icuSources/common/ucmndata.c

   1 /*
   2 ******************************************************************************
   3 *
   4 *   Copyright (C) 1999-2004, International Business Machines
   5 *   Corporation and others.  All Rights Reserved.
   6 *
   7 ******************************************************************************/
   8
   9
  10 /*------------------------------------------------------------------------------
  11  *
  12  *   UCommonData   An abstract interface for dealing with ICU Common Data Files.
  13  *                 ICU Common Data Files are a grouping of a number of individual
  14  *                 data items (resources, converters, tables, anything) into a
  15  *                 single file or dll.  The combined format includes a table of
  16  *                 contents for locating the individual items by name.
  17  *
  18  *                 Two formats for the table of contents are supported, which is
  19  *                 why there is an abstract inteface involved.
  20  *
  21  */
  22
  23 #include "unicode/utypes.h"
  24 #include "unicode/udata.h"
  25 #include "cstring.h"
  26 #include "ucmndata.h"
  27 #include "udatamem.h"
  28
  29 #if defined(UDATA_DEBUG) || defined(UDATA_DEBUG_DUMP)
  30 #   include <stdio.h>
  31 #endif
  32
  33 U_CFUNC uint16_t
  34 udata_getHeaderSize(const DataHeader *udh) {
  35     if(udh==NULL) {
  36         return 0;
  37     } else if(udh->info.isBigEndian==U_IS_BIG_ENDIAN) {
  38         /* same endianness */
  39         return udh->dataHeader.headerSize;
  40     } else {
  41         /* opposite endianness */
  42         uint16_t x=udh->dataHeader.headerSize;
  43         return (uint16_t)((x<<8)|(x>>8));
  44     }
  45 }
  46
  47 U_CFUNC uint16_t
  48 udata_getInfoSize(const UDataInfo *info) {
  49     if(info==NULL) {
  50         return 0;
  51     } else if(info->isBigEndian==U_IS_BIG_ENDIAN) {
  52         /* same endianness */
  53         return info->size;
  54     } else {
  55         /* opposite endianness */
  56         uint16_t x=info->size;
  57         return (uint16_t)((x<<8)|(x>>8));
  58     }
  59 }
  60
  61 /*-----------------------------------------------------------------------------*
  62  *                                                                             *
  63  *  Pointer TOCs.   TODO: This form of table-of-contents should be removed     *
  64  *                  because DLLs must be relocated on loading to correct the   *
  65  *                  pointer values and this operation makes shared memory      *
  66  *                  mapping of the data much less likely to work.              *
  67  *                                                                             *
  68  *-----------------------------------------------------------------------------*/
  69 typedef struct {
  70     const char       *entryName;
  71     const DataHeader *pHeader;
  72 } PointerTOCEntry;
  73
  74
  75 typedef struct  {
  76     uint32_t          count;
  77     uint32_t          reserved;
  78     PointerTOCEntry   entry[2];   /* Actual size is from count. */
  79 }  PointerTOC;
  80
  81
  82 /* definition of OffsetTOC struct types moved to ucmndata.h */
  83
  84 /*-----------------------------------------------------------------------------*
  85  *                                                                             *
  86  *    entry point lookup implementations                                       *
  87  *                                                                             *
  88  *-----------------------------------------------------------------------------*/
  89 static uint32_t offsetTOCEntryCount(const UDataMemory *pData) {
  90     int32_t          retVal=0;
  91     const UDataOffsetTOC *toc = (UDataOffsetTOC *)pData->toc;
  92     if (toc != NULL) {
  93         retVal = toc->count;
  94     }
  95     return retVal;
  96 }
  97
  98
  99 static const DataHeader *
 100 offsetTOCLookupFn(const UDataMemory *pData,
 101                   const char *tocEntryName,
 102                   int32_t *pLength,
 103                   UErrorCode *pErrorCode) {
 104     const UDataOffsetTOC  *toc = (UDataOffsetTOC *)pData->toc;
 105     if(toc!=NULL) {
 106         const char *base=(const char *)toc;
 107         uint32_t start, limit, number, lastNumber;
 108         int32_t strResult;
 109         const UDataOffsetTOCEntry *entry;
 110
 111         /* perform a binary search for the data in the common data's table of contents */
 112 #if defined (UDATA_DEBUG_DUMP)
 113         /* list the contents of the TOC each time .. not recommended */
 114         for(start=0;start<toc->count;start++) {
 115           fprintf(stderr, "\tx%d: %s\n", start, &base[toc->entry[start].nameOffset]);
 116         }
 117 #endif
 118
 119         start=0;
 120         limit=toc->count;         /* number of names in this table of contents */
 121         lastNumber=limit;
 122         entry=toc->entry;
 123         for (;;) {
 124             number = (start+limit)/2;
 125             if (lastNumber == number) { /* Have we moved? */
 126                 break;  /* We haven't moved, and it wasn't found; */
 127                         /* or the empty stub common data library was used during build. */
 128             }
 129             lastNumber = number;
 130             strResult = uprv_strcmp(tocEntryName, base+entry[number].nameOffset);
 131             if(strResult<0) {
 132                 limit=number;
 133             } else if (strResult>0) {
 134                 start=number;
 135             }
 136             else {
 137                 /* found it */
 138 #ifdef UDATA_DEBUG
 139                 fprintf(stderr, "%s: Found.\n", tocEntryName);
 140 #endif
 141                 entry += number; /* Alias the entry to the current entry. */
 142                 if((number+1) < toc->count) {
 143                     *pLength = (int32_t)(entry[1].dataOffset - entry->dataOffset);
 144                 } else {
 145                     *pLength = -1;
 146                 }
 147                 return (const DataHeader *)(base+entry->dataOffset);
 148             }
 149         }
 150 #ifdef UDATA_DEBUG
 151         fprintf(stderr, "%s: Not found.\n", tocEntryName);
 152 #endif
 153         return NULL;
 154     } else {
 155 #ifdef UDATA_DEBUG
 156         fprintf(stderr, "returning header\n");
 157 #endif
 158
 159         return pData->pHeader;
 160     }
 161 }
 162
 163
 164 static uint32_t pointerTOCEntryCount(const UDataMemory *pData) {
 165     const PointerTOC *toc = (PointerTOC *)pData->toc;
 166     return (uint32_t)((toc != NULL) ? (toc->count) : 0);
 167 }
 168
 169
 170 static const DataHeader *pointerTOCLookupFn(const UDataMemory *pData,
 171                    const char *name,
 172                    int32_t *pLength,
 173                    UErrorCode *pErrorCode) {
 174     if(pData->toc!=NULL) {
 175         const PointerTOC *toc = (PointerTOC *)pData->toc;
 176         uint32_t start, limit, number, lastNumber;
 177         int32_t strResult;
 178
 179 #if defined (UDATA_DEBUG_DUMP)
 180         /* list the contents of the TOC each time .. not recommended */
 181         for(start=0;start<toc->count;start++) {
 182             fprintf(stderr, "\tx%d: %s\n", start, toc->entry[start].entryName);
 183         }
 184 #endif
 185
 186         /* perform a binary search for the data in the common data's table of contents */
 187         start=0;
 188         limit=toc->count;
 189         lastNumber=limit;
 190
 191         for (;;) {
 192             number = (start+limit)/2;
 193             if (lastNumber == number) { /* Have we moved? */
 194                 break;  /* We haven't moved, and it wasn't found, */
 195                         /* or the empty stub common data library was used during build. */
 196             }
 197             lastNumber = number;
 198             strResult = uprv_strcmp(name, toc->entry[number].entryName);
 199             if(strResult<0) {
 200                 limit=number;
 201             } else if (strResult>0) {
 202                 start=number;
 203             }
 204             else {
 205                 /* found it */
 206 #ifdef UDATA_DEBUG
 207                 fprintf(STDErr, "%s: Found.\n", toc->entry[number].entryName);
 208 #endif
 209                 *pLength=-1;
 210                 return UDataMemory_normalizeDataPointer(toc->entry[number].pHeader);
 211             }
 212         }
 213 #ifdef UDATA_DEBUG
 214         fprintf(stderr, "%s: Not found.\n", name);
 215 #endif
 216         return NULL;
 217     } else {
 218         return pData->pHeader;
 219     }
 220 }
 221
 222 static const commonDataFuncs CmnDFuncs = {offsetTOCLookupFn,  offsetTOCEntryCount};
 223 static const commonDataFuncs ToCPFuncs = {pointerTOCLookupFn, pointerTOCEntryCount};
 224
 225
 226
 227 /*----------------------------------------------------------------------*
 228  *                                                                      *
 229  *  checkCommonData   Validate the format of a common data file.        *
 230  *                    Fill in the virtual function ptr based on TOC type *
 231  *                    If the data is invalid, close the UDataMemory     *
 232  *                    and set the appropriate error code.               *
 233  *                                                                      *
 234  *----------------------------------------------------------------------*/
 235 void udata_checkCommonData(UDataMemory *udm, UErrorCode *err) {
 236     if (U_FAILURE(*err)) {
 237         return;
 238     }
 239
 240     if(!(udm->pHeader->dataHeader.magic1==0xda &&
 241         udm->pHeader->dataHeader.magic2==0x27 &&
 242         udm->pHeader->info.isBigEndian==U_IS_BIG_ENDIAN &&
 243         udm->pHeader->info.charsetFamily==U_CHARSET_FAMILY)
 244         ) {
 245         /* header not valid */
 246         *err=U_INVALID_FORMAT_ERROR;
 247     }
 248     else if (udm->pHeader->info.dataFormat[0]==0x43 &&
 249         udm->pHeader->info.dataFormat[1]==0x6d &&
 250         udm->pHeader->info.dataFormat[2]==0x6e &&
 251         udm->pHeader->info.dataFormat[3]==0x44 &&
 252         udm->pHeader->info.formatVersion[0]==1
 253         ) {
 254         /* dataFormat="CmnD" */
 255         udm->vFuncs = &CmnDFuncs;
 256         udm->toc=(const char *)udm->pHeader+udata_getHeaderSize(udm->pHeader);
 257     }
 258     else if(udm->pHeader->info.dataFormat[0]==0x54 &&
 259         udm->pHeader->info.dataFormat[1]==0x6f &&
 260         udm->pHeader->info.dataFormat[2]==0x43 &&
 261         udm->pHeader->info.dataFormat[3]==0x50 &&
 262         udm->pHeader->info.formatVersion[0]==1
 263         ) {
 264         /* dataFormat="ToCP" */
 265         udm->vFuncs = &ToCPFuncs;
 266         udm->toc=(const char *)udm->pHeader+udata_getHeaderSize(udm->pHeader);
 267     }
 268     else {
 269         /* dataFormat not recognized */
 270         *err=U_INVALID_FORMAT_ERROR;
 271     }
 272
 273     if (U_FAILURE(*err)) {
 274         /* If the data is no good and we memory-mapped it ourselves,
 275          *  close the memory mapping so it doesn't leak.  Note that this has
 276          *  no effect on non-memory mapped data, other than clearing fields in udm.
 277          */
 278         udata_close(udm);
 279     }
 280 }
 281
 282 /*
 283  * TODO: Add a udata_swapPackageHeader() function that swaps an ICU .dat package
 284  * header but not its sub-items.
 285  * This function will be needed for automatic runtime swapping.
 286  * Sub-items should not be swapped to limit the swapping to the parts of the
 287  * package that are actually used.
 288  *
 289  * Since lengths of items are implicit in the order and offsets of their
 290  * ToC entries, and since offsets are relative to the start of the ToC,
 291  * a swapped version may need to generate a different data structure
 292  * with pointers to the original data items and with their lengths
 293  * (-1 for the last one if it is not known), and maybe even pointers to the
 294  * swapped versions of the items.
 295  * These pointers to swapped versions would establish a cache;
 296  * instead, each open data item could simply own the storage for its swapped
 297  * data. This fits better with the current design.
 298  *
 299  * markus 2003sep18 Jitterbug 2235
 300  */