(args, err);
1214
savebytes = (int8_t)(args->source - saveSource);
1216
if (U_SUCCESS(*err))
1218
if (uniChar < 0xfffe)
1220
*(args->target)++ = uniChar;
1223
*(args->offsets)++ = saveSource - pStartLMBCS;
1226
else if (uniChar == 0xfffe)
1228
*err = U_INVALID_CHAR_FOUND;
1230
else /* if (uniChar == 0xffff) */
1232
*err = U_ILLEGAL_CHAR_FOUND;
1236
/* if target ran out before source, return U_BUFFER_OVERFLOW_ERROR */
1237
if (U_SUCCESS(*err) && args->sourceLimit > args->source && args->targetLimit <= args->target)
1239
*err = U_BUFFER_OVERFLOW_ERROR;
1241
else if (U_FAILURE(*err))
1243
/* If character incomplete or unmappable/illegal, store it in toUBytes[] */
1244
args->converter->toULength = savebytes;
1245
if (savebytes > 0) {
1246
uprv_memcpy(args->converter->toUBytes, errSource, savebytes);
1248
if (*err == U_TRUNCATED_CHAR_FOUND) {
1249
*err = U_ZERO_ERROR;
1254
/* And now, the macroized declarations of data & functions: */
1255
DEFINE_LMBCS_OPEN(1)
1256
DEFINE_LMBCS_OPEN(2)
1257
DEFINE_LMBCS_OPEN(3)
1258
DEFINE_LMBCS_OPEN(4)
1259
DEFINE_LMBCS_OPEN(5)
1260
DEFINE_LMBCS_OPEN(6)
1261
DEFINE_LMBCS_OPEN(8)
1262
DEFINE_LMBCS_OPEN(11)
1263
DEFINE_LMBCS_OPEN(16)
1264
DEFINE_LMBCS_OPEN(17)
1265
DEFINE_LMBCS_OPEN(18)
1266
DEFINE_LMBCS_OPEN(19)
1269
DECLARE_LMBCS_DATA(1)
1270
DECLARE_LMBCS_DATA(2)
1271
DECLARE_LMBCS_DATA(3)
1272
DECLARE_LMBCS_DATA(4)
1273
DECLARE_LMBCS_DATA(5)
1274
DECLARE_LMBCS_DATA(6)
1275
DECLARE_LMBCS_DATA(8)
1276
DECLARE_LMBCS_DATA(11)
1277
DECLARE_LMBCS_DATA(16)
1278
DECLARE_LMBCS_DATA(17)
1279
DECLARE_LMBCS_DATA(18)
1280
DECLARE_LMBCS_DATA(19)
1282
#endif /* #if !UCONFIG_NO_LEGACY_CONVERSION */