1 /*
2 *******************************************************************************
3 * Copyright (C) 1999-2014, International Business Machines Corporation
4 *               and others. All Rights Reserved.
5 *******************************************************************************
6 *   file name:  uresdata.c
7 *   encoding:   US-ASCII
8 *   tab size:   8 (not used)
9 *   indentation:4
10 *
11 *   created on: 1999dec08
12 *   created by: Markus W. Scherer
13 * Modification History:
14 *
15 *   Date        Name        Description
16 *   06/20/2000  helena      OS/400 port changes; mostly typecast.
17 *   06/24/02    weiv        Added support for resource sharing
18 */
19 
20 #include "unicode/utypes.h"
21 #include "unicode/udata.h"
22 #include "unicode/ustring.h"
23 #include "unicode/utf16.h"
24 #include "cmemory.h"
25 #include "cstring.h"
26 #include "uarrsort.h"
27 #include "udataswp.h"
28 #include "ucol_swp.h"
29 #include "uinvchar.h"
30 #include "uresdata.h"
31 #include "uresimp.h"
32 #include "uassert.h"
33 
34 /*
35  * Resource access helpers
36  */
37 
38 /* get a const char* pointer to the key with the keyOffset byte offset from pRoot */
39 #define RES_GET_KEY16(pResData, keyOffset) \
40     ((keyOffset)<(pResData)->localKeyLimit ? \
41         (const char *)(pResData)->pRoot+(keyOffset) : \
42         (pResData)->poolBundleKeys+(keyOffset)-(pResData)->localKeyLimit)
43 
44 #define RES_GET_KEY32(pResData, keyOffset) \
45     ((keyOffset)>=0 ? \
46         (const char *)(pResData)->pRoot+(keyOffset) : \
47         (pResData)->poolBundleKeys+((keyOffset)&0x7fffffff))
48 
49 #define URESDATA_ITEM_NOT_FOUND -1
50 
51 /* empty resources, returned when the resource offset is 0 */
52 static const uint16_t gEmpty16=0;
53 
54 static const struct {
55     int32_t length;
56     int32_t res;
57 } gEmpty32={ 0, 0 };
58 
59 static const struct {
60     int32_t length;
61     UChar nul;
62     UChar pad;
63 } gEmptyString={ 0, 0, 0 };
64 
65 /*
66  * All the type-access functions assume that
67  * the resource is of the expected type.
68  */
69 
70 static int32_t
_res_findTableItem(const ResourceData * pResData,const uint16_t * keyOffsets,int32_t length,const char * key,const char ** realKey)71 _res_findTableItem(const ResourceData *pResData, const uint16_t *keyOffsets, int32_t length,
72                    const char *key, const char **realKey) {
73     const char *tableKey;
74     int32_t mid, start, limit;
75     int result;
76 
77     /* do a binary search for the key */
78     start=0;
79     limit=length;
80     while(start<limit) {
81         mid = (start + limit) / 2;
82         tableKey = RES_GET_KEY16(pResData, keyOffsets[mid]);
83         if (pResData->useNativeStrcmp) {
84             result = uprv_strcmp(key, tableKey);
85         } else {
86             result = uprv_compareInvCharsAsAscii(key, tableKey);
87         }
88         if (result < 0) {
89             limit = mid;
90         } else if (result > 0) {
91             start = mid + 1;
92         } else {
93             /* We found it! */
94             *realKey=tableKey;
95             return mid;
96         }
97     }
98     return URESDATA_ITEM_NOT_FOUND;  /* not found or table is empty. */
99 }
100 
101 static int32_t
_res_findTable32Item(const ResourceData * pResData,const int32_t * keyOffsets,int32_t length,const char * key,const char ** realKey)102 _res_findTable32Item(const ResourceData *pResData, const int32_t *keyOffsets, int32_t length,
103                      const char *key, const char **realKey) {
104     const char *tableKey;
105     int32_t mid, start, limit;
106     int result;
107 
108     /* do a binary search for the key */
109     start=0;
110     limit=length;
111     while(start<limit) {
112         mid = (start + limit) / 2;
113         tableKey = RES_GET_KEY32(pResData, keyOffsets[mid]);
114         if (pResData->useNativeStrcmp) {
115             result = uprv_strcmp(key, tableKey);
116         } else {
117             result = uprv_compareInvCharsAsAscii(key, tableKey);
118         }
119         if (result < 0) {
120             limit = mid;
121         } else if (result > 0) {
122             start = mid + 1;
123         } else {
124             /* We found it! */
125             *realKey=tableKey;
126             return mid;
127         }
128     }
129     return URESDATA_ITEM_NOT_FOUND;  /* not found or table is empty. */
130 }
131 
132 /* helper for res_load() ---------------------------------------------------- */
133 
134 static UBool U_CALLCONV
isAcceptable(void * context,const char * type,const char * name,const UDataInfo * pInfo)135 isAcceptable(void *context,
136              const char *type, const char *name,
137              const UDataInfo *pInfo) {
138     uprv_memcpy(context, pInfo->formatVersion, 4);
139     return (UBool)(
140         pInfo->size>=20 &&
141         pInfo->isBigEndian==U_IS_BIG_ENDIAN &&
142         pInfo->charsetFamily==U_CHARSET_FAMILY &&
143         pInfo->sizeofUChar==U_SIZEOF_UCHAR &&
144         pInfo->dataFormat[0]==0x52 &&   /* dataFormat="ResB" */
145         pInfo->dataFormat[1]==0x65 &&
146         pInfo->dataFormat[2]==0x73 &&
147         pInfo->dataFormat[3]==0x42 &&
148         (pInfo->formatVersion[0]==1 || pInfo->formatVersion[0]==2));
149 }
150 
151 /* semi-public functions ---------------------------------------------------- */
152 
153 static void
res_init(ResourceData * pResData,UVersionInfo formatVersion,const void * inBytes,int32_t length,UErrorCode * errorCode)154 res_init(ResourceData *pResData,
155          UVersionInfo formatVersion, const void *inBytes, int32_t length,
156          UErrorCode *errorCode) {
157     UResType rootType;
158 
159     /* get the root resource */
160     pResData->pRoot=(const int32_t *)inBytes;
161     pResData->rootRes=(Resource)*pResData->pRoot;
162     pResData->p16BitUnits=&gEmpty16;
163 
164     /* formatVersion 1.1 must have a root item and at least 5 indexes */
165     if(length>=0 && (length/4)<((formatVersion[0]==1 && formatVersion[1]==0) ? 1 : 1+5)) {
166         *errorCode=U_INVALID_FORMAT_ERROR;
167         res_unload(pResData);
168         return;
169     }
170 
171     /* currently, we accept only resources that have a Table as their roots */
172     rootType=(UResType)RES_GET_TYPE(pResData->rootRes);
173     if(!URES_IS_TABLE(rootType)) {
174         *errorCode=U_INVALID_FORMAT_ERROR;
175         res_unload(pResData);
176         return;
177     }
178 
179     if(formatVersion[0]==1 && formatVersion[1]==0) {
180         pResData->localKeyLimit=0x10000;  /* greater than any 16-bit key string offset */
181     } else {
182         /* bundles with formatVersion 1.1 and later contain an indexes[] array */
183         const int32_t *indexes=pResData->pRoot+1;
184         int32_t indexLength=indexes[URES_INDEX_LENGTH]&0xff;
185         if(indexLength<=URES_INDEX_MAX_TABLE_LENGTH) {
186             *errorCode=U_INVALID_FORMAT_ERROR;
187             res_unload(pResData);
188             return;
189         }
190         if( length>=0 &&
191             (length<((1+indexLength)<<2) ||
192              length<(indexes[URES_INDEX_BUNDLE_TOP]<<2))
193         ) {
194             *errorCode=U_INVALID_FORMAT_ERROR;
195             res_unload(pResData);
196             return;
197         }
198         if(indexes[URES_INDEX_KEYS_TOP]>(1+indexLength)) {
199             pResData->localKeyLimit=indexes[URES_INDEX_KEYS_TOP]<<2;
200         }
201         if(indexLength>URES_INDEX_ATTRIBUTES) {
202             int32_t att=indexes[URES_INDEX_ATTRIBUTES];
203             pResData->noFallback=(UBool)(att&URES_ATT_NO_FALLBACK);
204             pResData->isPoolBundle=(UBool)((att&URES_ATT_IS_POOL_BUNDLE)!=0);
205             pResData->usesPoolBundle=(UBool)((att&URES_ATT_USES_POOL_BUNDLE)!=0);
206         }
207         if((pResData->isPoolBundle || pResData->usesPoolBundle) && indexLength<=URES_INDEX_POOL_CHECKSUM) {
208             *errorCode=U_INVALID_FORMAT_ERROR;
209             res_unload(pResData);
210             return;
211         }
212         if( indexLength>URES_INDEX_16BIT_TOP &&
213             indexes[URES_INDEX_16BIT_TOP]>indexes[URES_INDEX_KEYS_TOP]
214         ) {
215             pResData->p16BitUnits=(const uint16_t *)(pResData->pRoot+indexes[URES_INDEX_KEYS_TOP]);
216         }
217     }
218 
219     if(formatVersion[0]==1 || U_CHARSET_FAMILY==U_ASCII_FAMILY) {
220         /*
221          * formatVersion 1: compare key strings in native-charset order
222          * formatVersion 2 and up: compare key strings in ASCII order
223          */
224         pResData->useNativeStrcmp=TRUE;
225     }
226 }
227 
228 U_CAPI void U_EXPORT2
res_read(ResourceData * pResData,const UDataInfo * pInfo,const void * inBytes,int32_t length,UErrorCode * errorCode)229 res_read(ResourceData *pResData,
230          const UDataInfo *pInfo, const void *inBytes, int32_t length,
231          UErrorCode *errorCode) {
232     UVersionInfo formatVersion;
233 
234     uprv_memset(pResData, 0, sizeof(ResourceData));
235     if(U_FAILURE(*errorCode)) {
236         return;
237     }
238     if(!isAcceptable(formatVersion, NULL, NULL, pInfo)) {
239         *errorCode=U_INVALID_FORMAT_ERROR;
240         return;
241     }
242     res_init(pResData, formatVersion, inBytes, length, errorCode);
243 }
244 
245 U_CFUNC void
res_load(ResourceData * pResData,const char * path,const char * name,UErrorCode * errorCode)246 res_load(ResourceData *pResData,
247          const char *path, const char *name, UErrorCode *errorCode) {
248     UVersionInfo formatVersion;
249 
250     uprv_memset(pResData, 0, sizeof(ResourceData));
251 
252     /* load the ResourceBundle file */
253     pResData->data=udata_openChoice(path, "res", name, isAcceptable, formatVersion, errorCode);
254     if(U_FAILURE(*errorCode)) {
255         return;
256     }
257 
258     /* get its memory and initialize *pResData */
259     res_init(pResData, formatVersion, udata_getMemory(pResData->data), -1, errorCode);
260 }
261 
262 U_CFUNC void
res_unload(ResourceData * pResData)263 res_unload(ResourceData *pResData) {
264     if(pResData->data!=NULL) {
265         udata_close(pResData->data);
266         pResData->data=NULL;
267     }
268 }
269 
270 static const int8_t gPublicTypes[URES_LIMIT] = {
271     URES_STRING,
272     URES_BINARY,
273     URES_TABLE,
274     URES_ALIAS,
275 
276     URES_TABLE,     /* URES_TABLE32 */
277     URES_TABLE,     /* URES_TABLE16 */
278     URES_STRING,    /* URES_STRING_V2 */
279     URES_INT,
280 
281     URES_ARRAY,
282     URES_ARRAY,     /* URES_ARRAY16 */
283     URES_NONE,
284     URES_NONE,
285 
286     URES_NONE,
287     URES_NONE,
288     URES_INT_VECTOR,
289     URES_NONE
290 };
291 
292 U_CAPI UResType U_EXPORT2
res_getPublicType(Resource res)293 res_getPublicType(Resource res) {
294     return (UResType)gPublicTypes[RES_GET_TYPE(res)];
295 }
296 
297 U_CAPI const UChar * U_EXPORT2
res_getString(const ResourceData * pResData,Resource res,int32_t * pLength)298 res_getString(const ResourceData *pResData, Resource res, int32_t *pLength) {
299     const UChar *p;
300     uint32_t offset=RES_GET_OFFSET(res);
301     int32_t length;
302     if(RES_GET_TYPE(res)==URES_STRING_V2) {
303         int32_t first;
304         p=(const UChar *)(pResData->p16BitUnits+offset);
305         first=*p;
306         if(!U16_IS_TRAIL(first)) {
307             length=u_strlen(p);
308         } else if(first<0xdfef) {
309             length=first&0x3ff;
310             ++p;
311         } else if(first<0xdfff) {
312             length=((first-0xdfef)<<16)|p[1];
313             p+=2;
314         } else {
315             length=((int32_t)p[1]<<16)|p[2];
316             p+=3;
317         }
318     } else if(res==offset) /* RES_GET_TYPE(res)==URES_STRING */ {
319         const int32_t *p32= res==0 ? &gEmptyString.length : pResData->pRoot+res;
320         length=*p32++;
321         p=(const UChar *)p32;
322     } else {
323         p=NULL;
324         length=0;
325     }
326     if(pLength) {
327         *pLength=length;
328     }
329     return p;
330 }
331 
332 U_CAPI const UChar * U_EXPORT2
res_getAlias(const ResourceData * pResData,Resource res,int32_t * pLength)333 res_getAlias(const ResourceData *pResData, Resource res, int32_t *pLength) {
334     const UChar *p;
335     uint32_t offset=RES_GET_OFFSET(res);
336     int32_t length;
337     if(RES_GET_TYPE(res)==URES_ALIAS) {
338         const int32_t *p32= offset==0 ? &gEmptyString.length : pResData->pRoot+offset;
339         length=*p32++;
340         p=(const UChar *)p32;
341     } else {
342         p=NULL;
343         length=0;
344     }
345     if(pLength) {
346         *pLength=length;
347     }
348     return p;
349 }
350 
351 U_CAPI const uint8_t * U_EXPORT2
res_getBinary(const ResourceData * pResData,Resource res,int32_t * pLength)352 res_getBinary(const ResourceData *pResData, Resource res, int32_t *pLength) {
353     const uint8_t *p;
354     uint32_t offset=RES_GET_OFFSET(res);
355     int32_t length;
356     if(RES_GET_TYPE(res)==URES_BINARY) {
357         const int32_t *p32= offset==0 ? (const int32_t*)&gEmpty32 : pResData->pRoot+offset;
358         length=*p32++;
359         p=(const uint8_t *)p32;
360     } else {
361         p=NULL;
362         length=0;
363     }
364     if(pLength) {
365         *pLength=length;
366     }
367     return p;
368 }
369 
370 
371 U_CAPI const int32_t * U_EXPORT2
res_getIntVector(const ResourceData * pResData,Resource res,int32_t * pLength)372 res_getIntVector(const ResourceData *pResData, Resource res, int32_t *pLength) {
373     const int32_t *p;
374     uint32_t offset=RES_GET_OFFSET(res);
375     int32_t length;
376     if(RES_GET_TYPE(res)==URES_INT_VECTOR) {
377         p= offset==0 ? (const int32_t *)&gEmpty32 : pResData->pRoot+offset;
378         length=*p++;
379     } else {
380         p=NULL;
381         length=0;
382     }
383     if(pLength) {
384         *pLength=length;
385     }
386     return p;
387 }
388 
389 U_CAPI int32_t U_EXPORT2
res_countArrayItems(const ResourceData * pResData,Resource res)390 res_countArrayItems(const ResourceData *pResData, Resource res) {
391     uint32_t offset=RES_GET_OFFSET(res);
392     switch(RES_GET_TYPE(res)) {
393     case URES_STRING:
394     case URES_STRING_V2:
395     case URES_BINARY:
396     case URES_ALIAS:
397     case URES_INT:
398     case URES_INT_VECTOR:
399         return 1;
400     case URES_ARRAY:
401     case URES_TABLE32:
402         return offset==0 ? 0 : *(pResData->pRoot+offset);
403     case URES_TABLE:
404         return offset==0 ? 0 : *((const uint16_t *)(pResData->pRoot+offset));
405     case URES_ARRAY16:
406     case URES_TABLE16:
407         return pResData->p16BitUnits[offset];
408     default:
409         return 0;
410     }
411 }
412 
413 U_CAPI Resource U_EXPORT2
res_getTableItemByKey(const ResourceData * pResData,Resource table,int32_t * indexR,const char ** key)414 res_getTableItemByKey(const ResourceData *pResData, Resource table,
415                       int32_t *indexR, const char **key) {
416     uint32_t offset=RES_GET_OFFSET(table);
417     int32_t length;
418     int32_t idx;
419     if(key == NULL || *key == NULL) {
420         return RES_BOGUS;
421     }
422     switch(RES_GET_TYPE(table)) {
423     case URES_TABLE: {
424         if (offset!=0) { /* empty if offset==0 */
425             const uint16_t *p= (const uint16_t *)(pResData->pRoot+offset);
426             length=*p++;
427             *indexR=idx=_res_findTableItem(pResData, p, length, *key, key);
428             if(idx>=0) {
429                 const Resource *p32=(const Resource *)(p+length+(~length&1));
430                 return p32[idx];
431             }
432         }
433         break;
434     }
435     case URES_TABLE16: {
436         const uint16_t *p=pResData->p16BitUnits+offset;
437         length=*p++;
438         *indexR=idx=_res_findTableItem(pResData, p, length, *key, key);
439         if(idx>=0) {
440             return URES_MAKE_RESOURCE(URES_STRING_V2, p[length+idx]);
441         }
442         break;
443     }
444     case URES_TABLE32: {
445         if (offset!=0) { /* empty if offset==0 */
446             const int32_t *p= pResData->pRoot+offset;
447             length=*p++;
448             *indexR=idx=_res_findTable32Item(pResData, p, length, *key, key);
449             if(idx>=0) {
450                 return (Resource)p[length+idx];
451             }
452         }
453         break;
454     }
455     default:
456         break;
457     }
458     return RES_BOGUS;
459 }
460 
461 U_CAPI Resource U_EXPORT2
res_getTableItemByIndex(const ResourceData * pResData,Resource table,int32_t indexR,const char ** key)462 res_getTableItemByIndex(const ResourceData *pResData, Resource table,
463                         int32_t indexR, const char **key) {
464     uint32_t offset=RES_GET_OFFSET(table);
465     int32_t length;
466     U_ASSERT(indexR>=0); /* to ensure the index is not negative */
467     switch(RES_GET_TYPE(table)) {
468     case URES_TABLE: {
469         if (offset != 0) { /* empty if offset==0 */
470             const uint16_t *p= (const uint16_t *)(pResData->pRoot+offset);
471             length=*p++;
472             if(indexR<length) {
473                 const Resource *p32=(const Resource *)(p+length+(~length&1));
474                 if(key!=NULL) {
475                     *key=RES_GET_KEY16(pResData, p[indexR]);
476                 }
477                 return p32[indexR];
478             }
479         }
480         break;
481     }
482     case URES_TABLE16: {
483         const uint16_t *p=pResData->p16BitUnits+offset;
484         length=*p++;
485         if(indexR<length) {
486             if(key!=NULL) {
487                 *key=RES_GET_KEY16(pResData, p[indexR]);
488             }
489             return URES_MAKE_RESOURCE(URES_STRING_V2, p[length+indexR]);
490         }
491         break;
492     }
493     case URES_TABLE32: {
494         if (offset != 0) { /* empty if offset==0 */
495             const int32_t *p= pResData->pRoot+offset;
496             length=*p++;
497             if(indexR<length) {
498                 if(key!=NULL) {
499                     *key=RES_GET_KEY32(pResData, p[indexR]);
500                 }
501                 return (Resource)p[length+indexR];
502             }
503         }
504         break;
505     }
506     default:
507         break;
508     }
509     return RES_BOGUS;
510 }
511 
512 U_CAPI Resource U_EXPORT2
res_getResource(const ResourceData * pResData,const char * key)513 res_getResource(const ResourceData *pResData, const char *key) {
514     const char *realKey=key;
515     int32_t idx;
516     return res_getTableItemByKey(pResData, pResData->rootRes, &idx, &realKey);
517 }
518 
519 U_CAPI Resource U_EXPORT2
res_getArrayItem(const ResourceData * pResData,Resource array,int32_t indexR)520 res_getArrayItem(const ResourceData *pResData, Resource array, int32_t indexR) {
521     uint32_t offset=RES_GET_OFFSET(array);
522     U_ASSERT(indexR>=0); /* to ensure the index is not negative */
523     switch(RES_GET_TYPE(array)) {
524     case URES_ARRAY: {
525         if (offset!=0) { /* empty if offset==0 */
526             const int32_t *p= pResData->pRoot+offset;
527             if(indexR<*p) {
528                 return (Resource)p[1+indexR];
529             }
530         }
531         break;
532     }
533     case URES_ARRAY16: {
534         const uint16_t *p=pResData->p16BitUnits+offset;
535         if(indexR<*p) {
536             return URES_MAKE_RESOURCE(URES_STRING_V2, p[1+indexR]);
537         }
538         break;
539     }
540     default:
541         break;
542     }
543     return RES_BOGUS;
544 }
545 
546 U_CFUNC Resource
res_findResource(const ResourceData * pResData,Resource r,char ** path,const char ** key)547 res_findResource(const ResourceData *pResData, Resource r, char** path, const char** key) {
548   /* we pass in a path. CollationElements/Sequence or zoneStrings/3/2 etc.
549    * iterates over a path and stops when a scalar resource is found. This
550    * CAN be an alias. Path gets set to the part that has not yet been processed.
551    */
552 
553   char *pathP = *path, *nextSepP = *path;
554   char *closeIndex = NULL;
555   Resource t1 = r;
556   Resource t2;
557   int32_t indexR = 0;
558   UResType type = (UResType)RES_GET_TYPE(t1);
559 
560   /* if you come in with an empty path, you'll be getting back the same resource */
561   if(!uprv_strlen(pathP)) {
562       return r;
563   }
564 
565   /* one needs to have an aggregate resource in order to search in it */
566   if(!URES_IS_CONTAINER(type)) {
567       return RES_BOGUS;
568   }
569 
570   while(nextSepP && *pathP && t1 != RES_BOGUS && URES_IS_CONTAINER(type)) {
571     /* Iteration stops if: the path has been consumed, we found a non-existing
572      * resource (t1 == RES_BOGUS) or we found a scalar resource (including alias)
573      */
574     nextSepP = uprv_strchr(pathP, RES_PATH_SEPARATOR);
575     /* if there are more separators, terminate string
576      * and set path to the remaining part of the string
577      */
578     if(nextSepP != NULL) {
579       *nextSepP = 0; /* overwrite the separator with a NUL to terminate the key */
580       *path = nextSepP+1;
581     } else {
582       *path = uprv_strchr(pathP, 0);
583     }
584 
585     /* if the resource is a table */
586     /* try the key based access */
587     if(URES_IS_TABLE(type)) {
588       *key = pathP;
589       t2 = res_getTableItemByKey(pResData, t1, &indexR, key);
590       if(t2 == RES_BOGUS) {
591         /* if we fail to get the resource by key, maybe we got an index */
592         indexR = uprv_strtol(pathP, &closeIndex, 10);
593         if(closeIndex != pathP) {
594           /* if we indeed have an index, try to get the item by index */
595           t2 = res_getTableItemByIndex(pResData, t1, indexR, key);
596         }
597       }
598     } else if(URES_IS_ARRAY(type)) {
599       indexR = uprv_strtol(pathP, &closeIndex, 10);
600       if(closeIndex != pathP) {
601         t2 = res_getArrayItem(pResData, t1, indexR);
602       } else {
603         t2 = RES_BOGUS; /* have an array, but don't have a valid index */
604       }
605       *key = NULL;
606     } else { /* can't do much here, except setting t2 to bogus */
607       t2 = RES_BOGUS;
608     }
609     t1 = t2;
610     type = (UResType)RES_GET_TYPE(t1);
611     /* position pathP to next resource key/index */
612     pathP = *path;
613   }
614 
615   return t1;
616 }
617 
618 /* resource bundle swapping ------------------------------------------------- */
619 
620 /*
621  * Need to always enumerate the entire item tree,
622  * track the lowest address of any item to use as the limit for char keys[],
623  * track the highest address of any item to return the size of the data.
624  *
625  * We should have thought of storing those in the data...
626  * It is possible to extend the data structure by putting additional values
627  * in places that are inaccessible by ordinary enumeration of the item tree.
628  * For example, additional integers could be stored at the beginning or
629  * end of the key strings; this could be indicated by a minor version number,
630  * and the data swapping would have to know about these values.
631  *
632  * The data structure does not forbid keys to be shared, so we must swap
633  * all keys once instead of each key when it is referenced.
634  *
635  * These swapping functions assume that a resource bundle always has a length
636  * that is a multiple of 4 bytes.
637  * Currently, this is trivially true because genrb writes bundle tree leaves
638  * physically first, before their branches, so that the root table with its
639  * array of resource items (uint32_t values) is always last.
640  */
641 
642 /* definitions for table sorting ------------------------ */
643 
644 /*
645  * row of a temporary array
646  *
647  * gets platform-endian key string indexes and sorting indexes;
648  * after sorting this array by keys, the actual key/value arrays are permutated
649  * according to the sorting indexes
650  */
651 typedef struct Row {
652     int32_t keyIndex, sortIndex;
653 } Row;
654 
655 static int32_t
ures_compareRows(const void * context,const void * left,const void * right)656 ures_compareRows(const void *context, const void *left, const void *right) {
657     const char *keyChars=(const char *)context;
658     return (int32_t)uprv_strcmp(keyChars+((const Row *)left)->keyIndex,
659                                 keyChars+((const Row *)right)->keyIndex);
660 }
661 
662 typedef struct TempTable {
663     const char *keyChars;
664     Row *rows;
665     int32_t *resort;
666     uint32_t *resFlags;
667     int32_t localKeyLimit;
668     uint8_t majorFormatVersion;
669 } TempTable;
670 
671 enum {
672     STACK_ROW_CAPACITY=200
673 };
674 
675 /* The table item key string is not locally available. */
676 static const char *const gUnknownKey="";
677 
678 /* resource table key for collation binaries: "%%CollationBin" */
679 static const UChar gCollationBinKey[]={
680     0x25, 0x25,
681     0x43, 0x6f, 0x6c, 0x6c, 0x61, 0x74, 0x69, 0x6f, 0x6e,
682     0x42, 0x69, 0x6e,
683     0
684 };
685 
686 /*
687  * swap one resource item
688  */
689 static void
ures_swapResource(const UDataSwapper * ds,const Resource * inBundle,Resource * outBundle,Resource res,const char * key,TempTable * pTempTable,UErrorCode * pErrorCode)690 ures_swapResource(const UDataSwapper *ds,
691                   const Resource *inBundle, Resource *outBundle,
692                   Resource res, /* caller swaps res itself */
693                   const char *key,
694                   TempTable *pTempTable,
695                   UErrorCode *pErrorCode) {
696     const Resource *p;
697     Resource *q;
698     int32_t offset, count;
699 
700     switch(RES_GET_TYPE(res)) {
701     case URES_TABLE16:
702     case URES_STRING_V2:
703     case URES_INT:
704     case URES_ARRAY16:
705         /* integer, or points to 16-bit units, nothing to do here */
706         return;
707     default:
708         break;
709     }
710 
711     /* all other types use an offset to point to their data */
712     offset=(int32_t)RES_GET_OFFSET(res);
713     if(offset==0) {
714         /* special offset indicating an empty item */
715         return;
716     }
717     if(pTempTable->resFlags[offset>>5]&((uint32_t)1<<(offset&0x1f))) {
718         /* we already swapped this resource item */
719         return;
720     } else {
721         /* mark it as swapped now */
722         pTempTable->resFlags[offset>>5]|=((uint32_t)1<<(offset&0x1f));
723     }
724 
725     p=inBundle+offset;
726     q=outBundle+offset;
727 
728     switch(RES_GET_TYPE(res)) {
729     case URES_ALIAS:
730         /* physically same value layout as string, fall through */
731     case URES_STRING:
732         count=udata_readInt32(ds, (int32_t)*p);
733         /* swap length */
734         ds->swapArray32(ds, p, 4, q, pErrorCode);
735         /* swap each UChar (the terminating NUL would not change) */
736         ds->swapArray16(ds, p+1, 2*count, q+1, pErrorCode);
737         break;
738     case URES_BINARY:
739         count=udata_readInt32(ds, (int32_t)*p);
740         /* swap length */
741         ds->swapArray32(ds, p, 4, q, pErrorCode);
742         /* no need to swap or copy bytes - ures_swap() copied them all */
743 
744         /* swap known formats */
745 #if !UCONFIG_NO_COLLATION
746         if( key!=NULL &&  /* the binary is in a table */
747             (key!=gUnknownKey ?
748                 /* its table key string is "%%CollationBin" */
749                 0==ds->compareInvChars(ds, key, -1,
750                                        gCollationBinKey, UPRV_LENGTHOF(gCollationBinKey)-1) :
751                 /* its table key string is unknown but it looks like a collation binary */
752                 ucol_looksLikeCollationBinary(ds, p+1, count))
753         ) {
754             ucol_swap(ds, p+1, count, q+1, pErrorCode);
755         }
756 #endif
757         break;
758     case URES_TABLE:
759     case URES_TABLE32:
760         {
761             const uint16_t *pKey16;
762             uint16_t *qKey16;
763 
764             const int32_t *pKey32;
765             int32_t *qKey32;
766 
767             Resource item;
768             int32_t i, oldIndex;
769 
770             if(RES_GET_TYPE(res)==URES_TABLE) {
771                 /* get table item count */
772                 pKey16=(const uint16_t *)p;
773                 qKey16=(uint16_t *)q;
774                 count=ds->readUInt16(*pKey16);
775 
776                 pKey32=qKey32=NULL;
777 
778                 /* swap count */
779                 ds->swapArray16(ds, pKey16++, 2, qKey16++, pErrorCode);
780 
781                 offset+=((1+count)+1)/2;
782             } else {
783                 /* get table item count */
784                 pKey32=(const int32_t *)p;
785                 qKey32=(int32_t *)q;
786                 count=udata_readInt32(ds, *pKey32);
787 
788                 pKey16=qKey16=NULL;
789 
790                 /* swap count */
791                 ds->swapArray32(ds, pKey32++, 4, qKey32++, pErrorCode);
792 
793                 offset+=1+count;
794             }
795 
796             if(count==0) {
797                 break;
798             }
799 
800             p=inBundle+offset; /* pointer to table resources */
801             q=outBundle+offset;
802 
803             /* recurse */
804             for(i=0; i<count; ++i) {
805                 const char *itemKey=gUnknownKey;
806                 if(pKey16!=NULL) {
807                     int32_t keyOffset=ds->readUInt16(pKey16[i]);
808                     if(keyOffset<pTempTable->localKeyLimit) {
809                         itemKey=(const char *)outBundle+keyOffset;
810                     }
811                 } else {
812                     int32_t keyOffset=udata_readInt32(ds, pKey32[i]);
813                     if(keyOffset>=0) {
814                         itemKey=(const char *)outBundle+keyOffset;
815                     }
816                 }
817                 item=ds->readUInt32(p[i]);
818                 ures_swapResource(ds, inBundle, outBundle, item, itemKey, pTempTable, pErrorCode);
819                 if(U_FAILURE(*pErrorCode)) {
820                     udata_printError(ds, "ures_swapResource(table res=%08x)[%d].recurse(%08x) failed\n",
821                                      res, i, item);
822                     return;
823                 }
824             }
825 
826             if(pTempTable->majorFormatVersion>1 || ds->inCharset==ds->outCharset) {
827                 /* no need to sort, just swap the offset/value arrays */
828                 if(pKey16!=NULL) {
829                     ds->swapArray16(ds, pKey16, count*2, qKey16, pErrorCode);
830                     ds->swapArray32(ds, p, count*4, q, pErrorCode);
831                 } else {
832                     /* swap key offsets and items as one array */
833                     ds->swapArray32(ds, pKey32, count*2*4, qKey32, pErrorCode);
834                 }
835                 break;
836             }
837 
838             /*
839              * We need to sort tables by outCharset key strings because they
840              * sort differently for different charset families.
841              * ures_swap() already set pTempTable->keyChars appropriately.
842              * First we set up a temporary table with the key indexes and
843              * sorting indexes and sort that.
844              * Then we permutate and copy/swap the actual values.
845              */
846             if(pKey16!=NULL) {
847                 for(i=0; i<count; ++i) {
848                     pTempTable->rows[i].keyIndex=ds->readUInt16(pKey16[i]);
849                     pTempTable->rows[i].sortIndex=i;
850                 }
851             } else {
852                 for(i=0; i<count; ++i) {
853                     pTempTable->rows[i].keyIndex=udata_readInt32(ds, pKey32[i]);
854                     pTempTable->rows[i].sortIndex=i;
855                 }
856             }
857             uprv_sortArray(pTempTable->rows, count, sizeof(Row),
858                            ures_compareRows, pTempTable->keyChars,
859                            FALSE, pErrorCode);
860             if(U_FAILURE(*pErrorCode)) {
861                 udata_printError(ds, "ures_swapResource(table res=%08x).uprv_sortArray(%d items) failed\n",
862                                  res, count);
863                 return;
864             }
865 
866             /*
867              * copy/swap/permutate items
868              *
869              * If we swap in-place, then the permutation must use another
870              * temporary array (pTempTable->resort)
871              * before the results are copied to the outBundle.
872              */
873             /* keys */
874             if(pKey16!=NULL) {
875                 uint16_t *rKey16;
876 
877                 if(pKey16!=qKey16) {
878                     rKey16=qKey16;
879                 } else {
880                     rKey16=(uint16_t *)pTempTable->resort;
881                 }
882                 for(i=0; i<count; ++i) {
883                     oldIndex=pTempTable->rows[i].sortIndex;
884                     ds->swapArray16(ds, pKey16+oldIndex, 2, rKey16+i, pErrorCode);
885                 }
886                 if(qKey16!=rKey16) {
887                     uprv_memcpy(qKey16, rKey16, 2*count);
888                 }
889             } else {
890                 int32_t *rKey32;
891 
892                 if(pKey32!=qKey32) {
893                     rKey32=qKey32;
894                 } else {
895                     rKey32=pTempTable->resort;
896                 }
897                 for(i=0; i<count; ++i) {
898                     oldIndex=pTempTable->rows[i].sortIndex;
899                     ds->swapArray32(ds, pKey32+oldIndex, 4, rKey32+i, pErrorCode);
900                 }
901                 if(qKey32!=rKey32) {
902                     uprv_memcpy(qKey32, rKey32, 4*count);
903                 }
904             }
905 
906             /* resources */
907             {
908                 Resource *r;
909 
910 
911                 if(p!=q) {
912                     r=q;
913                 } else {
914                     r=(Resource *)pTempTable->resort;
915                 }
916                 for(i=0; i<count; ++i) {
917                     oldIndex=pTempTable->rows[i].sortIndex;
918                     ds->swapArray32(ds, p+oldIndex, 4, r+i, pErrorCode);
919                 }
920                 if(q!=r) {
921                     uprv_memcpy(q, r, 4*count);
922                 }
923             }
924         }
925         break;
926     case URES_ARRAY:
927         {
928             Resource item;
929             int32_t i;
930 
931             count=udata_readInt32(ds, (int32_t)*p);
932             /* swap length */
933             ds->swapArray32(ds, p++, 4, q++, pErrorCode);
934 
935             /* recurse */
936             for(i=0; i<count; ++i) {
937                 item=ds->readUInt32(p[i]);
938                 ures_swapResource(ds, inBundle, outBundle, item, NULL, pTempTable, pErrorCode);
939                 if(U_FAILURE(*pErrorCode)) {
940                     udata_printError(ds, "ures_swapResource(array res=%08x)[%d].recurse(%08x) failed\n",
941                                      res, i, item);
942                     return;
943                 }
944             }
945 
946             /* swap items */
947             ds->swapArray32(ds, p, 4*count, q, pErrorCode);
948         }
949         break;
950     case URES_INT_VECTOR:
951         count=udata_readInt32(ds, (int32_t)*p);
952         /* swap length and each integer */
953         ds->swapArray32(ds, p, 4*(1+count), q, pErrorCode);
954         break;
955     default:
956         /* also catches RES_BOGUS */
957         *pErrorCode=U_UNSUPPORTED_ERROR;
958         break;
959     }
960 }
961 
962 U_CAPI int32_t U_EXPORT2
ures_swap(const UDataSwapper * ds,const void * inData,int32_t length,void * outData,UErrorCode * pErrorCode)963 ures_swap(const UDataSwapper *ds,
964           const void *inData, int32_t length, void *outData,
965           UErrorCode *pErrorCode) {
966     const UDataInfo *pInfo;
967     const Resource *inBundle;
968     Resource rootRes;
969     int32_t headerSize, maxTableLength;
970 
971     Row rows[STACK_ROW_CAPACITY];
972     int32_t resort[STACK_ROW_CAPACITY];
973     TempTable tempTable;
974 
975     const int32_t *inIndexes;
976 
977     /* the following integers count Resource item offsets (4 bytes each), not bytes */
978     int32_t bundleLength, indexLength, keysBottom, keysTop, resBottom, top;
979 
980     /* udata_swapDataHeader checks the arguments */
981     headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode);
982     if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
983         return 0;
984     }
985 
986     /* check data format and format version */
987     pInfo=(const UDataInfo *)((const char *)inData+4);
988     if(!(
989         pInfo->dataFormat[0]==0x52 &&   /* dataFormat="ResB" */
990         pInfo->dataFormat[1]==0x65 &&
991         pInfo->dataFormat[2]==0x73 &&
992         pInfo->dataFormat[3]==0x42 &&
993         ((pInfo->formatVersion[0]==1 && pInfo->formatVersion[1]>=1) ||  /* formatVersion 1.1+ or 2.x */
994          pInfo->formatVersion[0]==2)
995     )) {
996         udata_printError(ds, "ures_swap(): data format %02x.%02x.%02x.%02x (format version %02x.%02x) is not a resource bundle\n",
997                          pInfo->dataFormat[0], pInfo->dataFormat[1],
998                          pInfo->dataFormat[2], pInfo->dataFormat[3],
999                          pInfo->formatVersion[0], pInfo->formatVersion[1]);
1000         *pErrorCode=U_UNSUPPORTED_ERROR;
1001         return 0;
1002     }
1003     tempTable.majorFormatVersion=pInfo->formatVersion[0];
1004 
1005     /* a resource bundle must contain at least one resource item */
1006     if(length<0) {
1007         bundleLength=-1;
1008     } else {
1009         bundleLength=(length-headerSize)/4;
1010 
1011         /* formatVersion 1.1 must have a root item and at least 5 indexes */
1012         if(bundleLength<(1+5)) {
1013             udata_printError(ds, "ures_swap(): too few bytes (%d after header) for a resource bundle\n",
1014                              length-headerSize);
1015             *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1016             return 0;
1017         }
1018     }
1019 
1020     inBundle=(const Resource *)((const char *)inData+headerSize);
1021     rootRes=ds->readUInt32(*inBundle);
1022 
1023     /* formatVersion 1.1 adds the indexes[] array */
1024     inIndexes=(const int32_t *)(inBundle+1);
1025 
1026     indexLength=udata_readInt32(ds, inIndexes[URES_INDEX_LENGTH])&0xff;
1027     if(indexLength<=URES_INDEX_MAX_TABLE_LENGTH) {
1028         udata_printError(ds, "ures_swap(): too few indexes for a 1.1+ resource bundle\n");
1029         *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1030         return 0;
1031     }
1032     keysBottom=1+indexLength;
1033     keysTop=udata_readInt32(ds, inIndexes[URES_INDEX_KEYS_TOP]);
1034     if(indexLength>URES_INDEX_16BIT_TOP) {
1035         resBottom=udata_readInt32(ds, inIndexes[URES_INDEX_16BIT_TOP]);
1036     } else {
1037         resBottom=keysTop;
1038     }
1039     top=udata_readInt32(ds, inIndexes[URES_INDEX_BUNDLE_TOP]);
1040     maxTableLength=udata_readInt32(ds, inIndexes[URES_INDEX_MAX_TABLE_LENGTH]);
1041 
1042     if(0<=bundleLength && bundleLength<top) {
1043         udata_printError(ds, "ures_swap(): resource top %d exceeds bundle length %d\n",
1044                          top, bundleLength);
1045         *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1046         return 0;
1047     }
1048     if(keysTop>(1+indexLength)) {
1049         tempTable.localKeyLimit=keysTop<<2;
1050     } else {
1051         tempTable.localKeyLimit=0;
1052     }
1053 
1054     if(length>=0) {
1055         Resource *outBundle=(Resource *)((char *)outData+headerSize);
1056 
1057         /* track which resources we have already swapped */
1058         uint32_t stackResFlags[STACK_ROW_CAPACITY];
1059         int32_t resFlagsLength;
1060 
1061         /*
1062          * We need one bit per 4 resource bundle bytes so that we can track
1063          * every possible Resource for whether we have swapped it already.
1064          * Multiple Resource words can refer to the same bundle offsets
1065          * for sharing identical values.
1066          * We could optimize this by allocating only for locations above
1067          * where Resource values are stored (above keys & strings).
1068          */
1069         resFlagsLength=(length+31)>>5;          /* number of bytes needed */
1070         resFlagsLength=(resFlagsLength+3)&~3;   /* multiple of 4 bytes for uint32_t */
1071         if(resFlagsLength<=sizeof(stackResFlags)) {
1072             tempTable.resFlags=stackResFlags;
1073         } else {
1074             tempTable.resFlags=(uint32_t *)uprv_malloc(resFlagsLength);
1075             if(tempTable.resFlags==NULL) {
1076                 udata_printError(ds, "ures_swap(): unable to allocate memory for tracking resources\n");
1077                 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
1078                 return 0;
1079             }
1080         }
1081         uprv_memset(tempTable.resFlags, 0, resFlagsLength);
1082 
1083         /* copy the bundle for binary and inaccessible data */
1084         if(inData!=outData) {
1085             uprv_memcpy(outBundle, inBundle, 4*top);
1086         }
1087 
1088         /* swap the key strings, but not the padding bytes (0xaa) after the last string and its NUL */
1089         udata_swapInvStringBlock(ds, inBundle+keysBottom, 4*(keysTop-keysBottom),
1090                                     outBundle+keysBottom, pErrorCode);
1091         if(U_FAILURE(*pErrorCode)) {
1092             udata_printError(ds, "ures_swap().udata_swapInvStringBlock(keys[%d]) failed\n", 4*(keysTop-keysBottom));
1093             return 0;
1094         }
1095 
1096         /* swap the 16-bit units (strings, table16, array16) */
1097         if(keysTop<resBottom) {
1098             ds->swapArray16(ds, inBundle+keysTop, (resBottom-keysTop)*4, outBundle+keysTop, pErrorCode);
1099             if(U_FAILURE(*pErrorCode)) {
1100                 udata_printError(ds, "ures_swap().swapArray16(16-bit units[%d]) failed\n", 2*(resBottom-keysTop));
1101                 return 0;
1102             }
1103         }
1104 
1105         /* allocate the temporary table for sorting resource tables */
1106         tempTable.keyChars=(const char *)outBundle; /* sort by outCharset */
1107         if(tempTable.majorFormatVersion>1 || maxTableLength<=STACK_ROW_CAPACITY) {
1108             tempTable.rows=rows;
1109             tempTable.resort=resort;
1110         } else {
1111             tempTable.rows=(Row *)uprv_malloc(maxTableLength*sizeof(Row)+maxTableLength*4);
1112             if(tempTable.rows==NULL) {
1113                 udata_printError(ds, "ures_swap(): unable to allocate memory for sorting tables (max length: %d)\n",
1114                                  maxTableLength);
1115                 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
1116                 if(tempTable.resFlags!=stackResFlags) {
1117                     uprv_free(tempTable.resFlags);
1118                 }
1119                 return 0;
1120             }
1121             tempTable.resort=(int32_t *)(tempTable.rows+maxTableLength);
1122         }
1123 
1124         /* swap the resources */
1125         ures_swapResource(ds, inBundle, outBundle, rootRes, NULL, &tempTable, pErrorCode);
1126         if(U_FAILURE(*pErrorCode)) {
1127             udata_printError(ds, "ures_swapResource(root res=%08x) failed\n",
1128                              rootRes);
1129         }
1130 
1131         if(tempTable.rows!=rows) {
1132             uprv_free(tempTable.rows);
1133         }
1134         if(tempTable.resFlags!=stackResFlags) {
1135             uprv_free(tempTable.resFlags);
1136         }
1137 
1138         /* swap the root resource and indexes */
1139         ds->swapArray32(ds, inBundle, keysBottom*4, outBundle, pErrorCode);
1140     }
1141 
1142     return headerSize+4*top;
1143 }
1144