1 /* 2 ******************************************************************************* 3 * Copyright (C) 2006-2015, International Business Machines Corporation and 4 * others. All Rights Reserved. 5 ******************************************************************************* 6 */ 7 8 package com.ibm.icu.charset; 9 10 import java.io.IOException; 11 import java.nio.ByteBuffer; 12 import java.nio.CharBuffer; 13 import java.nio.IntBuffer; 14 15 import com.ibm.icu.charset.CharsetMBCS.MBCSHeader; 16 import com.ibm.icu.charset.CharsetMBCS.MBCSToUFallback; 17 import com.ibm.icu.charset.CharsetMBCS.UConverterMBCSTable; 18 import com.ibm.icu.impl.ICUBinary; 19 import com.ibm.icu.impl.InvalidFormatException; 20 21 /** 22 * ucnvmbcs.h 23 * 24 * ICU conversion (.cnv) data file structure, following the usual UDataInfo 25 * header. 26 * 27 * Format version: 6.2 28 * 29 * struct UConverterStaticData -- struct containing the converter name, IBM CCSID, 30 * min/max bytes per character, etc. 31 * see ucnv_bld.h 32 * 33 * -------------------- 34 * 35 * The static data is followed by conversionType-specific data structures. 36 * At the moment, there are only variations of MBCS converters. They all have 37 * the same toUnicode structures, while the fromUnicode structures for SBCS 38 * differ from those for other MBCS-style converters. 39 * 40 * _MBCSHeader.version 4.2 adds an optional conversion extension data structure. 41 * If it is present, then an ICU version reading header versions 4.0 or 4.1 42 * will be able to use the base table and ignore the extension. 43 * 44 * The unicodeMask in the static data is part of the base table data structure. 45 * Especially, the UCNV_HAS_SUPPLEMENTARY flag determines the length of the 46 * fromUnicode stage 1 array. 47 * The static data unicodeMask refers only to the base table's properties if 48 * a base table is included. 49 * In an extension-only file, the static data unicodeMask is 0. 50 * The extension data indexes have a separate field with the unicodeMask flags. 51 * 52 * MBCS-style data structure following the static data. 53 * Offsets are counted in bytes from the beginning of the MBCS header structure. 54 * Details about usage in comments in ucnvmbcs.c. 55 * 56 * struct _MBCSHeader (see the definition in this header file below) 57 * contains 32-bit fields as follows: 58 * 8 values: 59 * 0 uint8_t[4] MBCS version in UVersionInfo format (currently 4.2.0.0) 60 * 1 uint32_t countStates 61 * 2 uint32_t countToUFallbacks 62 * 3 uint32_t offsetToUCodeUnits 63 * 4 uint32_t offsetFromUTable 64 * 5 uint32_t offsetFromUBytes 65 * 6 uint32_t flags, bits: 66 * 31.. 8 offsetExtension -- _MBCSHeader.version 4.2 (ICU 2.8) and higher 67 * 0 for older versions and if 68 * there is not extension structure 69 * 7.. 0 outputType 70 * 7 uint32_t fromUBytesLength -- _MBCSHeader.version 4.1 (ICU 2.4) and higher 71 * counts bytes in fromUBytes[] 72 * 73 * if(outputType==MBCS_OUTPUT_EXT_ONLY) { 74 * -- base table name for extension-only table 75 * char baseTableName[variable]; -- with NUL plus padding for 4-alignment 76 * 77 * -- all _MBCSHeader fields except for version and flags are 0 78 * } else { 79 * -- normal base table with optional extension 80 * 81 * int32_t stateTable[countStates][256]; 82 * 83 * struct _MBCSToUFallback { (fallbacks are sorted by offset) 84 * uint32_t offset; 85 * UChar32 codePoint; 86 * } toUFallbacks[countToUFallbacks]; 87 * 88 * uint16_t unicodeCodeUnits[(offsetFromUTable-offsetToUCodeUnits)/2]; 89 * (padded to an even number of units) 90 * 91 * -- stage 1 tables 92 * if(staticData.unicodeMask&UCNV_HAS_SUPPLEMENTARY) { 93 * -- stage 1 table for all of Unicode 94 * uint16_t fromUTable[0x440]; (32-bit-aligned) 95 * } else { 96 * -- BMP-only tables have a smaller stage 1 table 97 * uint16_t fromUTable[0x40]; (32-bit-aligned) 98 * } 99 * 100 * -- stage 2 tables 101 * length determined by top of stage 1 and bottom of stage 3 tables 102 * if(outputType==MBCS_OUTPUT_1) { 103 * -- SBCS: pure indexes 104 * uint16_t stage 2 indexes[?]; 105 * } else { 106 * -- DBCS, MBCS, EBCDIC_STATEFUL, ...: roundtrip flags and indexes 107 * uint32_t stage 2 flags and indexes[?]; 108 * } 109 * 110 * -- stage 3 tables with byte results 111 * if(outputType==MBCS_OUTPUT_1) { 112 * -- SBCS: each 16-bit result contains flags and the result byte, see ucnvmbcs.c 113 * uint16_t fromUBytes[fromUBytesLength/2]; 114 * } else { 115 * -- DBCS, MBCS, EBCDIC_STATEFUL, ... 2/3/4 bytes result, see ucnvmbcs.c 116 * uint8_t fromUBytes[fromUBytesLength]; or 117 * uint16_t fromUBytes[fromUBytesLength/2]; or 118 * uint32_t fromUBytes[fromUBytesLength/4]; 119 * } 120 * } 121 * 122 * -- extension table, details see ucnv_ext.h 123 * int32_t indexes[>=32]; ... 124 */ 125 /* 126 * ucnv_ext.h 127 * 128 * See icuhtml/design/conversion/conversion_extensions.html 129 * 130 * Conversion extensions serve two purposes: 131 * 1. They support m:n mappings. 132 * 2. They support extension-only conversion files that are used together 133 * with the regular conversion data in base files. 134 * 135 * A base file may contain an extension table (explicitly requested or 136 * implicitly generated for m:n mappings), but its extension table is not 137 * used when an extension-only file is used. 138 * 139 * It is an error if a base file contains any regular (not extension) mapping 140 * from the same sequence as a mapping in the extension file 141 * because the base mapping would hide the extension mapping. 142 * 143 * 144 * Data for conversion extensions: 145 * 146 * One set of data structures per conversion direction (to/from Unicode). 147 * The data structures are sorted by input units to allow for binary search. 148 * Input sequences of more than one unit are handled like contraction tables 149 * in collation: 150 * The lookup value of a unit points to another table that is to be searched 151 * for the next unit, recursively. 152 * 153 * For conversion from Unicode, the initial code point is looked up in 154 * a 3-stage trie for speed, 155 * with an additional table of unique results to save space. 156 * 157 * Long output strings are stored in separate arrays, with length and index 158 * in the lookup tables. 159 * Output results also include a flag distinguishing roundtrip from 160 * (reverse) fallback mappings. 161 * 162 * Input Unicode strings must not begin or end with unpaired surrogates 163 * to avoid problems with matches on parts of surrogate pairs. 164 * 165 * Mappings from multiple characters (code points or codepage state 166 * table sequences) must be searched preferring the longest match. 167 * For this to work and be efficient, the variable-width table must contain 168 * all mappings that contain prefixes of the multiple characters. 169 * If an extension table is built on top of a base table in another file 170 * and a base table entry is a prefix of a multi-character mapping, then 171 * this is an error. 172 * 173 * 174 * Implementation note: 175 * 176 * Currently, the parser and several checks in the code limit the number 177 * of UChars or bytes in a mapping to 178 * UCNV_EXT_MAX_UCHARS and UCNV_EXT_MAX_BYTES, respectively, 179 * which are output value limits in the data structure. 180 * 181 * For input, this is not strictly necessary - it is a hard limit only for the 182 * buffers in UConverter that are used to store partial matches. 183 * 184 * Input sequences could otherwise be arbitrarily long if partial matches 185 * need not be stored (i.e., if a sequence does not span several buffers with too 186 * many units before the last buffer), although then results would differ 187 * depending on whether partial matches exceed the limits or not, 188 * which depends on the pattern of buffer sizes. 189 * 190 * 191 * Data structure: 192 * 193 * int32_t indexes[>=32]; 194 * 195 * Array of indexes and lengths etc. The length of the array is at least 32. 196 * The actual length is stored in indexes[0] to be forward compatible. 197 * 198 * Each index to another array is the number of bytes from indexes[]. 199 * Each length of an array is the number of array base units in that array. 200 * 201 * Some of the structures may not be present, in which case their indexes 202 * and lengths are 0. 203 * 204 * Usage of indexes[i]: 205 * [0] length of indexes[] 206 * 207 * // to Unicode table 208 * [1] index of toUTable[] (array of uint32_t) 209 * [2] length of toUTable[] 210 * [3] index of toUUChars[] (array of UChar) 211 * [4] length of toUUChars[] 212 * 213 * // from Unicode table, not for the initial code point 214 * [5] index of fromUTableUChars[] (array of UChar) 215 * [6] index of fromUTableValues[] (array of uint32_t) 216 * [7] length of fromUTableUChars[] and fromUTableValues[] 217 * [8] index of fromUBytes[] (array of char) 218 * [9] length of fromUBytes[] 219 * 220 * // from Unicode trie for initial-code point lookup 221 * [10] index of fromUStage12[] (combined array of uint16_t for stages 1 & 2) 222 * [11] length of stage 1 portion of fromUStage12[] 223 * [12] length of fromUStage12[] 224 * [13] index of fromUStage3[] (array of uint16_t indexes into fromUStage3b[]) 225 * [14] length of fromUStage3[] 226 * [15] index of fromUStage3b[] (array of uint32_t like fromUTableValues[]) 227 * [16] length of fromUStage3b[] 228 * 229 * [17] Bit field containing numbers of bytes: 230 * 31..24 reserved, 0 231 * 23..16 maximum input bytes 232 * 15.. 8 maximum output bytes 233 * 7.. 0 maximum bytes per UChar 234 * 235 * [18] Bit field containing numbers of UChars: 236 * 31..24 reserved, 0 237 * 23..16 maximum input UChars 238 * 15.. 8 maximum output UChars 239 * 7.. 0 maximum UChars per byte 240 * 241 * [19] Bit field containing flags: 242 * (extension table unicodeMask) 243 * 1 UCNV_HAS_SURROGATES flag for the extension table 244 * 0 UCNV_HAS_SUPPLEMENTARY flag for the extension table 245 * 246 * [20]..[30] reserved, 0 247 * [31] number of bytes for the entire extension structure 248 * [>31] reserved; there are indexes[0] indexes 249 * 250 * 251 * uint32_t toUTable[]; 252 * 253 * Array of byte/value pairs for lookups for toUnicode conversion. 254 * The array is partitioned into sections like collation contraction tables. 255 * Each section contains one word with the number of following words and 256 * a default value for when the lookup in this section yields no match. 257 * 258 * A section is sorted in ascending order of input bytes, 259 * allowing for fast linear or binary searches. 260 * The builder may store entries for a contiguous range of byte values 261 * (compare difference between the first and last one with count), 262 * which then allows for direct array access. 263 * The builder should always do this for the initial table section. 264 * 265 * Entries may have 0 values, see below. 266 * No two entries in a section have the same byte values. 267 * 268 * Each uint32_t contains an input byte value in bits 31..24 and the 269 * corresponding lookup value in bits 23..0. 270 * Interpret the value as follows: 271 * if(value==0) { 272 * no match, see below 273 * } else if(value<0x1f0000) { 274 * partial match - use value as index to the next toUTable section 275 * and match the next unit; (value indexes toUTable[value]) 276 * } else { 277 * if(bit 23 set) { 278 * roundtrip; 279 * } else { 280 * fallback; 281 * } 282 * unset value bit 23; 283 * if(value<=0x2fffff) { 284 * (value-0x1f0000) is a code point; (BMP: value<=0x1fffff) 285 * } else { 286 * bits 17..0 (value&0x3ffff) is an index to 287 * the result UChars in toUUChars[]; (0 indexes toUUChars[0]) 288 * length of the result=((value>>18)-12); (length=0..19) 289 * } 290 * } 291 * 292 * The first word in a section contains the number of following words in the 293 * input byte position (bits 31..24, number=1..0xff). 294 * The value of the initial word is used when the current byte is not found 295 * in this section. 296 * If the value is not 0, then it represents a result as above. 297 * If the value is 0, then the search has to return a shorter match with an 298 * earlier default value as the result, or result in "unmappable" even for the 299 * initial bytes. 300 * If the value is 0 for the initial toUTable entry, then the initial byte 301 * does not start any mapping input. 302 * 303 * 304 * UChar toUUChars[]; 305 * 306 * Contains toUnicode mapping results, stored as sequences of UChars. 307 * Indexes and lengths stored in the toUTable[]. 308 * 309 * 310 * UChar fromUTableUChars[]; 311 * uint32_t fromUTableValues[]; 312 * 313 * The fromUTable is split into two arrays, but works otherwise much like 314 * the toUTable. The array is partitioned into sections like collation 315 * contraction tables and toUTable. 316 * A row in the table consists of same-index entries in fromUTableUChars[] 317 * and fromUTableValues[]. 318 * 319 * Interpret a value as follows: 320 * if(value==0) { 321 * no match, see below 322 * } else if(value<=0xffffff) { (bits 31..24 are 0) 323 * partial match - use value as index to the next fromUTable section 324 * and match the next unit; (value indexes fromUTable[value]) 325 * } else { 326 * if(value==0x80000001) { 327 * return no mapping, but request for <subchar1>; 328 * } 329 * if(bit 31 set) { 330 * roundtrip; 331 * } else { 332 * fallback; 333 * } 334 * // bits 30..29 reserved, 0 335 * length=(value>>24)&0x1f; (bits 28..24) 336 * if(length==1..3) { 337 * bits 23..0 contain 1..3 bytes, padded with 00s on the left; 338 * } else { 339 * bits 23..0 (value&0xffffff) is an index to 340 * the result bytes in fromUBytes[]; (0 indexes fromUBytes[0]) 341 * } 342 * } 343 * 344 * The first pair in a section contains the number of following pairs in the 345 * UChar position (16 bits, number=1..0xffff). 346 * The value of the initial pair is used when the current UChar is not found 347 * in this section. 348 * If the value is not 0, then it represents a result as above. 349 * If the value is 0, then the search has to return a shorter match with an 350 * earlier default value as the result, or result in "unmappable" even for the 351 * initial UChars. 352 * 353 * If the from Unicode trie is present, then the from Unicode search tables 354 * are not used for initial code points. 355 * In this case, the first entries (index 0) in the tables are not used 356 * (reserved, set to 0) because a value of 0 is used in trie results 357 * to indicate no mapping. 358 * 359 * 360 * uint16_t fromUStage12[]; 361 * 362 * Stages 1 & 2 of a trie that maps an initial code point. 363 * Indexes in stage 1 are all offset by the length of stage 1 so that the 364 * same array pointer can be used for both stages. 365 * If (c>>10)>=(length of stage 1) then c does not start any mapping. 366 * Same bit distribution as for regular conversion tries. 367 * 368 * 369 * uint16_t fromUStage3[]; 370 * uint32_t fromUStage3b[]; 371 * 372 * Stage 3 of the trie. The first array simply contains indexes to the second, 373 * which contains words in the same format as fromUTableValues[]. 374 * Use a stage 3 granularity of 4, which allows for 256k stage 3 entries, 375 * and 16-bit entries in stage 3 allow for 64k stage 3b entries. 376 * The stage 3 granularity means that the stage 2 entry needs to be left-shifted. 377 * 378 * Two arrays are used because it is expected that more than half of the stage 3 379 * entries will be zero. The 16-bit index stage 3 array saves space even 380 * considering storing a total of 6 bytes per non-zero entry in both arrays 381 * together. 382 * Using a stage 3 granularity of >1 diminishes the compactability in that stage 383 * but provides a larger effective addressing space in stage 2. 384 * All but the final result stage use 16-bit entries to save space. 385 * 386 * fromUStage3b[] contains a zero for "no mapping" at its index 0, 387 * and may contain UCNV_EXT_FROM_U_SUBCHAR1 at index 1 for "<subchar1> SUB mapping" 388 * (i.e., "no mapping" with preference for <subchar1> rather than <subchar>), 389 * and all other items are unique non-zero results. 390 * 391 * The default value of a fromUTableValues[] section that is referenced 392 * _directly_ from a fromUStage3b[] item may also be UCNV_EXT_FROM_U_SUBCHAR1, 393 * but this value must not occur anywhere else in fromUTableValues[] 394 * because "no mapping" is always a property of a single code point, 395 * never of multiple. 396 * 397 * 398 * char fromUBytes[]; 399 * 400 * Contains fromUnicode mapping results, stored as sequences of chars. 401 * Indexes and lengths stored in the fromUTableValues[]. 402 */ 403 404 final class UConverterDataReader { 405 //private final static boolean debug = ICUDebug.enabled("UConverterDataReader"); 406 407 private static final class IsAcceptable implements ICUBinary.Authenticate { 408 // @Override when we switch to Java 6 409 public boolean isDataVersionAcceptable(byte formatVersion[]) { 410 return formatVersion[0] == 6; 411 } 412 } 413 private static final IsAcceptable IS_ACCEPTABLE = new IsAcceptable(); 414 415 /* 416 * UConverterDataReader(UConverterDataReader r) 417 { 418 byteBuffer = ICUBinary.getByteBufferFromInputStreamAndCloseStream(r.byteBuffer); 419 unicodeVersion = r.unicodeVersion; 420 } 421 */ 422 /** The buffer position after the static data. */ 423 private int posAfterStaticData; 424 425 /** 426 * <p>Protected constructor.</p> 427 * @param bytes ICU conversion data file 428 * @exception IOException throw if data file fails authentication 429 */ 430 protected UConverterDataReader(ByteBuffer bytes) 431 throws IOException{ 432 //if(debug) System.out.println("Bytes in buffer " + bytes.remaining()); 433 434 byteBuffer = bytes; 435 /*unicodeVersion = */ICUBinary.readHeader(byteBuffer, DATA_FORMAT_ID, IS_ACCEPTABLE); 436 437 //if(debug) System.out.println("Bytes left in byteBuffer " + byteBuffer.remaining()); 438 } 439 440 // protected methods ------------------------------------------------- 441 442 protected void readStaticData(UConverterStaticData sd) throws IOException 443 { 444 sd.structSize = byteBuffer.getInt(); 445 byte[] name = new byte[UConverterConstants.MAX_CONVERTER_NAME_LENGTH]; 446 byteBuffer.get(name); 447 sd.name = new String(name, "US-ASCII"); 448 sd.codepage = byteBuffer.getInt(); 449 sd.platform = byteBuffer.get(); 450 sd.conversionType = byteBuffer.get(); 451 sd.minBytesPerChar = byteBuffer.get(); 452 sd.maxBytesPerChar = byteBuffer.get(); 453 byteBuffer.get(sd.subChar); 454 sd.subCharLen = byteBuffer.get(); 455 sd.hasToUnicodeFallback = byteBuffer.get(); 456 sd.hasFromUnicodeFallback = byteBuffer.get(); 457 sd.unicodeMask = (short)(byteBuffer.get() & 0xff); 458 sd.subChar1 = byteBuffer.get(); 459 byteBuffer.get(sd.reserved); 460 posAfterStaticData = byteBuffer.position(); 461 } 462 463 int bytesReadAfterStaticData() { 464 return byteBuffer.position() - posAfterStaticData; 465 } 466 467 protected void readMBCSHeader(CharsetMBCS.MBCSHeader h) throws IOException 468 { 469 byteBuffer.get(h.version); 470 h.countStates = byteBuffer.getInt(); 471 h.countToUFallbacks = byteBuffer.getInt(); 472 h.offsetToUCodeUnits = byteBuffer.getInt(); 473 h.offsetFromUTable = byteBuffer.getInt(); 474 h.offsetFromUBytes = byteBuffer.getInt(); 475 h.flags = byteBuffer.getInt(); 476 h.fromUBytesLength = byteBuffer.getInt(); 477 if (h.version[0] == 5 && h.version[1] >= 3) { 478 h.options = byteBuffer.getInt(); 479 if ((h.options & CharsetMBCS.MBCS_OPT_NO_FROM_U) != 0) { 480 h.fullStage2Length = byteBuffer.getInt(); 481 } 482 } 483 } 484 485 protected void readMBCSTable(MBCSHeader header, UConverterMBCSTable mbcsTable) throws IOException 486 { 487 IntBuffer intBuffer = byteBuffer.asIntBuffer(); 488 mbcsTable.countStates = (byte) header.countStates; 489 mbcsTable.stateTable = new int[header.countStates][256]; 490 int i; 491 for(i = 0; i < header.countStates; ++i) { 492 intBuffer.get(mbcsTable.stateTable[i]); 493 } 494 495 mbcsTable.countToUFallbacks = header.countToUFallbacks; 496 mbcsTable.toUFallbacks = new MBCSToUFallback[header.countToUFallbacks]; 497 for(i = 0; i < header.countToUFallbacks; ++i) { 498 int offset = intBuffer.get(); 499 int codePoint = intBuffer.get(); 500 mbcsTable.toUFallbacks[i] = new MBCSToUFallback(offset, codePoint); 501 } 502 // Skip as many bytes as we have read from the IntBuffer. 503 int length = intBuffer.position() * 4; 504 ICUBinary.skipBytes(byteBuffer, length); 505 506 // Consider leaving some large arrays as CharBuffer/IntBuffer rather than 507 // reading them into Java arrays, to reduce initialization time and memory usage, 508 // at the cost of some performance. 509 // For example: unicodeCodeUnits, fromUnicodeTable, fromUnicodeInts. 510 // Take care not to modify the buffer contents for swaplfnl. 511 CharBuffer charBuffer = byteBuffer.asCharBuffer(); 512 length = header.offsetFromUTable - header.offsetToUCodeUnits; 513 assert (length & 1) == 0; 514 mbcsTable.unicodeCodeUnits = new char[length / 2]; 515 charBuffer.get(mbcsTable.unicodeCodeUnits); 516 // Skip as many bytes as we have read from the CharBuffer. 517 ICUBinary.skipBytes(byteBuffer, length); 518 519 length = header.offsetFromUBytes - header.offsetFromUTable; 520 assert (length & 1) == 0; 521 int fromUTableCharsLength; 522 if (mbcsTable.outputType == CharsetMBCS.MBCS_OUTPUT_1) { 523 // single-byte table stage1 + stage2 524 fromUTableCharsLength = length / 2; 525 } else if (mbcsTable.hasSupplementary()) { 526 // stage1 for Unicode limit 0x110000 >> 10 527 fromUTableCharsLength = 0x440; 528 } else { 529 // stage1 for BMP limit 0x10000 >> 10 530 fromUTableCharsLength = 0x40; 531 } 532 mbcsTable.fromUnicodeTable = new char[fromUTableCharsLength]; 533 charBuffer.get(mbcsTable.fromUnicodeTable); 534 if (mbcsTable.outputType != CharsetMBCS.MBCS_OUTPUT_1) { 535 // Read both stage1 and stage2 together into an int[] array. 536 // Keeping the short stage1 in the array avoids offsetting at runtime. 537 // The stage1 part of this array will not be used. 538 assert (length & 3) == 0; 539 mbcsTable.fromUnicodeTableInts = new int[length / 4]; 540 byteBuffer.asIntBuffer().get(mbcsTable.fromUnicodeTableInts); 541 } 542 // Skip as many bytes as are in stage1 + stage2. 543 ICUBinary.skipBytes(byteBuffer, length); 544 545 mbcsTable.fromUBytesLength = header.fromUBytesLength; 546 boolean noFromU = ((header.options & CharsetMBCS.MBCS_OPT_NO_FROM_U) != 0); 547 if (!noFromU) { 548 switch (mbcsTable.outputType) { 549 case CharsetMBCS.MBCS_OUTPUT_1: 550 case CharsetMBCS.MBCS_OUTPUT_2: 551 case CharsetMBCS.MBCS_OUTPUT_2_SISO: 552 case CharsetMBCS.MBCS_OUTPUT_3_EUC: 553 mbcsTable.fromUnicodeChars = ICUBinary.getChars( 554 byteBuffer, header.fromUBytesLength / 2, 0); 555 break; 556 case CharsetMBCS.MBCS_OUTPUT_3: 557 case CharsetMBCS.MBCS_OUTPUT_4_EUC: 558 mbcsTable.fromUnicodeBytes = new byte[header.fromUBytesLength]; 559 byteBuffer.get(mbcsTable.fromUnicodeBytes); 560 break; 561 case CharsetMBCS.MBCS_OUTPUT_4: 562 mbcsTable.fromUnicodeInts = ICUBinary.getInts( 563 byteBuffer, header.fromUBytesLength / 4, 0); 564 break; 565 default: 566 // Cannot occur, caller checked already. 567 assert false; 568 } 569 } else { 570 // Optional utf8Friendly mbcsIndex -- _MBCSHeader.version 4.3 (ICU 3.8) and higher. 571 // Needed for reconstituting omitted data. 572 mbcsTable.mbcsIndex = byteBuffer.asCharBuffer(); 573 } 574 } 575 576 protected String readBaseTableName() throws IOException 577 { 578 char c; 579 StringBuilder name = new StringBuilder(); 580 while((c = (char)byteBuffer.get()) != 0){ 581 name.append(c); 582 } 583 return name.toString(); 584 } 585 586 //protected int[] readExtIndexes(int skip) throws IOException 587 protected ByteBuffer readExtIndexes(int skip) throws IOException, InvalidFormatException 588 { 589 ICUBinary.skipBytes(byteBuffer, skip); 590 ByteBuffer b = ICUBinary.sliceWithOrder(byteBuffer); 591 int lengthOfIndexes = b.getInt(0); 592 if (lengthOfIndexes < 32) { 593 throw new InvalidFormatException(); 594 } 595 int numBytesExtensionStructure = b.getInt(31 * 4); 596 b.limit(numBytesExtensionStructure); 597 ICUBinary.skipBytes(byteBuffer, numBytesExtensionStructure); 598 return b; 599 } 600 601 /** 602 * Data formatVersion 6.1 and higher has a unicodeMask. 603 */ 604 boolean dataFormatHasUnicodeMask() { 605 int formatVersion0 = byteBuffer.get(16) & 0xff; 606 return formatVersion0 > 6 || (formatVersion0 == 6 && byteBuffer.get(17) != 0); 607 } 608 609 // private data members ------------------------------------------------- 610 611 /** 612 * ICU data file input stream 613 */ 614 private ByteBuffer byteBuffer; 615 616 // private VersionInfo unicodeVersion; 617 618 /** 619 * File format version that this class understands. 620 * No guarantees are made if a older version is used 621 * see store.c of gennorm for more information and values 622 */ 623 // DATA_FORMAT_ID_ values taken from icu4c isCnvAcceptable (ucnv_bld.c) 624 private static final int DATA_FORMAT_ID = 0x636e7674; // dataFormat="cnvt" 625 } 626
