J. Duke | 319a3b9 | 2007-12-01 00:00:00 +0000 | [diff] [blame^] | 1 | /* |
| 2 | * Portions Copyright 2003-2006 Sun Microsystems, Inc. All Rights Reserved. |
| 3 | * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. |
| 4 | * |
| 5 | * This code is free software; you can redistribute it and/or modify it |
| 6 | * under the terms of the GNU General Public License version 2 only, as |
| 7 | * published by the Free Software Foundation. Sun designates this |
| 8 | * particular file as subject to the "Classpath" exception as provided |
| 9 | * by Sun in the LICENSE file that accompanied this code. |
| 10 | * |
| 11 | * This code is distributed in the hope that it will be useful, but WITHOUT |
| 12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
| 13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
| 14 | * version 2 for more details (a copy is included in the LICENSE file that |
| 15 | * accompanied this code). |
| 16 | * |
| 17 | * You should have received a copy of the GNU General Public License version |
| 18 | * 2 along with this work; if not, write to the Free Software Foundation, |
| 19 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. |
| 20 | * |
| 21 | * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, |
| 22 | * CA 95054 USA or visit www.sun.com if you need additional information or |
| 23 | * have any questions. |
| 24 | */ |
| 25 | |
| 26 | /* |
| 27 | ******************************************************************************* |
| 28 | * (C) Copyright IBM Corp. 1996-2005 - All Rights Reserved * |
| 29 | * * |
| 30 | * The original version of this source code and documentation is copyrighted * |
| 31 | * and owned by IBM, These materials are provided under terms of a License * |
| 32 | * Agreement between IBM and Sun. This technology is protected by multiple * |
| 33 | * US and International patents. This notice and attribution to IBM may not * |
| 34 | * to removed. * |
| 35 | ******************************************************************************* |
| 36 | */ |
| 37 | |
| 38 | package sun.text.normalizer; |
| 39 | |
| 40 | import java.io.DataInputStream; |
| 41 | import java.io.InputStream; |
| 42 | import java.io.IOException; |
| 43 | |
| 44 | /** |
| 45 | * @author Ram Viswanadha |
| 46 | */ |
| 47 | |
| 48 | /* |
| 49 | * Description of the format of unorm.icu version 2.1. |
| 50 | * |
| 51 | * Main change from version 1 to version 2: |
| 52 | * Use of new, common Trie instead of normalization-specific tries. |
| 53 | * Change to version 2.1: add third/auxiliary trie with associated data. |
| 54 | * |
| 55 | * For more details of how to use the data structures see the code |
| 56 | * in unorm.cpp (runtime normalization code) and |
| 57 | * in gennorm.c and gennorm/store.c (build-time data generation). |
| 58 | * |
| 59 | * For the serialized format of Trie see Trie.c/TrieHeader. |
| 60 | * |
| 61 | * - Overall partition |
| 62 | * |
| 63 | * unorm.icu customarily begins with a UDataInfo structure, see udata.h and .c. |
| 64 | * After that there are the following structures: |
| 65 | * |
| 66 | * char indexes[INDEX_TOP]; -- INDEX_TOP=32, see enum in this file |
| 67 | * |
| 68 | * Trie normTrie; -- size in bytes=indexes[INDEX_TRIE_SIZE] |
| 69 | * |
| 70 | * char extraData[extraDataTop]; -- extraDataTop=indexes[INDEX_UCHAR_COUNT] |
| 71 | * extraData[0] contains the number of units for |
| 72 | * FC_NFKC_Closure (formatVersion>=2.1) |
| 73 | * |
| 74 | * char combiningTable[combiningTableTop]; -- combiningTableTop=indexes[INDEX_COMBINE_DATA_COUNT] |
| 75 | * combiningTableTop may include one 16-bit padding unit |
| 76 | * to make sure that fcdTrie is 32-bit-aligned |
| 77 | * |
| 78 | * Trie fcdTrie; -- size in bytes=indexes[INDEX_FCD_TRIE_SIZE] |
| 79 | * |
| 80 | * Trie auxTrie; -- size in bytes=indexes[INDEX_AUX_TRIE_SIZE] |
| 81 | * |
| 82 | * |
| 83 | * The indexes array contains lengths and sizes of the following arrays and structures |
| 84 | * as well as the following values: |
| 85 | * indexes[INDEX_COMBINE_FWD_COUNT]=combineFwdTop |
| 86 | * -- one more than the highest combining index computed for forward-only-combining characters |
| 87 | * indexes[INDEX_COMBINE_BOTH_COUNT]=combineBothTop-combineFwdTop |
| 88 | * -- number of combining indexes computed for both-ways-combining characters |
| 89 | * indexes[INDEX_COMBINE_BACK_COUNT]=combineBackTop-combineBothTop |
| 90 | * -- number of combining indexes computed for backward-only-combining characters |
| 91 | * |
| 92 | * indexes[INDEX_MIN_NF*_NO_MAYBE] (where *={ C, D, KC, KD }) |
| 93 | * -- first code point with a quick check NF* value of NO/MAYBE |
| 94 | * |
| 95 | * |
| 96 | * - Tries |
| 97 | * |
| 98 | * The main structures are two Trie tables ("compact arrays"), |
| 99 | * each with one index array and one data array. |
| 100 | * See Trie.h and Trie.c. |
| 101 | * |
| 102 | * |
| 103 | * - Tries in unorm.icu |
| 104 | * |
| 105 | * The first trie (normTrie above) |
| 106 | * provides data for the NF* quick checks and normalization. |
| 107 | * The second trie (fcdTrie above) provides data just for FCD checks. |
| 108 | * |
| 109 | * |
| 110 | * - norm32 data words from the first trie |
| 111 | * |
| 112 | * The norm32Table contains one 32-bit word "norm32" per code point. |
| 113 | * It contains the following bit fields: |
| 114 | * 31..16 extra data index, EXTRA_SHIFT is used to shift this field down |
| 115 | * if this index is <EXTRA_INDEX_TOP then it is an index into |
| 116 | * extraData[] where variable-length normalization data for this |
| 117 | * code point is found |
| 118 | * if this index is <EXTRA_INDEX_TOP+EXTRA_SURROGATE_TOP |
| 119 | * then this is a norm32 for a leading surrogate, and the index |
| 120 | * value is used together with the following trailing surrogate |
| 121 | * code unit in the second trie access |
| 122 | * if this index is >=EXTRA_INDEX_TOP+EXTRA_SURROGATE_TOP |
| 123 | * then this is a norm32 for a "special" character, |
| 124 | * i.e., the character is a Hangul syllable or a Jamo |
| 125 | * see EXTRA_HANGUL etc. |
| 126 | * generally, instead of extracting this index from the norm32 and |
| 127 | * comparing it with the above constants, |
| 128 | * the normalization code compares the entire norm32 value |
| 129 | * with MIN_SPECIAL, SURROGATES_TOP, MIN_HANGUL etc. |
| 130 | * |
| 131 | * 15..8 combining class (cc) according to UnicodeData.txt |
| 132 | * |
| 133 | * 7..6 COMBINES_ANY flags, used in composition to see if a character |
| 134 | * combines with any following or preceding character(s) |
| 135 | * at all |
| 136 | * 7 COMBINES_BACK |
| 137 | * 6 COMBINES_FWD |
| 138 | * |
| 139 | * 5..0 quick check flags, set for "no" or "maybe", with separate flags for |
| 140 | * each normalization form |
| 141 | * the higher bits are "maybe" flags; for NF*D there are no such flags |
| 142 | * the lower bits are "no" flags for all forms, in the same order |
| 143 | * as the "maybe" flags, |
| 144 | * which is (MSB to LSB): NFKD NFD NFKC NFC |
| 145 | * 5..4 QC_ANY_MAYBE |
| 146 | * 3..0 QC_ANY_NO |
| 147 | * see further related constants |
| 148 | * |
| 149 | * |
| 150 | * - Extra data per code point |
| 151 | * |
| 152 | * "Extra data" is referenced by the index in norm32. |
| 153 | * It is variable-length data. It is only present, and only those parts |
| 154 | * of it are, as needed for a given character. |
| 155 | * The norm32 extra data index is added to the beginning of extraData[] |
| 156 | * to get to a vector of 16-bit words with data at the following offsets: |
| 157 | * |
| 158 | * [-1] Combining index for composition. |
| 159 | * Stored only if norm32&COMBINES_ANY . |
| 160 | * [0] Lengths of the canonical and compatibility decomposition strings. |
| 161 | * Stored only if there are decompositions, i.e., |
| 162 | * if norm32&(QC_NFD|QC_NFKD) |
| 163 | * High byte: length of NFKD, or 0 if none |
| 164 | * Low byte: length of NFD, or 0 if none |
| 165 | * Each length byte also has another flag: |
| 166 | * Bit 7 of a length byte is set if there are non-zero |
| 167 | * combining classes (cc's) associated with the respective |
| 168 | * decomposition. If this flag is set, then the decomposition |
| 169 | * is preceded by a 16-bit word that contains the |
| 170 | * leading and trailing cc's. |
| 171 | * Bits 6..0 of a length byte are the length of the |
| 172 | * decomposition string, not counting the cc word. |
| 173 | * [1..n] NFD |
| 174 | * [n+1..] NFKD |
| 175 | * |
| 176 | * Each of the two decompositions consists of up to two parts: |
| 177 | * - The 16-bit words with the leading and trailing cc's. |
| 178 | * This is only stored if bit 7 of the corresponding length byte |
| 179 | * is set. In this case, at least one of the cc's is not zero. |
| 180 | * High byte: leading cc==cc of the first code point in the decomposition string |
| 181 | * Low byte: trailing cc==cc of the last code point in the decomposition string |
| 182 | * - The decomposition string in UTF-16, with length code units. |
| 183 | * |
| 184 | * |
| 185 | * - Combining indexes and combiningTable[] |
| 186 | * |
| 187 | * Combining indexes are stored at the [-1] offset of the extra data |
| 188 | * if the character combines forward or backward with any other characters. |
| 189 | * They are used for (re)composition in NF*C. |
| 190 | * Values of combining indexes are arranged according to whether a character |
| 191 | * combines forward, backward, or both ways: |
| 192 | * forward-only < both ways < backward-only |
| 193 | * |
| 194 | * The index values for forward-only and both-ways combining characters |
| 195 | * are indexes into the combiningTable[]. |
| 196 | * The index values for backward-only combining characters are simply |
| 197 | * incremented from the preceding index values to be unique. |
| 198 | * |
| 199 | * In the combiningTable[], a variable-length list |
| 200 | * of variable-length (back-index, code point) pair entries is stored |
| 201 | * for each forward-combining character. |
| 202 | * |
| 203 | * These back-indexes are the combining indexes of both-ways or backward-only |
| 204 | * combining characters that the forward-combining character combines with. |
| 205 | * |
| 206 | * Each list is sorted in ascending order of back-indexes. |
| 207 | * Each list is terminated with the last back-index having bit 15 set. |
| 208 | * |
| 209 | * Each pair (back-index, code point) takes up either 2 or 3 |
| 210 | * 16-bit words. |
| 211 | * The first word of a list entry is the back-index, with its bit 15 set if |
| 212 | * this is the last pair in the list. |
| 213 | * |
| 214 | * The second word contains flags in bits 15..13 that determine |
| 215 | * if there is a third word and how the combined character is encoded: |
| 216 | * 15 set if there is a third word in this list entry |
| 217 | * 14 set if the result is a supplementary character |
| 218 | * 13 set if the result itself combines forward |
| 219 | * |
| 220 | * According to these bits 15..14 of the second word, |
| 221 | * the result character is encoded as follows: |
| 222 | * 00 or 01 The result is <=0x1fff and stored in bits 12..0 of |
| 223 | * the second word. |
| 224 | * 10 The result is 0x2000..0xffff and stored in the third word. |
| 225 | * Bits 12..0 of the second word are not used. |
| 226 | * 11 The result is a supplementary character. |
| 227 | * Bits 9..0 of the leading surrogate are in bits 9..0 of |
| 228 | * the second word. |
| 229 | * Add 0xd800 to these bits to get the complete surrogate. |
| 230 | * Bits 12..10 of the second word are not used. |
| 231 | * The trailing surrogate is stored in the third word. |
| 232 | * |
| 233 | * |
| 234 | * - FCD trie |
| 235 | * |
| 236 | * The FCD trie is very simple. |
| 237 | * It is a folded trie with 16-bit data words. |
| 238 | * In each word, the high byte contains the leading cc of the character, |
| 239 | * and the low byte contains the trailing cc of the character. |
| 240 | * These cc's are the cc's of the first and last code points in the |
| 241 | * canonical decomposition of the character. |
| 242 | * |
| 243 | * Since all 16 bits are used for cc's, lead surrogates must be tested |
| 244 | * by checking the code unit instead of the trie data. |
| 245 | * This is done only if the 16-bit data word is not zero. |
| 246 | * If the code unit is a leading surrogate and the data word is not zero, |
| 247 | * then instead of cc's it contains the offset for the second trie lookup. |
| 248 | * |
| 249 | * |
| 250 | * - Auxiliary trie and data |
| 251 | * |
| 252 | * |
| 253 | * The auxiliary 16-bit trie contains data for additional properties. |
| 254 | * Bits |
| 255 | * 15..13 reserved |
| 256 | * 12 not NFC_Skippable (f) (formatVersion>=2.2) |
| 257 | * 11 flag: not a safe starter for canonical closure |
| 258 | * 10 composition exclusion |
| 259 | * 9.. 0 index into extraData[] to FC_NFKC_Closure string |
| 260 | * (not for lead surrogate), |
| 261 | * or lead surrogate offset (for lead surrogate, if 9..0 not zero) |
| 262 | * |
| 263 | * Conditions for "NF* Skippable" from Mark Davis' com.ibm.text.UCD.NFSkippable: |
| 264 | * (used in NormalizerTransliterator) |
| 265 | * |
| 266 | * A skippable character is |
| 267 | * a) unassigned, or ALL of the following: |
| 268 | * b) of combining class 0. |
| 269 | * c) not decomposed by this normalization form. |
| 270 | * AND if NFC or NFKC, |
| 271 | * d) can never compose with a previous character. |
| 272 | * e) can never compose with a following character. |
| 273 | * f) can never change if another character is added. |
| 274 | * Example: a-breve might satisfy all but f, but if you |
| 275 | * add an ogonek it changes to a-ogonek + breve |
| 276 | * |
| 277 | * a)..e) must be tested from norm32. |
| 278 | * Since f) is more complicated, the (not-)NFC_Skippable flag (f) is built |
| 279 | * into the auxiliary trie. |
| 280 | * The same bit is used for NFC and NFKC; (c) differs for them. |
| 281 | * As usual, we build the "not skippable" flags so that unassigned |
| 282 | * code points get a 0 bit. |
| 283 | * This bit is only valid after (a)..(e) test FALSE; test NFD_NO before (f) as well. |
| 284 | * Test Hangul LV syllables entirely in code. |
| 285 | * |
| 286 | * |
| 287 | * - FC_NFKC_Closure strings in extraData[] |
| 288 | * |
| 289 | * Strings are either stored as a single code unit or as the length |
| 290 | * followed by that many units. |
| 291 | * |
| 292 | */ |
| 293 | final class NormalizerDataReader implements ICUBinary.Authenticate { |
| 294 | |
| 295 | /** |
| 296 | * <p>Protected constructor.</p> |
| 297 | * @param inputStream ICU uprop.dat file input stream |
| 298 | * @exception IOException throw if data file fails authentication |
| 299 | * @draft 2.1 |
| 300 | */ |
| 301 | protected NormalizerDataReader(InputStream inputStream) |
| 302 | throws IOException{ |
| 303 | |
| 304 | unicodeVersion = ICUBinary.readHeader(inputStream, DATA_FORMAT_ID, this); |
| 305 | dataInputStream = new DataInputStream(inputStream); |
| 306 | } |
| 307 | |
| 308 | // protected methods ------------------------------------------------- |
| 309 | |
| 310 | protected int[] readIndexes(int length)throws IOException{ |
| 311 | int[] indexes = new int[length]; |
| 312 | //Read the indexes |
| 313 | for (int i = 0; i <length ; i++) { |
| 314 | indexes[i] = dataInputStream.readInt(); |
| 315 | } |
| 316 | return indexes; |
| 317 | } |
| 318 | /** |
| 319 | * <p>Reads unorm.icu, parse it into blocks of data to be stored in |
| 320 | * NormalizerImpl.</P |
| 321 | * @param normBytes |
| 322 | * @param fcdBytes |
| 323 | * @param auxBytes |
| 324 | * @param extraData |
| 325 | * @param combiningTable |
| 326 | * @exception thrown when data reading fails |
| 327 | * @draft 2.1 |
| 328 | */ |
| 329 | protected void read(byte[] normBytes, byte[] fcdBytes, byte[] auxBytes, |
| 330 | char[] extraData, char[] combiningTable) |
| 331 | throws IOException{ |
| 332 | |
| 333 | //Read the bytes that make up the normTrie |
| 334 | dataInputStream.read(normBytes); |
| 335 | |
| 336 | //normTrieStream= new ByteArrayInputStream(normBytes); |
| 337 | |
| 338 | //Read the extra data |
| 339 | for(int i=0;i<extraData.length;i++){ |
| 340 | extraData[i]=dataInputStream.readChar(); |
| 341 | } |
| 342 | |
| 343 | //Read the combining class table |
| 344 | for(int i=0; i<combiningTable.length; i++){ |
| 345 | combiningTable[i]=dataInputStream.readChar(); |
| 346 | } |
| 347 | |
| 348 | //Read the fcdTrie |
| 349 | dataInputStream.read(fcdBytes); |
| 350 | |
| 351 | |
| 352 | //Read the AuxTrie |
| 353 | dataInputStream.read(auxBytes); |
| 354 | } |
| 355 | |
| 356 | public byte[] getDataFormatVersion(){ |
| 357 | return DATA_FORMAT_VERSION; |
| 358 | } |
| 359 | |
| 360 | public boolean isDataVersionAcceptable(byte version[]) |
| 361 | { |
| 362 | return version[0] == DATA_FORMAT_VERSION[0] |
| 363 | && version[2] == DATA_FORMAT_VERSION[2] |
| 364 | && version[3] == DATA_FORMAT_VERSION[3]; |
| 365 | } |
| 366 | |
| 367 | public byte[] getUnicodeVersion(){ |
| 368 | return unicodeVersion; |
| 369 | } |
| 370 | // private data members ------------------------------------------------- |
| 371 | |
| 372 | |
| 373 | /** |
| 374 | * ICU data file input stream |
| 375 | */ |
| 376 | private DataInputStream dataInputStream; |
| 377 | |
| 378 | private byte[] unicodeVersion; |
| 379 | |
| 380 | /** |
| 381 | * File format version that this class understands. |
| 382 | * No guarantees are made if a older version is used |
| 383 | * see store.c of gennorm for more information and values |
| 384 | */ |
| 385 | private static final byte DATA_FORMAT_ID[] = {(byte)0x4E, (byte)0x6F, |
| 386 | (byte)0x72, (byte)0x6D}; |
| 387 | private static final byte DATA_FORMAT_VERSION[] = {(byte)0x2, (byte)0x2, |
| 388 | (byte)0x5, (byte)0x2}; |
| 389 | |
| 390 | } |