1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/intl/icu/source/common/ucmndata.c Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,382 @@
1.4 +/*
1.5 +******************************************************************************
1.6 +*
1.7 +* Copyright (C) 1999-2011, International Business Machines
1.8 +* Corporation and others. All Rights Reserved.
1.9 +*
1.10 +******************************************************************************/
1.11 +
1.12 +
1.13 +/*------------------------------------------------------------------------------
1.14 + *
1.15 + * UCommonData An abstract interface for dealing with ICU Common Data Files.
1.16 + * ICU Common Data Files are a grouping of a number of individual
1.17 + * data items (resources, converters, tables, anything) into a
1.18 + * single file or dll. The combined format includes a table of
1.19 + * contents for locating the individual items by name.
1.20 + *
1.21 + * Two formats for the table of contents are supported, which is
1.22 + * why there is an abstract inteface involved.
1.23 + *
1.24 + */
1.25 +
1.26 +#include "unicode/utypes.h"
1.27 +#include "unicode/udata.h"
1.28 +#include "cstring.h"
1.29 +#include "ucmndata.h"
1.30 +#include "udatamem.h"
1.31 +
1.32 +#if defined(UDATA_DEBUG) || defined(UDATA_DEBUG_DUMP)
1.33 +# include <stdio.h>
1.34 +#endif
1.35 +
1.36 +U_CFUNC uint16_t
1.37 +udata_getHeaderSize(const DataHeader *udh) {
1.38 + if(udh==NULL) {
1.39 + return 0;
1.40 + } else if(udh->info.isBigEndian==U_IS_BIG_ENDIAN) {
1.41 + /* same endianness */
1.42 + return udh->dataHeader.headerSize;
1.43 + } else {
1.44 + /* opposite endianness */
1.45 + uint16_t x=udh->dataHeader.headerSize;
1.46 + return (uint16_t)((x<<8)|(x>>8));
1.47 + }
1.48 +}
1.49 +
1.50 +U_CFUNC uint16_t
1.51 +udata_getInfoSize(const UDataInfo *info) {
1.52 + if(info==NULL) {
1.53 + return 0;
1.54 + } else if(info->isBigEndian==U_IS_BIG_ENDIAN) {
1.55 + /* same endianness */
1.56 + return info->size;
1.57 + } else {
1.58 + /* opposite endianness */
1.59 + uint16_t x=info->size;
1.60 + return (uint16_t)((x<<8)|(x>>8));
1.61 + }
1.62 +}
1.63 +
1.64 +/*-----------------------------------------------------------------------------*
1.65 + * *
1.66 + * Pointer TOCs. TODO: This form of table-of-contents should be removed *
1.67 + * because DLLs must be relocated on loading to correct the *
1.68 + * pointer values and this operation makes shared memory *
1.69 + * mapping of the data much less likely to work. *
1.70 + * *
1.71 + *-----------------------------------------------------------------------------*/
1.72 +typedef struct {
1.73 + const char *entryName;
1.74 + const DataHeader *pHeader;
1.75 +} PointerTOCEntry;
1.76 +
1.77 +
1.78 +typedef struct {
1.79 + uint32_t count;
1.80 + uint32_t reserved;
1.81 + PointerTOCEntry entry[2]; /* Actual size is from count. */
1.82 +} PointerTOC;
1.83 +
1.84 +
1.85 +/* definition of OffsetTOC struct types moved to ucmndata.h */
1.86 +
1.87 +/*-----------------------------------------------------------------------------*
1.88 + * *
1.89 + * entry point lookup implementations *
1.90 + * *
1.91 + *-----------------------------------------------------------------------------*/
1.92 +
1.93 +#ifndef MIN
1.94 +#define MIN(a,b) (((a)<(b)) ? (a) : (b))
1.95 +#endif
1.96 +
1.97 +/**
1.98 + * Compare strings where we know the shared prefix length,
1.99 + * and advance the prefix length as we find that the strings share even more characters.
1.100 + */
1.101 +static int32_t
1.102 +strcmpAfterPrefix(const char *s1, const char *s2, int32_t *pPrefixLength) {
1.103 + int32_t pl=*pPrefixLength;
1.104 + int32_t cmp=0;
1.105 + s1+=pl;
1.106 + s2+=pl;
1.107 + for(;;) {
1.108 + int32_t c1=(uint8_t)*s1++;
1.109 + int32_t c2=(uint8_t)*s2++;
1.110 + cmp=c1-c2;
1.111 + if(cmp!=0 || c1==0) { /* different or done */
1.112 + break;
1.113 + }
1.114 + ++pl; /* increment shared same-prefix length */
1.115 + }
1.116 + *pPrefixLength=pl;
1.117 + return cmp;
1.118 +}
1.119 +
1.120 +static int32_t
1.121 +offsetTOCPrefixBinarySearch(const char *s, const char *names,
1.122 + const UDataOffsetTOCEntry *toc, int32_t count) {
1.123 + int32_t start=0;
1.124 + int32_t limit=count;
1.125 + /*
1.126 + * Remember the shared prefix between s, start and limit,
1.127 + * and don't compare that shared prefix again.
1.128 + * The shared prefix should get longer as we narrow the [start, limit[ range.
1.129 + */
1.130 + int32_t startPrefixLength=0;
1.131 + int32_t limitPrefixLength=0;
1.132 + if(count==0) {
1.133 + return -1;
1.134 + }
1.135 + /*
1.136 + * Prime the prefix lengths so that we don't keep prefixLength at 0 until
1.137 + * both the start and limit indexes have moved.
1.138 + * At the same time, we find if s is one of the start and (limit-1) names,
1.139 + * and if not, exclude them from the actual binary search.
1.140 + */
1.141 + if(0==strcmpAfterPrefix(s, names+toc[0].nameOffset, &startPrefixLength)) {
1.142 + return 0;
1.143 + }
1.144 + ++start;
1.145 + --limit;
1.146 + if(0==strcmpAfterPrefix(s, names+toc[limit].nameOffset, &limitPrefixLength)) {
1.147 + return limit;
1.148 + }
1.149 + while(start<limit) {
1.150 + int32_t i=(start+limit)/2;
1.151 + int32_t prefixLength=MIN(startPrefixLength, limitPrefixLength);
1.152 + int32_t cmp=strcmpAfterPrefix(s, names+toc[i].nameOffset, &prefixLength);
1.153 + if(cmp<0) {
1.154 + limit=i;
1.155 + limitPrefixLength=prefixLength;
1.156 + } else if(cmp==0) {
1.157 + return i;
1.158 + } else {
1.159 + start=i+1;
1.160 + startPrefixLength=prefixLength;
1.161 + }
1.162 + }
1.163 + return -1;
1.164 +}
1.165 +
1.166 +static int32_t
1.167 +pointerTOCPrefixBinarySearch(const char *s, const PointerTOCEntry *toc, int32_t count) {
1.168 + int32_t start=0;
1.169 + int32_t limit=count;
1.170 + /*
1.171 + * Remember the shared prefix between s, start and limit,
1.172 + * and don't compare that shared prefix again.
1.173 + * The shared prefix should get longer as we narrow the [start, limit[ range.
1.174 + */
1.175 + int32_t startPrefixLength=0;
1.176 + int32_t limitPrefixLength=0;
1.177 + if(count==0) {
1.178 + return -1;
1.179 + }
1.180 + /*
1.181 + * Prime the prefix lengths so that we don't keep prefixLength at 0 until
1.182 + * both the start and limit indexes have moved.
1.183 + * At the same time, we find if s is one of the start and (limit-1) names,
1.184 + * and if not, exclude them from the actual binary search.
1.185 + */
1.186 + if(0==strcmpAfterPrefix(s, toc[0].entryName, &startPrefixLength)) {
1.187 + return 0;
1.188 + }
1.189 + ++start;
1.190 + --limit;
1.191 + if(0==strcmpAfterPrefix(s, toc[limit].entryName, &limitPrefixLength)) {
1.192 + return limit;
1.193 + }
1.194 + while(start<limit) {
1.195 + int32_t i=(start+limit)/2;
1.196 + int32_t prefixLength=MIN(startPrefixLength, limitPrefixLength);
1.197 + int32_t cmp=strcmpAfterPrefix(s, toc[i].entryName, &prefixLength);
1.198 + if(cmp<0) {
1.199 + limit=i;
1.200 + limitPrefixLength=prefixLength;
1.201 + } else if(cmp==0) {
1.202 + return i;
1.203 + } else {
1.204 + start=i+1;
1.205 + startPrefixLength=prefixLength;
1.206 + }
1.207 + }
1.208 + return -1;
1.209 +}
1.210 +
1.211 +static uint32_t offsetTOCEntryCount(const UDataMemory *pData) {
1.212 + int32_t retVal=0;
1.213 + const UDataOffsetTOC *toc = (UDataOffsetTOC *)pData->toc;
1.214 + if (toc != NULL) {
1.215 + retVal = toc->count;
1.216 + }
1.217 + return retVal;
1.218 +}
1.219 +
1.220 +static const DataHeader *
1.221 +offsetTOCLookupFn(const UDataMemory *pData,
1.222 + const char *tocEntryName,
1.223 + int32_t *pLength,
1.224 + UErrorCode *pErrorCode) {
1.225 + const UDataOffsetTOC *toc = (UDataOffsetTOC *)pData->toc;
1.226 + if(toc!=NULL) {
1.227 + const char *base=(const char *)toc;
1.228 + int32_t number, count=(int32_t)toc->count;
1.229 +
1.230 + /* perform a binary search for the data in the common data's table of contents */
1.231 +#if defined (UDATA_DEBUG_DUMP)
1.232 + /* list the contents of the TOC each time .. not recommended */
1.233 + for(number=0; number<count; ++number) {
1.234 + fprintf(stderr, "\tx%d: %s\n", number, &base[toc->entry[number].nameOffset]);
1.235 + }
1.236 +#endif
1.237 + number=offsetTOCPrefixBinarySearch(tocEntryName, base, toc->entry, count);
1.238 + if(number>=0) {
1.239 + /* found it */
1.240 + const UDataOffsetTOCEntry *entry=toc->entry+number;
1.241 +#ifdef UDATA_DEBUG
1.242 + fprintf(stderr, "%s: Found.\n", tocEntryName);
1.243 +#endif
1.244 + if((number+1) < count) {
1.245 + *pLength = (int32_t)(entry[1].dataOffset - entry->dataOffset);
1.246 + } else {
1.247 + *pLength = -1;
1.248 + }
1.249 + return (const DataHeader *)(base+entry->dataOffset);
1.250 + } else {
1.251 +#ifdef UDATA_DEBUG
1.252 + fprintf(stderr, "%s: Not found.\n", tocEntryName);
1.253 +#endif
1.254 + return NULL;
1.255 + }
1.256 + } else {
1.257 +#ifdef UDATA_DEBUG
1.258 + fprintf(stderr, "returning header\n");
1.259 +#endif
1.260 +
1.261 + return pData->pHeader;
1.262 + }
1.263 +}
1.264 +
1.265 +
1.266 +static uint32_t pointerTOCEntryCount(const UDataMemory *pData) {
1.267 + const PointerTOC *toc = (PointerTOC *)pData->toc;
1.268 + return (uint32_t)((toc != NULL) ? (toc->count) : 0);
1.269 +}
1.270 +
1.271 +
1.272 +static const DataHeader *pointerTOCLookupFn(const UDataMemory *pData,
1.273 + const char *name,
1.274 + int32_t *pLength,
1.275 + UErrorCode *pErrorCode) {
1.276 + if(pData->toc!=NULL) {
1.277 + const PointerTOC *toc = (PointerTOC *)pData->toc;
1.278 + int32_t number, count=(int32_t)toc->count;
1.279 +
1.280 +#if defined (UDATA_DEBUG_DUMP)
1.281 + /* list the contents of the TOC each time .. not recommended */
1.282 + for(number=0; number<count; ++number) {
1.283 + fprintf(stderr, "\tx%d: %s\n", number, toc->entry[number].entryName);
1.284 + }
1.285 +#endif
1.286 + number=pointerTOCPrefixBinarySearch(name, toc->entry, count);
1.287 + if(number>=0) {
1.288 + /* found it */
1.289 +#ifdef UDATA_DEBUG
1.290 + fprintf(stderr, "%s: Found.\n", toc->entry[number].entryName);
1.291 +#endif
1.292 + *pLength=-1;
1.293 + return UDataMemory_normalizeDataPointer(toc->entry[number].pHeader);
1.294 + } else {
1.295 +#ifdef UDATA_DEBUG
1.296 + fprintf(stderr, "%s: Not found.\n", name);
1.297 +#endif
1.298 + return NULL;
1.299 + }
1.300 + } else {
1.301 + return pData->pHeader;
1.302 + }
1.303 +}
1.304 +
1.305 +static const commonDataFuncs CmnDFuncs = {offsetTOCLookupFn, offsetTOCEntryCount};
1.306 +static const commonDataFuncs ToCPFuncs = {pointerTOCLookupFn, pointerTOCEntryCount};
1.307 +
1.308 +
1.309 +
1.310 +/*----------------------------------------------------------------------*
1.311 + * *
1.312 + * checkCommonData Validate the format of a common data file. *
1.313 + * Fill in the virtual function ptr based on TOC type *
1.314 + * If the data is invalid, close the UDataMemory *
1.315 + * and set the appropriate error code. *
1.316 + * *
1.317 + *----------------------------------------------------------------------*/
1.318 +U_CFUNC void udata_checkCommonData(UDataMemory *udm, UErrorCode *err) {
1.319 + if (U_FAILURE(*err)) {
1.320 + return;
1.321 + }
1.322 +
1.323 + if(udm==NULL || udm->pHeader==NULL) {
1.324 + *err=U_INVALID_FORMAT_ERROR;
1.325 + } else if(!(udm->pHeader->dataHeader.magic1==0xda &&
1.326 + udm->pHeader->dataHeader.magic2==0x27 &&
1.327 + udm->pHeader->info.isBigEndian==U_IS_BIG_ENDIAN &&
1.328 + udm->pHeader->info.charsetFamily==U_CHARSET_FAMILY)
1.329 + ) {
1.330 + /* header not valid */
1.331 + *err=U_INVALID_FORMAT_ERROR;
1.332 + }
1.333 + else if (udm->pHeader->info.dataFormat[0]==0x43 &&
1.334 + udm->pHeader->info.dataFormat[1]==0x6d &&
1.335 + udm->pHeader->info.dataFormat[2]==0x6e &&
1.336 + udm->pHeader->info.dataFormat[3]==0x44 &&
1.337 + udm->pHeader->info.formatVersion[0]==1
1.338 + ) {
1.339 + /* dataFormat="CmnD" */
1.340 + udm->vFuncs = &CmnDFuncs;
1.341 + udm->toc=(const char *)udm->pHeader+udata_getHeaderSize(udm->pHeader);
1.342 + }
1.343 + else if(udm->pHeader->info.dataFormat[0]==0x54 &&
1.344 + udm->pHeader->info.dataFormat[1]==0x6f &&
1.345 + udm->pHeader->info.dataFormat[2]==0x43 &&
1.346 + udm->pHeader->info.dataFormat[3]==0x50 &&
1.347 + udm->pHeader->info.formatVersion[0]==1
1.348 + ) {
1.349 + /* dataFormat="ToCP" */
1.350 + udm->vFuncs = &ToCPFuncs;
1.351 + udm->toc=(const char *)udm->pHeader+udata_getHeaderSize(udm->pHeader);
1.352 + }
1.353 + else {
1.354 + /* dataFormat not recognized */
1.355 + *err=U_INVALID_FORMAT_ERROR;
1.356 + }
1.357 +
1.358 + if (U_FAILURE(*err)) {
1.359 + /* If the data is no good and we memory-mapped it ourselves,
1.360 + * close the memory mapping so it doesn't leak. Note that this has
1.361 + * no effect on non-memory mapped data, other than clearing fields in udm.
1.362 + */
1.363 + udata_close(udm);
1.364 + }
1.365 +}
1.366 +
1.367 +/*
1.368 + * TODO: Add a udata_swapPackageHeader() function that swaps an ICU .dat package
1.369 + * header but not its sub-items.
1.370 + * This function will be needed for automatic runtime swapping.
1.371 + * Sub-items should not be swapped to limit the swapping to the parts of the
1.372 + * package that are actually used.
1.373 + *
1.374 + * Since lengths of items are implicit in the order and offsets of their
1.375 + * ToC entries, and since offsets are relative to the start of the ToC,
1.376 + * a swapped version may need to generate a different data structure
1.377 + * with pointers to the original data items and with their lengths
1.378 + * (-1 for the last one if it is not known), and maybe even pointers to the
1.379 + * swapped versions of the items.
1.380 + * These pointers to swapped versions would establish a cache;
1.381 + * instead, each open data item could simply own the storage for its swapped
1.382 + * data. This fits better with the current design.
1.383 + *
1.384 + * markus 2003sep18 Jitterbug 2235
1.385 + */
