ucnv_lmb.c revision c69afcec261fc345fda8daf46f0ea6b4351dc777
1/* 2********************************************************************** 3* Copyright (C) 2000-2007, 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 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#define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0])) 41 42/* 43 LMBCS 44 45 (Lotus Multi-Byte Character Set) 46 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. 50 51 The code in this file provides an implementation for an ICU converter of 52 LMBCS to and from Unicode. 53 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. 57 58 LMBCS was originally designed with these four sometimes-competing design goals: 59 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. 65 66 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. 70 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. 73 74 The general LMBCS code unit is from 1-3 bytes. We can describe the 3 bytes as 75 follows: 76 77 [G] D1 [D2] 78 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: 81*/ 82#define ULMBCS_CHARSIZE_MAX 3 83/* 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 90/* 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: 93 94*/ 95typedef uint8_t ulmbcs_byte_t; 96 97/* 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: 102*/ 103 104#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 */ 116 117/* 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 123 0x80 or greater. 124 125 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: */ 128 129#define ULMBCS_DOUBLEOPTGROUP_START 0x10 130 131/* 132However, to fully understand LMBCS, you must also understand a series of 133exceptions & optimizations made in service of the design goals. 134 135First, those of you who are character set mavens may have noticed that 136the 'double-byte' character sets are actually multi-byte character sets 137that can have 1 or two bytes, even in the upper-ascii range. To force 138each group byte to introduce a fixed-width encoding (to make it faster to 139count characters), we use a convention of doubling up on the group byte 140to introduce any single-byte character > 0x80 in an otherwise double-byte 141character set. So, for example, the LMBCS sequence x10 x10 xAE is the 142same as '0xAE' in the Japanese code page 943. 143 144Next, you will notice that the list of group bytes has some gaps. 145These are used in various ways. 146 147We reserve a few special single byte values for common control 148characters. These are in the same place as their ANSI eqivalents for speed. 149*/ 150 151#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 */ 154 155/* Then, 1-2-3 reserved a special single-byte character to put at the 156beginning of internal 'system' range names: */ 157 158#define ULMBCS_123SYSTEMRANGE 0x19 159 160/* Then we needed a place to put all the other ansi control characters 161that must be moved to different values because LMBCS reserves those 162values for other purposes. To represent the control characters, we start 163with a first byte of 0xF & add the control chaarcter value as the 164second byte */ 165#define ULMBCS_GRP_CTRL 0x0F 166 167/* For the C0 controls (less than 0x20), we add 0x20 to preserve the 168useful doctrine that any byte less than 0x20 in a LMBCS char must be 169the first byte of a character:*/ 170#define ULMBCS_CTRLOFFSET 0x20 171 172/* 173Where to put the characters that aren't part of any of the 12 national 174character sets? The first thing that was done, in the earlier years of 175LMBCS, was to use up the spaces of the form 176 177 [G] D1, 178 179 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.*/ 184 185#define ULMBCS_GRP_EXCEPT 0x00 186/* 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 193representation 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 195LMBCS from encoding any Unicode values of the form U+F6xx, but that's OK: 1960xF6xx is in the middle of the Private Use Area.)*/ 197#define ULMBCS_UNICOMPATZERO 0xF6 198 199/* It is also useful in our code to have a constant for the size of 200a LMBCS char that holds a literal Unicode value */ 201#define ULMBCS_UNICODE_SIZE 3 202 203/* 204To squish the LMBCS representations down even further, and to make 205translations even faster,sometimes the optimization group byte can be dropped 206from a LMBCS character. This is decided on a process-by-process basis. The 207group byte that is dropped is called the 'optimization group'. 208 209For 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 212file. 213 214 In any case, when using ICU, you either pass in the 215optimization group as part of the name of the converter (LMBCS-1, LMBCS-2, 216etc.). Using plain 'LMBCS' as the name of the converter will give you 217LMBCS-1. 218 219 220*** Implementation strategy *** 221 222 223Because of the extensive use of other character sets, the LMBCS converter 224keeps a mapping between optimization groups and IBM character sets, so that 225ICU converters can be created and used as needed. */ 226 227/* As you can see, even though any byte below 0x20 could be an optimization 228byte, only those at 0x13 or below can map to an actual converter. To limit 229some loops and searches, we define a value for that last group converter:*/ 230 231#define ULMBCS_GRP_LAST 0x13 /* last LMBCS group that has a converter */ 232 233static 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" 254 255 /* The rest are null, including the 0x0014 Unicode compatibility region 256 and 0x0019, the 1-2-3 system range control char */ 257}; 258 259 260/* That's approximately all the data that's needed for translating 261 LMBCS to Unicode. 262 263 264However, to translate Unicode to LMBCS, we need some more support. 265 266That's because there are often more than one possible mappings from a Unicode 267code point back into LMBCS. The first thing we do is look up into a table 268to figure out if there are more than one possible mappings. This table, 269arranged by Unicode values (including ranges) either lists which group 270to use, or says that it could go into one or more of the SBCS sets, or 271into one or more of the DBCS sets. (If the character exists in both DBCS & 272SBCS, the table will place it in the SBCS sets, to make the LMBCS code point 273length as small as possible. Here's the two special markers we use to indicate 274ambiguous mappings: */ 275 276#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 281 (example: Unihan) */ 282 283/* 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)) 289 290 291/* The table & some code to use it: */ 292 293 294static const struct _UniLMBCSGrpMap 295{ 296 const UChar uniStartRange; 297 const UChar uniEndRange; 298 const ulmbcs_byte_t GrpType; 299} UniLMBCSGrpMap[] 300= 301{ 302 303 {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} 382}; 383 384static ulmbcs_byte_t 385FindLMBCSUniRange(UChar uniChar) 386{ 387 const struct _UniLMBCSGrpMap * pTable = UniLMBCSGrpMap; 388 389 while (uniChar > pTable->uniEndRange) 390 { 391 pTable++; 392 } 393 394 if (uniChar >= pTable->uniStartRange) 395 { 396 return pTable->GrpType; 397 } 398 return ULMBCS_GRP_UNICODE; 399} 400 401/* 402We also ask the creator of a converter to send in a preferred locale 403that we can use in resolving ambiguous mappings. They send the locale 404in as a string, and we map it, if possible, to one of the 405LMBCS groups. We use this table, and the associated code, to 406do the lookup: */ 407 408/************************************************** 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 415 Simplified. 416 417 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 *************************************************/ 422 423static const struct _LocaleLMBCSGrpMap 424{ 425 const char *LocaleID; 426 const ulmbcs_byte_t OptGroup; 427} LocaleLMBCSGrpMap[] = 428{ 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} 471}; 472 473 474static ulmbcs_byte_t 475FindLMBCSLocale(const char *LocaleID) 476{ 477 const struct _LocaleLMBCSGrpMap *pTable = LocaleLMBCSGrpMap; 478 479 if ((!LocaleID) || (!*LocaleID)) 480 { 481 return 0; 482 } 483 484 while (pTable->LocaleID) 485 { 486 if (*pTable->LocaleID == *LocaleID) /* Check only first char for speed */ 487 { 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; 491 } 492 else 493 if (*pTable->LocaleID > *LocaleID) /* Sorted alphabetically - exit */ 494 break; 495 pTable++; 496 } 497 return ULMBCS_GRP_L1; 498} 499 500 501/* 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 512*/ 513 514typedef struct 515 { 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 */ 519 } 520UConverterDataLMBCS; 521 522 523#define DECLARE_LMBCS_DATA(n) \ 524static const UConverterImpl _LMBCSImpl##n={\ 525 UCNV_LMBCS_##n,\ 526 NULL,NULL,\ 527 _LMBCSOpen##n,\ 528 _LMBCSClose,\ 529 NULL,\ 530 _LMBCSToUnicodeWithOffsets,\ 531 _LMBCSToUnicodeWithOffsets,\ 532 _LMBCSFromUnicode,\ 533 _LMBCSFromUnicode,\ 534 NULL,\ 535 NULL,\ 536 NULL,\ 537 NULL,\ 538 _LMBCSSafeClone,\ 539 ucnv_getCompleteUnicodeSet\ 540};\ 541static const UConverterStaticData _LMBCSStaticData##n={\ 542 sizeof(UConverterStaticData),\ 543 "LMBCS-" #n,\ 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} \ 546};\ 547const UConverterSharedData _LMBCSData##n={\ 548 sizeof(UConverterSharedData), ~((uint32_t) 0),\ 549 NULL, NULL, &_LMBCSStaticData##n, FALSE, &_LMBCSImpl##n, \ 550 0 \ 551}; 552 553 /* The only function we needed to duplicate 12 times was the 'open' 554function, which will do basically the same thing except set a different 555optimization group. So, we put the common stuff into a worker function, 556and set up another macro to stamp out the 12 open functions:*/ 557#define DEFINE_LMBCS_OPEN(n) \ 558static void \ 559 _LMBCSOpen##n(UConverter* _this,const char* name,const char* locale,uint32_t options,UErrorCode* err) \ 560{ _LMBCSOpenWorker(_this, name,locale,options, err, n);} 561 562 563 564/* Here's the open worker & the common close function */ 565static void 566_LMBCSOpenWorker(UConverter* _this, 567 const char* name, 568 const char* locale, 569 uint32_t options, 570 UErrorCode* err, 571 ulmbcs_byte_t OptGroup 572 ) 573{ 574 UConverterDataLMBCS * extraInfo = (UConverterDataLMBCS*)uprv_malloc (sizeof (UConverterDataLMBCS)); 575 if(extraInfo != NULL) 576 { 577 ulmbcs_byte_t i; 578 579 uprv_memset(extraInfo, 0, sizeof(UConverterDataLMBCS)); 580 581 for (i=0; i <= ULMBCS_GRP_LAST && U_SUCCESS(*err); i++) 582 { 583 if(OptGroupByteToCPName[i] != NULL) { 584 extraInfo->OptGrpConverter[i] = ucnv_loadSharedData(OptGroupByteToCPName[i], NULL, err); 585 } 586 } 587 588 if(U_SUCCESS(*err)) { 589 extraInfo->OptGroup = OptGroup; 590 extraInfo->localeConverterIndex = FindLMBCSLocale(locale); 591 } else { 592 /* one of the subconverters could not be loaded, unload the previous ones */ 593 while(i > 0) { 594 if(extraInfo->OptGrpConverter[--i] != NULL) { 595 ucnv_unloadSharedDataIfReady(extraInfo->OptGrpConverter[i]); 596 extraInfo->OptGrpConverter[i] = NULL; 597 } 598 } 599 } 600 } 601 else 602 { 603 *err = U_MEMORY_ALLOCATION_ERROR; 604 } 605 _this->extraInfo = extraInfo; 606} 607 608static void 609_LMBCSClose(UConverter * _this) 610{ 611 if (_this->extraInfo != NULL) 612 { 613 ulmbcs_byte_t Ix; 614 UConverterDataLMBCS * extraInfo = (UConverterDataLMBCS *) _this->extraInfo; 615 616 for (Ix=0; Ix <= ULMBCS_GRP_LAST; Ix++) 617 { 618 if (extraInfo->OptGrpConverter[Ix] != NULL) 619 ucnv_unloadSharedDataIfReady(extraInfo->OptGrpConverter[Ix]); 620 } 621 if (!_this->isExtraLocal) { 622 uprv_free (_this->extraInfo); 623 } 624 } 625} 626 627typedef struct LMBCSClone { 628 UConverter cnv; 629 UConverterDataLMBCS lmbcs; 630} LMBCSClone; 631 632static UConverter * 633_LMBCSSafeClone(const UConverter *cnv, 634 void *stackBuffer, 635 int32_t *pBufferSize, 636 UErrorCode *status) { 637 LMBCSClone *newLMBCS; 638 UConverterDataLMBCS *extraInfo; 639 int32_t i; 640 641 if(*pBufferSize<=0) { 642 *pBufferSize=(int32_t)sizeof(LMBCSClone); 643 return NULL; 644 } 645 646 extraInfo=(UConverterDataLMBCS *)cnv->extraInfo; 647 newLMBCS=(LMBCSClone *)stackBuffer; 648 649 /* ucnv.c/ucnv_safeClone() copied the main UConverter already */ 650 651 uprv_memcpy(&newLMBCS->lmbcs, extraInfo, sizeof(UConverterDataLMBCS)); 652 653 /* share the subconverters */ 654 for(i = 0; i <= ULMBCS_GRP_LAST; ++i) { 655 if(extraInfo->OptGrpConverter[i] != NULL) { 656 ucnv_incrementRefCount(extraInfo->OptGrpConverter[i]); 657 } 658 } 659 660 newLMBCS->cnv.extraInfo = &newLMBCS->lmbcs; 661 newLMBCS->cnv.isExtraLocal = TRUE; 662 return &newLMBCS->cnv; 663} 664 665/* 666 * There used to be a _LMBCSGetUnicodeSet() function here (up to svn revision 20117) 667 * which added all code points except for U+F6xx 668 * because those cannot be represented in the Unicode group. 669 * However, it turns out that windows-950 has roundtrips for all of U+F6xx 670 * which means that LMBCS can convert all Unicode code points after all. 671 * We now simply use ucnv_getCompleteUnicodeSet(). 672 */ 673 674/* 675 Here's the basic helper function that we use when converting from 676 Unicode to LMBCS, and we suspect that a Unicode character will fit into 677 one of the 12 groups. The return value is the number of bytes written 678 starting at pStartLMBCS (if any). 679*/ 680 681static size_t 682LMBCSConversionWorker ( 683 UConverterDataLMBCS * extraInfo, /* subconverters, opt & locale groups */ 684 ulmbcs_byte_t group, /* The group to try */ 685 ulmbcs_byte_t * pStartLMBCS, /* where to put the results */ 686 UChar * pUniChar, /* The input unicode character */ 687 ulmbcs_byte_t * lastConverterIndex, /* output: track last successful group used */ 688 UBool * groups_tried /* output: track any unsuccessful groups */ 689) 690{ 691 ulmbcs_byte_t * pLMBCS = pStartLMBCS; 692 UConverterSharedData * xcnv = extraInfo->OptGrpConverter[group]; 693 694 int bytesConverted; 695 uint32_t value; 696 ulmbcs_byte_t firstByte; 697 698 U_ASSERT(xcnv); 699 U_ASSERT(group<ULMBCS_GRP_UNICODE); 700 701 bytesConverted = ucnv_MBCSFromUChar32(xcnv, *pUniChar, &value, FALSE); 702 703 /* get the first result byte */ 704 if(bytesConverted > 0) { 705 firstByte = (ulmbcs_byte_t)(value >> ((bytesConverted - 1) * 8)); 706 } else { 707 /* most common failure mode is an unassigned character */ 708 groups_tried[group] = TRUE; 709 return 0; 710 } 711 712 *lastConverterIndex = group; 713 714 /* All initial byte values in lower ascii range should have been caught by now, 715 except with the exception group. 716 */ 717 U_ASSERT((firstByte <= ULMBCS_C0END) || (firstByte >= ULMBCS_C1START) || (group == ULMBCS_GRP_EXCEPT)); 718 719 /* use converted data: first write 0, 1 or two group bytes */ 720 if (group != ULMBCS_GRP_EXCEPT && extraInfo->OptGroup != group) 721 { 722 *pLMBCS++ = group; 723 if (bytesConverted == 1 && group >= ULMBCS_DOUBLEOPTGROUP_START) 724 { 725 *pLMBCS++ = group; 726 } 727 } 728 729 /* don't emit control chars */ 730 if ( bytesConverted == 1 && firstByte < 0x20 ) 731 return 0; 732 733 734 /* then move over the converted data */ 735 switch(bytesConverted) 736 { 737 case 4: 738 *pLMBCS++ = (ulmbcs_byte_t)(value >> 24); 739 case 3: 740 *pLMBCS++ = (ulmbcs_byte_t)(value >> 16); 741 case 2: 742 *pLMBCS++ = (ulmbcs_byte_t)(value >> 8); 743 case 1: 744 *pLMBCS++ = (ulmbcs_byte_t)value; 745 default: 746 /* will never occur */ 747 break; 748 } 749 750 return (pLMBCS - pStartLMBCS); 751} 752 753 754/* This is a much simpler version of above, when we 755know we are writing LMBCS using the Unicode group 756*/ 757static size_t 758LMBCSConvertUni(ulmbcs_byte_t * pLMBCS, UChar uniChar) 759{ 760 /* encode into LMBCS Unicode range */ 761 uint8_t LowCh = (uint8_t)(uniChar & 0x00FF); 762 uint8_t HighCh = (uint8_t)(uniChar >> 8); 763 764 *pLMBCS++ = ULMBCS_GRP_UNICODE; 765 766 if (LowCh == 0) 767 { 768 *pLMBCS++ = ULMBCS_UNICOMPATZERO; 769 *pLMBCS++ = HighCh; 770 } 771 else 772 { 773 *pLMBCS++ = HighCh; 774 *pLMBCS++ = LowCh; 775 } 776 return ULMBCS_UNICODE_SIZE; 777} 778 779 780 781/* The main Unicode to LMBCS conversion function */ 782static void 783_LMBCSFromUnicode(UConverterFromUnicodeArgs* args, 784 UErrorCode* err) 785{ 786 ulmbcs_byte_t lastConverterIndex = 0; 787 UChar uniChar; 788 ulmbcs_byte_t LMBCS[ULMBCS_CHARSIZE_MAX]; 789 ulmbcs_byte_t * pLMBCS; 790 int bytes_written; 791 UBool groups_tried[ULMBCS_GRP_LAST+1]; 792 UConverterDataLMBCS * extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo; 793 int sourceIndex = 0; 794 795 796 /* Basic strategy: attempt to fill in local LMBCS 1-char buffer.(LMBCS) 797 If that succeeds, see if it will all fit into the target & copy it over 798 if it does. 799 800 We try conversions in the following order: 801 802 1. Single-byte ascii & special fixed control chars (&null) 803 2. Look up group in table & try that (could be 804 A) Unicode group 805 B) control group, 806 C) national encoding, 807 or ambiguous SBCS or MBCS group (on to step 4...) 808 809 3. If its ambiguous, try this order: 810 A) The optimization group 811 B) The locale group 812 C) The last group that succeeded with this string. 813 D) every other group that's relevent (single or double) 814 E) If its single-byte ambiguous, try the exceptions group 815 816 4. And as a grand fallback: Unicode 817 */ 818 819 while (args->source < args->sourceLimit && !U_FAILURE(*err)) 820 { 821 if (args->target >= args->targetLimit) 822 { 823 *err = U_BUFFER_OVERFLOW_ERROR; 824 break; 825 } 826 uniChar = *(args->source); 827 bytes_written = 0; 828 pLMBCS = LMBCS; 829 830 /* check cases in rough order of how common they are, for speed */ 831 832 /* single byte matches: strategy 1 */ 833 834 if (((uniChar > ULMBCS_C0END) && (uniChar < ULMBCS_C1START)) || 835 uniChar == 0 || uniChar == ULMBCS_HT || uniChar == ULMBCS_CR || 836 uniChar == ULMBCS_LF || uniChar == ULMBCS_123SYSTEMRANGE 837 ) 838 { 839 *pLMBCS++ = (ulmbcs_byte_t ) uniChar; 840 bytes_written = 1; 841 } 842 843 844 if (!bytes_written) 845 { 846 /* Check by UNICODE range (Strategy 2) */ 847 ulmbcs_byte_t group = FindLMBCSUniRange(uniChar); 848 849 if (group == ULMBCS_GRP_UNICODE) /* (Strategy 2A) */ 850 { 851 pLMBCS += LMBCSConvertUni(pLMBCS,uniChar); 852 853 bytes_written = pLMBCS - LMBCS; 854 } 855 else if (group == ULMBCS_GRP_CTRL) /* (Strategy 2B) */ 856 { 857 /* Handle control characters here */ 858 if (uniChar <= ULMBCS_C0END) 859 { 860 *pLMBCS++ = ULMBCS_GRP_CTRL; 861 *pLMBCS++ = (ulmbcs_byte_t)(ULMBCS_CTRLOFFSET + uniChar); 862 } 863 else if (uniChar >= ULMBCS_C1START && uniChar <= ULMBCS_C1START + ULMBCS_CTRLOFFSET) 864 { 865 *pLMBCS++ = ULMBCS_GRP_CTRL; 866 *pLMBCS++ = (ulmbcs_byte_t ) (uniChar & 0x00FF); 867 } 868 bytes_written = pLMBCS - LMBCS; 869 } 870 else if (group < ULMBCS_GRP_UNICODE) /* (Strategy 2C) */ 871 { 872 /* a specific converter has been identified - use it */ 873 bytes_written = LMBCSConversionWorker ( 874 extraInfo, group, pLMBCS, &uniChar, 875 &lastConverterIndex, groups_tried); 876 } 877 if (!bytes_written) /* the ambiguous group cases (Strategy 3) */ 878 { 879 uprv_memset(groups_tried, 0, sizeof(groups_tried)); 880 881 /* check for non-default optimization group (Strategy 3A )*/ 882 if (extraInfo->OptGroup != 1 883 && ULMBCS_AMBIGUOUS_MATCH(group, extraInfo->OptGroup)) 884 { 885 bytes_written = LMBCSConversionWorker (extraInfo, 886 extraInfo->OptGroup, pLMBCS, &uniChar, 887 &lastConverterIndex, groups_tried); 888 } 889 /* check for locale optimization group (Strategy 3B) */ 890 if (!bytes_written 891 && (extraInfo->localeConverterIndex) 892 && (ULMBCS_AMBIGUOUS_MATCH(group, extraInfo->localeConverterIndex))) 893 { 894 bytes_written = LMBCSConversionWorker (extraInfo, 895 extraInfo->localeConverterIndex, pLMBCS, &uniChar, 896 &lastConverterIndex, groups_tried); 897 } 898 /* check for last optimization group used for this string (Strategy 3C) */ 899 if (!bytes_written 900 && (lastConverterIndex) 901 && (ULMBCS_AMBIGUOUS_MATCH(group, lastConverterIndex))) 902 { 903 bytes_written = LMBCSConversionWorker (extraInfo, 904 lastConverterIndex, pLMBCS, &uniChar, 905 &lastConverterIndex, groups_tried); 906 907 } 908 if (!bytes_written) 909 { 910 /* just check every possible matching converter (Strategy 3D) */ 911 ulmbcs_byte_t grp_start; 912 ulmbcs_byte_t grp_end; 913 ulmbcs_byte_t grp_ix; 914 grp_start = (ulmbcs_byte_t)((group == ULMBCS_AMBIGUOUS_MBCS) 915 ? ULMBCS_DOUBLEOPTGROUP_START 916 : ULMBCS_GRP_L1); 917 grp_end = (ulmbcs_byte_t)((group == ULMBCS_AMBIGUOUS_MBCS) 918 ? ULMBCS_GRP_LAST 919 : ULMBCS_GRP_TH); 920 for (grp_ix = grp_start; 921 grp_ix <= grp_end && !bytes_written; 922 grp_ix++) 923 { 924 if (extraInfo->OptGrpConverter [grp_ix] && !groups_tried [grp_ix]) 925 { 926 bytes_written = LMBCSConversionWorker (extraInfo, 927 grp_ix, pLMBCS, &uniChar, 928 &lastConverterIndex, groups_tried); 929 } 930 } 931 /* a final conversion fallback to the exceptions group if its likely 932 to be single byte (Strategy 3E) */ 933 if (!bytes_written && grp_start == ULMBCS_GRP_L1) 934 { 935 bytes_written = LMBCSConversionWorker (extraInfo, 936 ULMBCS_GRP_EXCEPT, pLMBCS, &uniChar, 937 &lastConverterIndex, groups_tried); 938 } 939 } 940 /* all of our other strategies failed. Fallback to Unicode. (Strategy 4)*/ 941 if (!bytes_written) 942 { 943 944 pLMBCS += LMBCSConvertUni(pLMBCS, uniChar); 945 bytes_written = pLMBCS - LMBCS; 946 } 947 } 948 } 949 950 /* we have a translation. increment source and write as much as posible to target */ 951 args->source++; 952 pLMBCS = LMBCS; 953 while (args->target < args->targetLimit && bytes_written--) 954 { 955 *(args->target)++ = *pLMBCS++; 956 if (args->offsets) 957 { 958 *(args->offsets)++ = sourceIndex; 959 } 960 } 961 sourceIndex++; 962 if (bytes_written > 0) 963 { 964 /* write any bytes that didn't fit in target to the error buffer, 965 common code will move this to target if we get called back with 966 enough target room 967 */ 968 uint8_t * pErrorBuffer = args->converter->charErrorBuffer; 969 *err = U_BUFFER_OVERFLOW_ERROR; 970 args->converter->charErrorBufferLength = (int8_t)bytes_written; 971 while (bytes_written--) 972 { 973 *pErrorBuffer++ = *pLMBCS++; 974 } 975 } 976 } 977} 978 979 980/* Now, the Unicode from LMBCS section */ 981 982 983/* A function to call when we are looking at the Unicode group byte in LMBCS */ 984static UChar 985GetUniFromLMBCSUni(char const ** ppLMBCSin) /* Called with LMBCS-style Unicode byte stream */ 986{ 987 uint8_t HighCh = *(*ppLMBCSin)++; /* Big-endian Unicode in LMBCS compatibility group*/ 988 uint8_t LowCh = *(*ppLMBCSin)++; 989 990 if (HighCh == ULMBCS_UNICOMPATZERO ) 991 { 992 HighCh = LowCh; 993 LowCh = 0; /* zero-byte in LSB special character */ 994 } 995 return (UChar)((HighCh << 8) | LowCh); 996} 997 998 999 1000/* CHECK_SOURCE_LIMIT: Helper macro to verify that there are at least'index' 1001 bytes left in source up to sourceLimit.Errors appropriately if not. 1002 If we reach the limit, then update the source pointer to there to consume 1003 all input as required by ICU converter semantics. 1004*/ 1005 1006#define CHECK_SOURCE_LIMIT(index) \ 1007 if (args->source+index > args->sourceLimit){\ 1008 *err = U_TRUNCATED_CHAR_FOUND;\ 1009 args->source = args->sourceLimit;\ 1010 return 0xffff;} 1011 1012/* Return the Unicode representation for the current LMBCS character */ 1013 1014static UChar32 1015_LMBCSGetNextUCharWorker(UConverterToUnicodeArgs* args, 1016 UErrorCode* err) 1017{ 1018 UChar32 uniChar = 0; /* an output UNICODE char */ 1019 ulmbcs_byte_t CurByte; /* A byte from the input stream */ 1020 1021 /* error check */ 1022 if (args->source >= args->sourceLimit) 1023 { 1024 *err = U_ILLEGAL_ARGUMENT_ERROR; 1025 return 0xffff; 1026 } 1027 /* Grab first byte & save address for error recovery */ 1028 CurByte = *((ulmbcs_byte_t *) (args->source++)); 1029 1030 /* 1031 * at entry of each if clause: 1032 * 1. 'CurByte' points at the first byte of a LMBCS character 1033 * 2. '*source'points to the next byte of the source stream after 'CurByte' 1034 * 1035 * the job of each if clause is: 1036 * 1. set '*source' to point at the beginning of next char (nop if LMBCS char is only 1 byte) 1037 * 2. set 'uniChar' up with the right Unicode value, or set 'err' appropriately 1038 */ 1039 1040 /* First lets check the simple fixed values. */ 1041 1042 if(((CurByte > ULMBCS_C0END) && (CurByte < ULMBCS_C1START)) /* ascii range */ 1043 || (CurByte == 0) 1044 || CurByte == ULMBCS_HT || CurByte == ULMBCS_CR 1045 || CurByte == ULMBCS_LF || CurByte == ULMBCS_123SYSTEMRANGE) 1046 { 1047 uniChar = CurByte; 1048 } 1049 else 1050 { 1051 UConverterDataLMBCS * extraInfo; 1052 ulmbcs_byte_t group; 1053 UConverterSharedData *cnv; 1054 1055 if (CurByte == ULMBCS_GRP_CTRL) /* Control character group - no opt group update */ 1056 { 1057 ulmbcs_byte_t C0C1byte; 1058 CHECK_SOURCE_LIMIT(1); 1059 C0C1byte = *(args->source)++; 1060 uniChar = (C0C1byte < ULMBCS_C1START) ? C0C1byte - ULMBCS_CTRLOFFSET : C0C1byte; 1061 } 1062 else 1063 if (CurByte == ULMBCS_GRP_UNICODE) /* Unicode compatibility group: BigEndian UTF16 */ 1064 { 1065 CHECK_SOURCE_LIMIT(2); 1066 1067 /* don't check for error indicators fffe/ffff below */ 1068 return GetUniFromLMBCSUni(&(args->source)); 1069 } 1070 else if (CurByte <= ULMBCS_CTRLOFFSET) 1071 { 1072 group = CurByte; /* group byte is in the source */ 1073 extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo; 1074 if (group > ULMBCS_GRP_LAST || (cnv = extraInfo->OptGrpConverter[group]) == NULL) 1075 { 1076 /* this is not a valid group byte - no converter*/ 1077 *err = U_INVALID_CHAR_FOUND; 1078 } 1079 else if (group >= ULMBCS_DOUBLEOPTGROUP_START) /* double byte conversion */ 1080 { 1081 1082 CHECK_SOURCE_LIMIT(2); 1083 1084 /* check for LMBCS doubled-group-byte case */ 1085 if (*args->source == group) { 1086 /* single byte */ 1087 ++args->source; 1088 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source, 1, FALSE); 1089 ++args->source; 1090 } else { 1091 /* double byte */ 1092 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source, 2, FALSE); 1093 args->source += 2; 1094 } 1095 } 1096 else { /* single byte conversion */ 1097 CHECK_SOURCE_LIMIT(1); 1098 CurByte = *(args->source)++; 1099 1100 if (CurByte >= ULMBCS_C1START) 1101 { 1102 uniChar = _MBCS_SINGLE_SIMPLE_GET_NEXT_BMP(cnv, CurByte); 1103 } 1104 else 1105 { 1106 /* The non-optimizable oddballs where there is an explicit byte 1107 * AND the second byte is not in the upper ascii range 1108 */ 1109 char bytes[2]; 1110 1111 extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo; 1112 cnv = extraInfo->OptGrpConverter [ULMBCS_GRP_EXCEPT]; 1113 1114 /* Lookup value must include opt group */ 1115 bytes[0] = group; 1116 bytes[1] = CurByte; 1117 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, bytes, 2, FALSE); 1118 } 1119 } 1120 } 1121 else if (CurByte >= ULMBCS_C1START) /* group byte is implicit */ 1122 { 1123 extraInfo = (UConverterDataLMBCS *) args->converter->extraInfo; 1124 group = extraInfo->OptGroup; 1125 cnv = extraInfo->OptGrpConverter[group]; 1126 if (group >= ULMBCS_DOUBLEOPTGROUP_START) /* double byte conversion */ 1127 { 1128 if (!ucnv_MBCSIsLeadByte(cnv, CurByte)) 1129 { 1130 CHECK_SOURCE_LIMIT(0); 1131 1132 /* let the MBCS conversion consume CurByte again */ 1133 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source - 1, 1, FALSE); 1134 } 1135 else 1136 { 1137 CHECK_SOURCE_LIMIT(1); 1138 /* let the MBCS conversion consume CurByte again */ 1139 uniChar = ucnv_MBCSSimpleGetNextUChar(cnv, args->source - 1, 2, FALSE); 1140 ++args->source; 1141 } 1142 } 1143 else /* single byte conversion */ 1144 { 1145 uniChar = _MBCS_SINGLE_SIMPLE_GET_NEXT_BMP(cnv, CurByte); 1146 } 1147 } 1148 } 1149 return uniChar; 1150} 1151 1152 1153/* The exported function that converts lmbcs to one or more 1154 UChars - currently UTF-16 1155*/ 1156static void 1157_LMBCSToUnicodeWithOffsets(UConverterToUnicodeArgs* args, 1158 UErrorCode* err) 1159{ 1160 char LMBCS [ULMBCS_CHARSIZE_MAX]; 1161 UChar uniChar; /* one output UNICODE char */ 1162 const char * saveSource; /* beginning of current code point */ 1163 const char * pStartLMBCS = args->source; /* beginning of whole string */ 1164 const char * errSource = NULL; /* pointer to actual input in case an error occurs */ 1165 int8_t savebytes = 0; 1166 1167 /* Process from source to limit, or until error */ 1168 while (U_SUCCESS(*err) && args->sourceLimit > args->source && args->targetLimit > args->target) 1169 { 1170 saveSource = args->source; /* beginning of current code point */ 1171 1172 if (args->converter->toULength) /* reassemble char from previous call */ 1173 { 1174 const char *saveSourceLimit; 1175 size_t size_old = args->converter->toULength; 1176 1177 /* limit from source is either remainder of temp buffer, or user limit on source */ 1178 size_t size_new_maybe_1 = sizeof(LMBCS) - size_old; 1179 size_t size_new_maybe_2 = args->sourceLimit - args->source; 1180 size_t size_new = (size_new_maybe_1 < size_new_maybe_2) ? size_new_maybe_1 : size_new_maybe_2; 1181 1182 1183 uprv_memcpy(LMBCS, args->converter->toUBytes, size_old); 1184 uprv_memcpy(LMBCS + size_old, args->source, size_new); 1185 saveSourceLimit = args->sourceLimit; 1186 args->source = errSource = LMBCS; 1187 args->sourceLimit = LMBCS+size_old+size_new; 1188 savebytes = (int8_t)(size_old+size_new); 1189 uniChar = (UChar) _LMBCSGetNextUCharWorker(args, err); 1190 args->source = saveSource + ((args->source - LMBCS) - size_old); 1191 args->sourceLimit = saveSourceLimit; 1192 1193 if (*err == U_TRUNCATED_CHAR_FOUND) 1194 { 1195 /* evil special case: source buffers so small a char spans more than 2 buffers */ 1196 args->converter->toULength = savebytes; 1197 uprv_memcpy(args->converter->toUBytes, LMBCS, savebytes); 1198 args->source = args->sourceLimit; 1199 *err = U_ZERO_ERROR; 1200 return; 1201 } 1202 else 1203 { 1204 /* clear the partial-char marker */ 1205 args->converter->toULength = 0; 1206 } 1207 } 1208 else 1209 { 1210 errSource = saveSource; 1211 uniChar = (UChar) _LMBCSGetNextUCharWorker(args, err); 1212 savebytes = (int8_t)(args->source - saveSource); 1213 } 1214 if (U_SUCCESS(*err)) 1215 { 1216 if (uniChar < 0xfffe) 1217 { 1218 *(args->target)++ = uniChar; 1219 if(args->offsets) 1220 { 1221 *(args->offsets)++ = saveSource - pStartLMBCS; 1222 } 1223 } 1224 else if (uniChar == 0xfffe) 1225 { 1226 *err = U_INVALID_CHAR_FOUND; 1227 } 1228 else /* if (uniChar == 0xffff) */ 1229 { 1230 *err = U_ILLEGAL_CHAR_FOUND; 1231 } 1232 } 1233 } 1234 /* if target ran out before source, return U_BUFFER_OVERFLOW_ERROR */ 1235 if (U_SUCCESS(*err) && args->sourceLimit > args->source && args->targetLimit <= args->target) 1236 { 1237 *err = U_BUFFER_OVERFLOW_ERROR; 1238 } 1239 else if (U_FAILURE(*err)) 1240 { 1241 /* If character incomplete or unmappable/illegal, store it in toUBytes[] */ 1242 args->converter->toULength = savebytes; 1243 if (savebytes > 0) { 1244 uprv_memcpy(args->converter->toUBytes, errSource, savebytes); 1245 } 1246 if (*err == U_TRUNCATED_CHAR_FOUND) { 1247 *err = U_ZERO_ERROR; 1248 } 1249 } 1250} 1251 1252/* And now, the macroized declarations of data & functions: */ 1253DEFINE_LMBCS_OPEN(1) 1254DEFINE_LMBCS_OPEN(2) 1255DEFINE_LMBCS_OPEN(3) 1256DEFINE_LMBCS_OPEN(4) 1257DEFINE_LMBCS_OPEN(5) 1258DEFINE_LMBCS_OPEN(6) 1259DEFINE_LMBCS_OPEN(8) 1260DEFINE_LMBCS_OPEN(11) 1261DEFINE_LMBCS_OPEN(16) 1262DEFINE_LMBCS_OPEN(17) 1263DEFINE_LMBCS_OPEN(18) 1264DEFINE_LMBCS_OPEN(19) 1265 1266 1267DECLARE_LMBCS_DATA(1) 1268DECLARE_LMBCS_DATA(2) 1269DECLARE_LMBCS_DATA(3) 1270DECLARE_LMBCS_DATA(4) 1271DECLARE_LMBCS_DATA(5) 1272DECLARE_LMBCS_DATA(6) 1273DECLARE_LMBCS_DATA(8) 1274DECLARE_LMBCS_DATA(11) 1275DECLARE_LMBCS_DATA(16) 1276DECLARE_LMBCS_DATA(17) 1277DECLARE_LMBCS_DATA(18) 1278DECLARE_LMBCS_DATA(19) 1279 1280#endif /* #if !UCONFIG_NO_LEGACY_CONVERSION */ 1281