1/*
2*******************************************************************************
3*
4*   Copyright (C) 2003-2012, International Business Machines
5*   Corporation and others.  All Rights Reserved.
6*
7*******************************************************************************
8*   file name:  ucol_swp.cpp
9*   encoding:   US-ASCII
10*   tab size:   8 (not used)
11*   indentation:4
12*
13*   created on: 2003sep10
14*   created by: Markus W. Scherer
15*
16*   Swap collation binaries.
17*/
18
19#include "unicode/udata.h" /* UDataInfo */
20#include "utrie.h"
21#include "udataswp.h"
22#include "cmemory.h"
23#include "ucol_data.h"
24#include "ucol_swp.h"
25
26/* swapping ----------------------------------------------------------------- */
27
28/*
29 * This performs data swapping for a folded trie (see utrie.c for details).
30 */
31
32U_CAPI int32_t U_EXPORT2
33utrie_swap(const UDataSwapper *ds,
34           const void *inData, int32_t length, void *outData,
35           UErrorCode *pErrorCode) {
36    const UTrieHeader *inTrie;
37    UTrieHeader trie;
38    int32_t size;
39    UBool dataIs32;
40
41    if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
42        return 0;
43    }
44    if(ds==NULL || inData==NULL || (length>=0 && outData==NULL)) {
45        *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
46        return 0;
47    }
48
49    /* setup and swapping */
50    if(length>=0 && (uint32_t)length<sizeof(UTrieHeader)) {
51        *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
52        return 0;
53    }
54
55    inTrie=(const UTrieHeader *)inData;
56    trie.signature=ds->readUInt32(inTrie->signature);
57    trie.options=ds->readUInt32(inTrie->options);
58    trie.indexLength=udata_readInt32(ds, inTrie->indexLength);
59    trie.dataLength=udata_readInt32(ds, inTrie->dataLength);
60
61    if( trie.signature!=0x54726965 ||
62        (trie.options&UTRIE_OPTIONS_SHIFT_MASK)!=UTRIE_SHIFT ||
63        ((trie.options>>UTRIE_OPTIONS_INDEX_SHIFT)&UTRIE_OPTIONS_SHIFT_MASK)!=UTRIE_INDEX_SHIFT ||
64        trie.indexLength<UTRIE_BMP_INDEX_LENGTH ||
65        (trie.indexLength&(UTRIE_SURROGATE_BLOCK_COUNT-1))!=0 ||
66        trie.dataLength<UTRIE_DATA_BLOCK_LENGTH ||
67        (trie.dataLength&(UTRIE_DATA_GRANULARITY-1))!=0 ||
68        ((trie.options&UTRIE_OPTIONS_LATIN1_IS_LINEAR)!=0 && trie.dataLength<(UTRIE_DATA_BLOCK_LENGTH+0x100))
69    ) {
70        *pErrorCode=U_INVALID_FORMAT_ERROR; /* not a UTrie */
71        return 0;
72    }
73
74    dataIs32=(UBool)((trie.options&UTRIE_OPTIONS_DATA_IS_32_BIT)!=0);
75    size=sizeof(UTrieHeader)+trie.indexLength*2+trie.dataLength*(dataIs32?4:2);
76
77    if(length>=0) {
78        UTrieHeader *outTrie;
79
80        if(length<size) {
81            *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
82            return 0;
83        }
84
85        outTrie=(UTrieHeader *)outData;
86
87        /* swap the header */
88        ds->swapArray32(ds, inTrie, sizeof(UTrieHeader), outTrie, pErrorCode);
89
90        /* swap the index and the data */
91        if(dataIs32) {
92            ds->swapArray16(ds, inTrie+1, trie.indexLength*2, outTrie+1, pErrorCode);
93            ds->swapArray32(ds, (const uint16_t *)(inTrie+1)+trie.indexLength, trie.dataLength*4,
94                                     (uint16_t *)(outTrie+1)+trie.indexLength, pErrorCode);
95        } else {
96            ds->swapArray16(ds, inTrie+1, (trie.indexLength+trie.dataLength)*2, outTrie+1, pErrorCode);
97        }
98    }
99
100    return size;
101}
102
103#if !UCONFIG_NO_COLLATION
104
105/* Modified copy of the beginning of ucol_swapBinary(). */
106U_CAPI UBool U_EXPORT2
107ucol_looksLikeCollationBinary(const UDataSwapper *ds,
108                              const void *inData, int32_t length) {
109    const UCATableHeader *inHeader;
110    UCATableHeader header;
111
112    if(ds==NULL || inData==NULL || length<-1) {
113        return FALSE;
114    }
115
116    inHeader=(const UCATableHeader *)inData;
117
118    /*
119     * The collation binary must contain at least the UCATableHeader,
120     * starting with its size field.
121     * sizeof(UCATableHeader)==42*4 in ICU 2.8
122     * check the length against the header size before reading the size field
123     */
124    uprv_memset(&header, 0, sizeof(header));
125    if(length<0) {
126        header.size=udata_readInt32(ds, inHeader->size);
127    } else if((length<(42*4) || length<(header.size=udata_readInt32(ds, inHeader->size)))) {
128        return FALSE;
129    }
130
131    header.magic=ds->readUInt32(inHeader->magic);
132    if(!(
133        header.magic==UCOL_HEADER_MAGIC &&
134        inHeader->formatVersion[0]==3 /*&&
135        inHeader->formatVersion[1]>=0*/
136    )) {
137        return FALSE;
138    }
139
140    if(inHeader->isBigEndian!=ds->inIsBigEndian || inHeader->charSetFamily!=ds->inCharset) {
141        return FALSE;
142    }
143
144    return TRUE;
145}
146
147/* swap a header-less collation binary, inside a resource bundle or ucadata.icu */
148U_CAPI int32_t U_EXPORT2
149ucol_swapBinary(const UDataSwapper *ds,
150                const void *inData, int32_t length, void *outData,
151                UErrorCode *pErrorCode) {
152    const uint8_t *inBytes;
153    uint8_t *outBytes;
154
155    const UCATableHeader *inHeader;
156    UCATableHeader *outHeader;
157    UCATableHeader header;
158
159    uint32_t count;
160
161    /* argument checking in case we were not called from ucol_swap() */
162    if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
163        return 0;
164    }
165    if(ds==NULL || inData==NULL || length<-1 || (length>0 && outData==NULL)) {
166        *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
167        return 0;
168    }
169
170    inBytes=(const uint8_t *)inData;
171    outBytes=(uint8_t *)outData;
172
173    inHeader=(const UCATableHeader *)inData;
174    outHeader=(UCATableHeader *)outData;
175
176    /*
177     * The collation binary must contain at least the UCATableHeader,
178     * starting with its size field.
179     * sizeof(UCATableHeader)==42*4 in ICU 2.8
180     * check the length against the header size before reading the size field
181     */
182    uprv_memset(&header, 0, sizeof(header));
183    if(length<0) {
184        header.size=udata_readInt32(ds, inHeader->size);
185    } else if((length<(42*4) || length<(header.size=udata_readInt32(ds, inHeader->size)))) {
186        udata_printError(ds, "ucol_swapBinary(): too few bytes (%d after header) for collation data\n",
187                         length);
188        *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
189        return 0;
190    }
191
192    header.magic=ds->readUInt32(inHeader->magic);
193    if(!(
194        header.magic==UCOL_HEADER_MAGIC &&
195        inHeader->formatVersion[0]==3 /*&&
196        inHeader->formatVersion[1]>=0*/
197    )) {
198        udata_printError(ds, "ucol_swapBinary(): magic 0x%08x or format version %02x.%02x is not a collation binary\n",
199                         header.magic,
200                         inHeader->formatVersion[0], inHeader->formatVersion[1]);
201        *pErrorCode=U_UNSUPPORTED_ERROR;
202        return 0;
203    }
204
205    if(inHeader->isBigEndian!=ds->inIsBigEndian || inHeader->charSetFamily!=ds->inCharset) {
206        udata_printError(ds, "ucol_swapBinary(): endianness %d or charset %d does not match the swapper\n",
207                         inHeader->isBigEndian, inHeader->charSetFamily);
208        *pErrorCode=U_INVALID_FORMAT_ERROR;
209        return 0;
210    }
211
212    if(length>=0) {
213        /* copy everything, takes care of data that needs no swapping */
214        if(inBytes!=outBytes) {
215            uprv_memcpy(outBytes, inBytes, header.size);
216        }
217
218        /* swap the necessary pieces in the order of their occurrence in the data */
219
220        /* read more of the UCATableHeader (the size field was read above) */
221        header.options=                 ds->readUInt32(inHeader->options);
222        header.UCAConsts=               ds->readUInt32(inHeader->UCAConsts);
223        header.contractionUCACombos=    ds->readUInt32(inHeader->contractionUCACombos);
224        header.mappingPosition=         ds->readUInt32(inHeader->mappingPosition);
225        header.expansion=               ds->readUInt32(inHeader->expansion);
226        header.contractionIndex=        ds->readUInt32(inHeader->contractionIndex);
227        header.contractionCEs=          ds->readUInt32(inHeader->contractionCEs);
228        header.contractionSize=         ds->readUInt32(inHeader->contractionSize);
229        header.endExpansionCE=          ds->readUInt32(inHeader->endExpansionCE);
230        header.expansionCESize=         ds->readUInt32(inHeader->expansionCESize);
231        header.endExpansionCECount=     udata_readInt32(ds, inHeader->endExpansionCECount);
232        header.contractionUCACombosSize=udata_readInt32(ds, inHeader->contractionUCACombosSize);
233        header.scriptToLeadByte=        ds->readUInt32(inHeader->scriptToLeadByte);
234        header.leadByteToScript=        ds->readUInt32(inHeader->leadByteToScript);
235
236        /* swap the 32-bit integers in the header */
237        ds->swapArray32(ds, inHeader, (int32_t)((const char *)&inHeader->jamoSpecial-(const char *)inHeader),
238                           outHeader, pErrorCode);
239        ds->swapArray32(ds, &(inHeader->scriptToLeadByte), sizeof(header.scriptToLeadByte) + sizeof(header.leadByteToScript),
240                           &(outHeader->scriptToLeadByte), pErrorCode);
241        /* set the output platform properties */
242        outHeader->isBigEndian=ds->outIsBigEndian;
243        outHeader->charSetFamily=ds->outCharset;
244
245        /* swap the options */
246        if(header.options!=0) {
247            ds->swapArray32(ds, inBytes+header.options, header.expansion-header.options,
248                               outBytes+header.options, pErrorCode);
249        }
250
251        /* swap the expansions */
252        if(header.mappingPosition!=0 && header.expansion!=0) {
253            if(header.contractionIndex!=0) {
254                /* expansions bounded by contractions */
255                count=header.contractionIndex-header.expansion;
256            } else {
257                /* no contractions: expansions bounded by the main trie */
258                count=header.mappingPosition-header.expansion;
259            }
260            ds->swapArray32(ds, inBytes+header.expansion, (int32_t)count,
261                               outBytes+header.expansion, pErrorCode);
262        }
263
264        /* swap the contractions */
265        if(header.contractionSize!=0) {
266            /* contractionIndex: UChar[] */
267            ds->swapArray16(ds, inBytes+header.contractionIndex, header.contractionSize*2,
268                               outBytes+header.contractionIndex, pErrorCode);
269
270            /* contractionCEs: CEs[] */
271            ds->swapArray32(ds, inBytes+header.contractionCEs, header.contractionSize*4,
272                               outBytes+header.contractionCEs, pErrorCode);
273        }
274
275        /* swap the main trie */
276        if(header.mappingPosition!=0) {
277            count=header.endExpansionCE-header.mappingPosition;
278            utrie_swap(ds, inBytes+header.mappingPosition, (int32_t)count,
279                          outBytes+header.mappingPosition, pErrorCode);
280        }
281
282        /* swap the max expansion table */
283        if(header.endExpansionCECount!=0) {
284            ds->swapArray32(ds, inBytes+header.endExpansionCE, header.endExpansionCECount*4,
285                               outBytes+header.endExpansionCE, pErrorCode);
286        }
287
288        /* expansionCESize, unsafeCP, contrEndCP: uint8_t[], no need to swap */
289
290        /* swap UCA constants */
291        if(header.UCAConsts!=0) {
292            /*
293             * if UCAConsts!=0 then contractionUCACombos because we are swapping
294             * the UCA data file, and we know that the UCA contains contractions
295             */
296            count=header.contractionUCACombos-header.UCAConsts;
297            ds->swapArray32(ds, inBytes+header.UCAConsts, header.contractionUCACombos-header.UCAConsts,
298                               outBytes+header.UCAConsts, pErrorCode);
299        }
300
301        /* swap UCA contractions */
302        if(header.contractionUCACombosSize!=0) {
303            count=header.contractionUCACombosSize*inHeader->contractionUCACombosWidth*U_SIZEOF_UCHAR;
304            ds->swapArray16(ds, inBytes+header.contractionUCACombos, (int32_t)count,
305                               outBytes+header.contractionUCACombos, pErrorCode);
306        }
307
308        /* swap the script to lead bytes */
309        if(header.scriptToLeadByte!=0) {
310            int indexCount = ds->readUInt16(*((uint16_t*)(inBytes+header.scriptToLeadByte))); // each entry = 2 * uint16
311            int dataCount = ds->readUInt16(*((uint16_t*)(inBytes+header.scriptToLeadByte + 2))); // each entry = uint16
312            ds->swapArray16(ds, inBytes+header.scriptToLeadByte,
313                                4 + (4 * indexCount) + (2 * dataCount),
314                                outBytes+header.scriptToLeadByte, pErrorCode);
315        }
316
317        /* swap the lead byte to scripts */
318        if(header.leadByteToScript!=0) {
319            int indexCount = ds->readUInt16(*((uint16_t*)(inBytes+header.leadByteToScript))); // each entry = uint16
320            int dataCount = ds->readUInt16(*((uint16_t*)(inBytes+header.leadByteToScript + 2))); // each entry = uint16
321            ds->swapArray16(ds, inBytes+header.leadByteToScript,
322                                4 + (2 * indexCount) + (2 * dataCount),
323                                outBytes+header.leadByteToScript, pErrorCode);
324        }
325    }
326
327    return header.size;
328}
329
330/* swap ICU collation data like ucadata.icu */
331U_CAPI int32_t U_EXPORT2
332ucol_swap(const UDataSwapper *ds,
333          const void *inData, int32_t length, void *outData,
334          UErrorCode *pErrorCode) {
335
336    const UDataInfo *pInfo;
337    int32_t headerSize, collationSize;
338
339    /* udata_swapDataHeader checks the arguments */
340    headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode);
341    if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
342        return 0;
343    }
344
345    /* check data format and format version */
346    pInfo=(const UDataInfo *)((const char *)inData+4);
347    if(!(
348        pInfo->dataFormat[0]==0x55 &&   /* dataFormat="UCol" */
349        pInfo->dataFormat[1]==0x43 &&
350        pInfo->dataFormat[2]==0x6f &&
351        pInfo->dataFormat[3]==0x6c &&
352        pInfo->formatVersion[0]==3 /*&&
353        pInfo->formatVersion[1]>=0*/
354    )) {
355        udata_printError(ds, "ucol_swap(): data format %02x.%02x.%02x.%02x (format version %02x.%02x) is not a collation file\n",
356                         pInfo->dataFormat[0], pInfo->dataFormat[1],
357                         pInfo->dataFormat[2], pInfo->dataFormat[3],
358                         pInfo->formatVersion[0], pInfo->formatVersion[1]);
359        *pErrorCode=U_UNSUPPORTED_ERROR;
360        return 0;
361    }
362
363    collationSize=ucol_swapBinary(ds,
364                        (const char *)inData+headerSize,
365                        length>=0 ? length-headerSize : -1,
366                        (char *)outData+headerSize,
367                        pErrorCode);
368    if(U_SUCCESS(*pErrorCode)) {
369        return headerSize+collationSize;
370    } else {
371        return 0;
372    }
373}
374
375/* swap inverse UCA collation data (invuca.icu) */
376U_CAPI int32_t U_EXPORT2
377ucol_swapInverseUCA(const UDataSwapper *ds,
378                    const void *inData, int32_t length, void *outData,
379                    UErrorCode *pErrorCode) {
380    const UDataInfo *pInfo;
381    int32_t headerSize;
382
383    const uint8_t *inBytes;
384    uint8_t *outBytes;
385
386    const InverseUCATableHeader *inHeader;
387    InverseUCATableHeader *outHeader;
388    InverseUCATableHeader header={ 0,0,0,0,0,{0,0,0,0},{0,0,0,0,0,0,0,0} };
389
390    /* udata_swapDataHeader checks the arguments */
391    headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode);
392    if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
393        return 0;
394    }
395
396    /* check data format and format version */
397    pInfo=(const UDataInfo *)((const char *)inData+4);
398    if(!(
399        pInfo->dataFormat[0]==0x49 &&   /* dataFormat="InvC" */
400        pInfo->dataFormat[1]==0x6e &&
401        pInfo->dataFormat[2]==0x76 &&
402        pInfo->dataFormat[3]==0x43 &&
403        pInfo->formatVersion[0]==2 &&
404        pInfo->formatVersion[1]>=1
405    )) {
406        udata_printError(ds, "ucol_swapInverseUCA(): data format %02x.%02x.%02x.%02x (format version %02x.%02x) is not an inverse UCA collation file\n",
407                         pInfo->dataFormat[0], pInfo->dataFormat[1],
408                         pInfo->dataFormat[2], pInfo->dataFormat[3],
409                         pInfo->formatVersion[0], pInfo->formatVersion[1]);
410        *pErrorCode=U_UNSUPPORTED_ERROR;
411        return 0;
412    }
413
414    inBytes=(const uint8_t *)inData+headerSize;
415    outBytes=(uint8_t *)outData+headerSize;
416
417    inHeader=(const InverseUCATableHeader *)inBytes;
418    outHeader=(InverseUCATableHeader *)outBytes;
419
420    /*
421     * The inverse UCA collation binary must contain at least the InverseUCATableHeader,
422     * starting with its size field.
423     * sizeof(UCATableHeader)==8*4 in ICU 2.8
424     * check the length against the header size before reading the size field
425     */
426    if(length<0) {
427        header.byteSize=udata_readInt32(ds, inHeader->byteSize);
428    } else if(
429        ((length-headerSize)<(8*4) ||
430         (uint32_t)(length-headerSize)<(header.byteSize=udata_readInt32(ds, inHeader->byteSize)))
431    ) {
432        udata_printError(ds, "ucol_swapInverseUCA(): too few bytes (%d after header) for inverse UCA collation data\n",
433                         length);
434        *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
435        return 0;
436    }
437
438    if(length>=0) {
439        /* copy everything, takes care of data that needs no swapping */
440        if(inBytes!=outBytes) {
441            uprv_memcpy(outBytes, inBytes, header.byteSize);
442        }
443
444        /* swap the necessary pieces in the order of their occurrence in the data */
445
446        /* read more of the InverseUCATableHeader (the byteSize field was read above) */
447        header.tableSize=   ds->readUInt32(inHeader->tableSize);
448        header.contsSize=   ds->readUInt32(inHeader->contsSize);
449        header.table=       ds->readUInt32(inHeader->table);
450        header.conts=       ds->readUInt32(inHeader->conts);
451
452        /* swap the 32-bit integers in the header */
453        ds->swapArray32(ds, inHeader, 5*4, outHeader, pErrorCode);
454
455        /* swap the inverse table; tableSize counts uint32_t[3] rows */
456        ds->swapArray32(ds, inBytes+header.table, header.tableSize*3*4,
457                           outBytes+header.table, pErrorCode);
458
459        /* swap the continuation table; contsSize counts UChars */
460        ds->swapArray16(ds, inBytes+header.conts, header.contsSize*U_SIZEOF_UCHAR,
461                           outBytes+header.conts, pErrorCode);
462    }
463
464    return headerSize+header.byteSize;
465}
466
467#endif /* #if !UCONFIG_NO_COLLATION */
468