2 *******************************************************************************
4 * Copyright (C) 1999-2015, International Business Machines
5 * Corporation and others. All Rights Reserved.
7 *******************************************************************************
8 * file name: package.cpp
10 * tab size: 8 (not used)
13 * created on: 2005aug25
14 * created by: Markus W. Scherer
16 * Read, modify, and write ICU .dat data package files.
17 * This is an integral part of the icupkg tool, moved to the toolutil library
18 * because parts of tool implementations tend to be later shared by
20 * Subsumes functionality and implementation code from
21 * gencmn, decmn, and icuswap tools.
24 #include "unicode/utypes.h"
25 #include "unicode/putil.h"
26 #include "unicode/udata.h"
41 static const int32_t kItemsChunk
= 256; /* How much to increase the filesarray by each time */
43 // general definitions ----------------------------------------------------- ***
45 /* UDataInfo cf. udata.h */
46 static const UDataInfo dataInfo
={
47 (uint16_t)sizeof(UDataInfo
),
52 (uint8_t)sizeof(UChar
),
55 {0x43, 0x6d, 0x6e, 0x44}, /* dataFormat="CmnD" */
56 {1, 0, 0, 0}, /* formatVersion */
57 {3, 0, 0, 0} /* dataVersion */
61 static void U_CALLCONV
62 printPackageError(void *context
, const char *fmt
, va_list args
) {
63 vfprintf((FILE *)context
, fmt
, args
);
68 readSwapUInt16(uint16_t x
) {
69 return (uint16_t)((x
<<8)|(x
>>8));
72 // platform types ---------------------------------------------------------- ***
74 static const char *types
="lb?e";
76 enum { TYPE_L
, TYPE_B
, TYPE_LE
, TYPE_E
, TYPE_COUNT
};
79 makeTypeEnum(uint8_t charset
, UBool isBigEndian
) {
80 return 2*(int32_t)charset
+isBigEndian
;
84 makeTypeEnum(char type
) {
86 type
== 'l' ? TYPE_L
:
87 type
== 'b' ? TYPE_B
:
88 type
== 'e' ? TYPE_E
:
93 makeTypeLetter(uint8_t charset
, UBool isBigEndian
) {
94 return types
[makeTypeEnum(charset
, isBigEndian
)];
98 makeTypeLetter(int32_t typeEnum
) {
99 return types
[typeEnum
];
103 makeTypeProps(char type
, uint8_t &charset
, UBool
&isBigEndian
) {
104 int32_t typeEnum
=makeTypeEnum(type
);
105 charset
=(uint8_t)(typeEnum
>>1);
106 isBigEndian
=(UBool
)(typeEnum
&1);
109 U_CFUNC
const UDataInfo
*
110 getDataInfo(const uint8_t *data
, int32_t length
,
111 int32_t &infoLength
, int32_t &headerLength
,
112 UErrorCode
*pErrorCode
) {
113 const DataHeader
*pHeader
;
114 const UDataInfo
*pInfo
;
116 if(pErrorCode
==NULL
|| U_FAILURE(*pErrorCode
)) {
120 (length
>=0 && length
<(int32_t)sizeof(DataHeader
))
122 *pErrorCode
=U_ILLEGAL_ARGUMENT_ERROR
;
126 pHeader
=(const DataHeader
*)data
;
127 pInfo
=&pHeader
->info
;
128 if( (length
>=0 && length
<(int32_t)sizeof(DataHeader
)) ||
129 pHeader
->dataHeader
.magic1
!=0xda ||
130 pHeader
->dataHeader
.magic2
!=0x27 ||
131 pInfo
->sizeofUChar
!=2
133 *pErrorCode
=U_UNSUPPORTED_ERROR
;
137 if(pInfo
->isBigEndian
==U_IS_BIG_ENDIAN
) {
138 headerLength
=pHeader
->dataHeader
.headerSize
;
139 infoLength
=pInfo
->size
;
141 headerLength
=readSwapUInt16(pHeader
->dataHeader
.headerSize
);
142 infoLength
=readSwapUInt16(pInfo
->size
);
145 if( headerLength
<(int32_t)sizeof(DataHeader
) ||
146 infoLength
<(int32_t)sizeof(UDataInfo
) ||
147 headerLength
<(int32_t)(sizeof(pHeader
->dataHeader
)+infoLength
) ||
148 (length
>=0 && length
<headerLength
)
150 *pErrorCode
=U_UNSUPPORTED_ERROR
;
158 getTypeEnumForInputData(const uint8_t *data
, int32_t length
,
159 UErrorCode
*pErrorCode
) {
160 const UDataInfo
*pInfo
;
161 int32_t infoLength
, headerLength
;
163 /* getDataInfo() checks for illegal arguments */
164 pInfo
=getDataInfo(data
, length
, infoLength
, headerLength
, pErrorCode
);
169 return makeTypeEnum(pInfo
->charsetFamily
, (UBool
)pInfo
->isBigEndian
);
172 // file handling ----------------------------------------------------------- ***
175 extractPackageName(const char *filename
,
176 char pkg
[], int32_t capacity
) {
177 const char *basename
;
180 basename
=findBasename(filename
);
181 len
=(int32_t)strlen(basename
)-4; /* -4: subtract the length of ".dat" */
183 if(len
<=0 || 0!=strcmp(basename
+len
, ".dat")) {
184 fprintf(stderr
, "icupkg: \"%s\" is not recognized as a package filename (must end with .dat)\n",
186 exit(U_ILLEGAL_ARGUMENT_ERROR
);
190 fprintf(stderr
, "icupkg: the package name \"%s\" is too long (>=%ld)\n",
191 basename
, (long)capacity
);
192 exit(U_ILLEGAL_ARGUMENT_ERROR
);
195 memcpy(pkg
, basename
, len
);
200 getFileLength(FILE *f
) {
203 fseek(f
, 0, SEEK_END
);
204 length
=(int32_t)ftell(f
);
205 fseek(f
, 0, SEEK_SET
);
210 * Turn tree separators and alternate file separators into normal file separators.
212 #if U_TREE_ENTRY_SEP_CHAR==U_FILE_SEP_CHAR && U_FILE_ALT_SEP_CHAR==U_FILE_SEP_CHAR
213 #define treeToPath(s)
216 treeToPath(char *s
) {
219 for(t
=s
; *t
!=0; ++t
) {
220 if(*t
==U_TREE_ENTRY_SEP_CHAR
|| *t
==U_FILE_ALT_SEP_CHAR
) {
228 * Turn file separators into tree separators.
230 #if U_TREE_ENTRY_SEP_CHAR==U_FILE_SEP_CHAR && U_FILE_ALT_SEP_CHAR==U_FILE_SEP_CHAR
231 #define pathToTree(s)
234 pathToTree(char *s
) {
237 for(t
=s
; *t
!=0; ++t
) {
238 if(*t
==U_FILE_SEP_CHAR
|| *t
==U_FILE_ALT_SEP_CHAR
) {
239 *t
=U_TREE_ENTRY_SEP_CHAR
;
246 * Prepend the path (if any) to the name and run the name through treeToName().
249 makeFullFilename(const char *path
, const char *name
,
250 char *filename
, int32_t capacity
) {
253 // prepend the path unless NULL or empty
254 if(path
!=NULL
&& path
[0]!=0) {
255 if((int32_t)(strlen(path
)+1)>=capacity
) {
256 fprintf(stderr
, "pathname too long: \"%s\"\n", path
);
257 exit(U_BUFFER_OVERFLOW_ERROR
);
259 strcpy(filename
, path
);
261 // make sure the path ends with a file separator
262 s
=strchr(filename
, 0);
263 if(*(s
-1)!=U_FILE_SEP_CHAR
&& *(s
-1)!=U_FILE_ALT_SEP_CHAR
) {
264 *s
++=U_FILE_SEP_CHAR
;
270 // turn the name into a filename, turn tree separators into file separators
271 if((int32_t)((s
-filename
)+strlen(name
))>=capacity
) {
272 fprintf(stderr
, "path/filename too long: \"%s%s\"\n", filename
, name
);
273 exit(U_BUFFER_OVERFLOW_ERROR
);
280 makeFullFilenameAndDirs(const char *path
, const char *name
,
281 char *filename
, int32_t capacity
) {
283 UErrorCode errorCode
;
285 makeFullFilename(path
, name
, filename
, capacity
);
287 // make tree directories
288 errorCode
=U_ZERO_ERROR
;
289 sep
=strchr(filename
, 0)-strlen(name
);
290 while((sep
=strchr(sep
, U_FILE_SEP_CHAR
))!=NULL
) {
292 *sep
=0; // truncate temporarily
293 uprv_mkdir(filename
, &errorCode
);
294 if(U_FAILURE(errorCode
)) {
295 fprintf(stderr
, "icupkg: unable to create tree directory \"%s\"\n", filename
);
296 exit(U_FILE_ACCESS_ERROR
);
299 *sep
++=U_FILE_SEP_CHAR
; // restore file separator character
304 readFile(const char *path
, const char *name
, int32_t &length
, char &type
) {
307 UErrorCode errorCode
;
308 int32_t fileLength
, typeEnum
;
310 makeFullFilename(path
, name
, filename
, (int32_t)sizeof(filename
));
312 /* open the input file, get its length, allocate memory for it, read the file */
313 file
=fopen(filename
, "rb");
315 fprintf(stderr
, "icupkg: unable to open input file \"%s\"\n", filename
);
316 exit(U_FILE_ACCESS_ERROR
);
319 /* get the file length */
320 fileLength
=getFileLength(file
);
321 if(ferror(file
) || fileLength
<=0) {
322 fprintf(stderr
, "icupkg: empty input file \"%s\"\n", filename
);
324 exit(U_FILE_ACCESS_ERROR
);
327 /* allocate the buffer, pad to multiple of 16 */
328 length
=(fileLength
+0xf)&~0xf;
329 icu::LocalMemory
<uint8_t> data((uint8_t *)uprv_malloc(length
));
332 fprintf(stderr
, "icupkg: malloc error allocating %d bytes.\n", (int)length
);
333 exit(U_MEMORY_ALLOCATION_ERROR
);
337 if(fileLength
!=(int32_t)fread(data
.getAlias(), 1, fileLength
, file
)) {
338 fprintf(stderr
, "icupkg: error reading \"%s\"\n", filename
);
340 exit(U_FILE_ACCESS_ERROR
);
343 /* pad the file to a multiple of 16 using the usual padding byte */
344 if(fileLength
<length
) {
345 memset(data
.getAlias()+fileLength
, 0xaa, length
-fileLength
);
350 // minimum check for ICU-format data
351 errorCode
=U_ZERO_ERROR
;
352 typeEnum
=getTypeEnumForInputData(data
.getAlias(), length
, &errorCode
);
353 if(typeEnum
<0 || U_FAILURE(errorCode
)) {
354 fprintf(stderr
, "icupkg: not an ICU data file: \"%s\"\n", filename
);
355 #if !UCONFIG_NO_LEGACY_CONVERSION
356 exit(U_INVALID_FORMAT_ERROR
);
358 fprintf(stderr
, "U_INVALID_FORMAT_ERROR occurred but UCONFIG_NO_LEGACY_CONVERSION is on so this is expected.\n");
362 type
=makeTypeLetter(typeEnum
);
364 return data
.orphan();
367 // .dat package file representation ---------------------------------------- ***
371 static int32_t U_CALLCONV
372 compareItems(const void * /*context*/, const void *left
, const void *right
) {
375 return (int32_t)strcmp(((Item
*)left
)->name
, ((Item
*)right
)->name
);
383 : doAutoPrefix(FALSE
), prefixEndsWithType(FALSE
) {
388 inCharset
=U_CHARSET_FAMILY
;
389 inIsBigEndian
=U_IS_BIG_ENDIAN
;
395 inStringTop
=outStringTop
=0;
398 findPrefix
=findSuffix
=NULL
;
399 findPrefixLength
=findSuffixLength
=0;
402 // create a header for an empty package
404 pHeader
=(DataHeader
*)header
;
405 pHeader
->dataHeader
.magic1
=0xda;
406 pHeader
->dataHeader
.magic2
=0x27;
407 memcpy(&pHeader
->info
, &dataInfo
, sizeof(dataInfo
));
408 headerLength
=(int32_t)(4+sizeof(dataInfo
));
409 if(headerLength
&0xf) {
410 /* NUL-pad the header to a multiple of 16 */
411 int32_t length
=(headerLength
+0xf)&~0xf;
412 memset(header
+headerLength
, 0, length
-headerLength
);
415 pHeader
->dataHeader
.headerSize
=(uint16_t)headerLength
;
418 Package::~Package() {
423 for(idx
=0; idx
<itemCount
; ++idx
) {
424 if(items
[idx
].isDataOwned
) {
425 uprv_free(items
[idx
].data
);
429 uprv_free((void*)items
);
433 Package::setPrefix(const char *p
) {
434 if(strlen(p
)>=sizeof(pkgPrefix
)) {
435 fprintf(stderr
, "icupkg: --toc_prefix %s too long\n", p
);
436 exit(U_ILLEGAL_ARGUMENT_ERROR
);
438 strcpy(pkgPrefix
, p
);
442 Package::readPackage(const char *filename
) {
444 const UDataInfo
*pInfo
;
445 UErrorCode errorCode
;
447 const uint8_t *inBytes
;
449 int32_t length
, offset
, i
;
450 int32_t itemLength
, typeEnum
;
453 const UDataOffsetTOCEntry
*inEntries
;
455 extractPackageName(filename
, inPkgName
, (int32_t)sizeof(inPkgName
));
458 inData
=readFile(NULL
, filename
, inLength
, type
);
462 * swap the header - even if the swapping itself is a no-op
463 * because it tells us the header length
465 errorCode
=U_ZERO_ERROR
;
466 makeTypeProps(type
, inCharset
, inIsBigEndian
);
467 ds
=udata_openSwapper(inIsBigEndian
, inCharset
, U_IS_BIG_ENDIAN
, U_CHARSET_FAMILY
, &errorCode
);
468 if(U_FAILURE(errorCode
)) {
469 fprintf(stderr
, "icupkg: udata_openSwapper(\"%s\") failed - %s\n",
470 filename
, u_errorName(errorCode
));
474 ds
->printError
=printPackageError
;
475 ds
->printErrorContext
=stderr
;
477 headerLength
=sizeof(header
);
478 if(length
<headerLength
) {
481 headerLength
=udata_swapDataHeader(ds
, inData
, headerLength
, header
, &errorCode
);
482 if(U_FAILURE(errorCode
)) {
486 /* check data format and format version */
487 pInfo
=(const UDataInfo
*)((const char *)inData
+4);
489 pInfo
->dataFormat
[0]==0x43 && /* dataFormat="CmnD" */
490 pInfo
->dataFormat
[1]==0x6d &&
491 pInfo
->dataFormat
[2]==0x6e &&
492 pInfo
->dataFormat
[3]==0x44 &&
493 pInfo
->formatVersion
[0]==1
495 fprintf(stderr
, "icupkg: data format %02x.%02x.%02x.%02x (format version %02x) is not recognized as an ICU .dat package\n",
496 pInfo
->dataFormat
[0], pInfo
->dataFormat
[1],
497 pInfo
->dataFormat
[2], pInfo
->dataFormat
[3],
498 pInfo
->formatVersion
[0]);
499 exit(U_UNSUPPORTED_ERROR
);
501 inIsBigEndian
=(UBool
)pInfo
->isBigEndian
;
502 inCharset
=pInfo
->charsetFamily
;
504 inBytes
=(const uint8_t *)inData
+headerLength
;
505 inEntries
=(const UDataOffsetTOCEntry
*)(inBytes
+4);
507 /* check that the itemCount fits, then the ToC table, then at least the header of the last item */
508 length
-=headerLength
;
510 /* itemCount does not fit */
513 itemCount
=udata_readInt32(ds
, *(const int32_t *)inBytes
);
514 setItemCapacity(itemCount
); /* resize so there's space */
517 } else if(length
<(4+8*itemCount
)) {
518 /* ToC table does not fit */
521 /* offset of the last item plus at least 20 bytes for its header */
522 offset
=20+(int32_t)ds
->readUInt32(inEntries
[itemCount
-1].dataOffset
);
526 fprintf(stderr
, "icupkg: too few bytes (%ld after header) for a .dat package\n",
528 exit(U_INDEX_OUTOFBOUNDS_ERROR
);
530 /* do not modify the package length variable until the last item's length is set */
534 fprintf(stderr
, "icupkg: --auto_toc_prefix[_with_type] but the input package is empty\n");
535 exit(U_INVALID_FORMAT_ERROR
);
538 char prefix
[MAX_PKG_NAME_LENGTH
+4];
539 char *s
, *inItemStrings
;
541 if(itemCount
>itemMax
) {
542 fprintf(stderr
, "icupkg: too many items, maximum is %d\n", itemMax
);
543 exit(U_BUFFER_OVERFLOW_ERROR
);
546 /* swap the item name strings */
547 int32_t stringsOffset
=4+8*itemCount
;
548 itemLength
=(int32_t)(ds
->readUInt32(inEntries
[0].dataOffset
))-stringsOffset
;
550 // don't include padding bytes at the end of the item names
551 while(itemLength
>0 && inBytes
[stringsOffset
+itemLength
-1]!=0) {
555 if((inStringTop
+itemLength
)>STRING_STORE_SIZE
) {
556 fprintf(stderr
, "icupkg: total length of item name strings too long\n");
557 exit(U_BUFFER_OVERFLOW_ERROR
);
560 inItemStrings
=inStrings
+inStringTop
;
561 ds
->swapInvChars(ds
, inBytes
+stringsOffset
, itemLength
, inItemStrings
, &errorCode
);
562 if(U_FAILURE(errorCode
)) {
563 fprintf(stderr
, "icupkg failed to swap the input .dat package item name strings\n");
564 exit(U_INVALID_FORMAT_ERROR
);
566 inStringTop
+=itemLength
;
568 // reset the Item entries
569 memset(items
, 0, itemCount
*sizeof(Item
));
572 * Get the common prefix of the items.
573 * New-style ICU .dat packages use tree separators ('/') between package names,
574 * tree names, and item names,
575 * while old-style ICU .dat packages (before multi-tree support)
576 * use an underscore ('_') between package and item names.
578 offset
=(int32_t)ds
->readUInt32(inEntries
[0].nameOffset
)-stringsOffset
;
579 s
=inItemStrings
+offset
; // name of the first entry
580 int32_t prefixLength
;
582 // Use the first entry's prefix. Must be a new-style package.
583 const char *prefixLimit
=strchr(s
, U_TREE_ENTRY_SEP_CHAR
);
584 if(prefixLimit
==NULL
) {
586 "icupkg: --auto_toc_prefix[_with_type] but "
587 "the first entry \"%s\" does not contain a '%c'\n",
588 s
, U_TREE_ENTRY_SEP_CHAR
);
589 exit(U_INVALID_FORMAT_ERROR
);
591 prefixLength
=(int32_t)(prefixLimit
-s
);
592 if(prefixLength
==0 || prefixLength
>=UPRV_LENGTHOF(pkgPrefix
)) {
594 "icupkg: --auto_toc_prefix[_with_type] but "
595 "the prefix of the first entry \"%s\" is empty or too long\n",
597 exit(U_INVALID_FORMAT_ERROR
);
599 if(prefixEndsWithType
&& s
[prefixLength
-1]!=type
) {
601 "icupkg: --auto_toc_prefix_with_type but "
602 "the prefix of the first entry \"%s\" does not end with '%c'\n",
604 exit(U_INVALID_FORMAT_ERROR
);
606 memcpy(pkgPrefix
, s
, prefixLength
);
607 pkgPrefix
[prefixLength
]=0;
608 memcpy(prefix
, s
, ++prefixLength
); // include the /
610 // Use the package basename as prefix.
611 int32_t inPkgNameLength
=strlen(inPkgName
);
612 memcpy(prefix
, inPkgName
, inPkgNameLength
);
613 prefixLength
=inPkgNameLength
;
615 if( (int32_t)strlen(s
)>=(inPkgNameLength
+2) &&
616 0==memcmp(s
, inPkgName
, inPkgNameLength
) &&
617 s
[inPkgNameLength
]=='_'
619 // old-style .dat package
620 prefix
[prefixLength
++]='_';
622 // new-style .dat package
623 prefix
[prefixLength
++]=U_TREE_ENTRY_SEP_CHAR
;
624 // if it turns out to not contain U_TREE_ENTRY_SEP_CHAR
625 // then the test in the loop below will fail
628 prefix
[prefixLength
]=0;
630 /* read the ToC table */
631 for(i
=0; i
<itemCount
; ++i
) {
632 // skip the package part of the item name, error if it does not match the actual package name
633 // or if nothing follows the package name
634 offset
=(int32_t)ds
->readUInt32(inEntries
[i
].nameOffset
)-stringsOffset
;
635 s
=inItemStrings
+offset
;
636 if(0!=strncmp(s
, prefix
, prefixLength
) || s
[prefixLength
]==0) {
637 fprintf(stderr
, "icupkg: input .dat item name \"%s\" does not start with \"%s\"\n",
639 exit(U_INVALID_FORMAT_ERROR
);
641 items
[i
].name
=s
+prefixLength
;
643 // set the item's data
644 items
[i
].data
=(uint8_t *)inBytes
+ds
->readUInt32(inEntries
[i
].dataOffset
);
646 items
[i
-1].length
=(int32_t)(items
[i
].data
-items
[i
-1].data
);
648 // set the previous item's platform type
649 typeEnum
=getTypeEnumForInputData(items
[i
-1].data
, items
[i
-1].length
, &errorCode
);
650 if(typeEnum
<0 || U_FAILURE(errorCode
)) {
651 fprintf(stderr
, "icupkg: not an ICU data file: item \"%s\" in \"%s\"\n", items
[i
-1].name
, filename
);
652 exit(U_INVALID_FORMAT_ERROR
);
654 items
[i
-1].type
=makeTypeLetter(typeEnum
);
656 items
[i
].isDataOwned
=FALSE
;
658 // set the last item's length
659 items
[itemCount
-1].length
=length
-ds
->readUInt32(inEntries
[itemCount
-1].dataOffset
);
661 // set the last item's platform type
662 typeEnum
=getTypeEnumForInputData(items
[itemCount
-1].data
, items
[itemCount
-1].length
, &errorCode
);
663 if(typeEnum
<0 || U_FAILURE(errorCode
)) {
664 fprintf(stderr
, "icupkg: not an ICU data file: item \"%s\" in \"%s\"\n", items
[i
-1].name
, filename
);
665 exit(U_INVALID_FORMAT_ERROR
);
667 items
[itemCount
-1].type
=makeTypeLetter(typeEnum
);
669 if(type
!=U_ICUDATA_TYPE_LETTER
[0]) {
670 // sort the item names for the local charset
675 udata_closeSwapper(ds
);
679 Package::getInType() {
680 return makeTypeLetter(inCharset
, inIsBigEndian
);
684 Package::writePackage(const char *filename
, char outType
, const char *comment
) {
685 char prefix
[MAX_PKG_NAME_LENGTH
+4];
686 UDataOffsetTOCEntry entry
;
687 UDataSwapper
*dsLocalToOut
, *ds
[TYPE_COUNT
];
691 UErrorCode errorCode
;
692 int32_t i
, length
, prefixLength
, maxItemLength
, basenameOffset
, offset
, outInt32
;
694 UBool outIsBigEndian
;
696 extractPackageName(filename
, prefix
, MAX_PKG_NAME_LENGTH
);
698 // if there is an explicit comment, then use it, else use what's in the current header
700 /* get the header size minus the current comment */
704 pHeader
=(DataHeader
*)header
;
705 headerLength
=4+pHeader
->info
.size
;
706 length
=(int32_t)strlen(comment
);
707 if((int32_t)(headerLength
+length
)>=(int32_t)sizeof(header
)) {
708 fprintf(stderr
, "icupkg: comment too long\n");
709 exit(U_BUFFER_OVERFLOW_ERROR
);
711 memcpy(header
+headerLength
, comment
, length
+1);
712 headerLength
+=length
;
713 if(headerLength
&0xf) {
714 /* NUL-pad the header to a multiple of 16 */
715 length
=(headerLength
+0xf)&~0xf;
716 memset(header
+headerLength
, 0, length
-headerLength
);
719 pHeader
->dataHeader
.headerSize
=(uint16_t)headerLength
;
722 makeTypeProps(outType
, outCharset
, outIsBigEndian
);
724 // open (TYPE_COUNT-2) swappers
725 // one is a no-op for local type==outType
726 // one type (TYPE_LE) is bogus
727 errorCode
=U_ZERO_ERROR
;
728 i
=makeTypeEnum(outType
);
729 ds
[TYPE_B
]= i
==TYPE_B
? NULL
: udata_openSwapper(TRUE
, U_ASCII_FAMILY
, outIsBigEndian
, outCharset
, &errorCode
);
730 ds
[TYPE_L
]= i
==TYPE_L
? NULL
: udata_openSwapper(FALSE
, U_ASCII_FAMILY
, outIsBigEndian
, outCharset
, &errorCode
);
732 ds
[TYPE_E
]= i
==TYPE_E
? NULL
: udata_openSwapper(TRUE
, U_EBCDIC_FAMILY
, outIsBigEndian
, outCharset
, &errorCode
);
733 if(U_FAILURE(errorCode
)) {
734 fprintf(stderr
, "icupkg: udata_openSwapper() failed - %s\n", u_errorName(errorCode
));
737 for(i
=0; i
<TYPE_COUNT
; ++i
) {
739 ds
[i
]->printError
=printPackageError
;
740 ds
[i
]->printErrorContext
=stderr
;
744 dsLocalToOut
=ds
[makeTypeEnum(U_CHARSET_FAMILY
, U_IS_BIG_ENDIAN
)];
746 // create the file and write its contents
747 file
=fopen(filename
, "wb");
749 fprintf(stderr
, "icupkg: unable to create file \"%s\"\n", filename
);
750 exit(U_FILE_ACCESS_ERROR
);
753 // swap and write the header
754 if(dsLocalToOut
!=NULL
) {
755 udata_swapDataHeader(dsLocalToOut
, header
, headerLength
, header
, &errorCode
);
756 if(U_FAILURE(errorCode
)) {
757 fprintf(stderr
, "icupkg: udata_swapDataHeader(local to out) failed - %s\n", u_errorName(errorCode
));
761 length
=(int32_t)fwrite(header
, 1, headerLength
, file
);
762 if(length
!=headerLength
) {
763 fprintf(stderr
, "icupkg: unable to write complete header to file \"%s\"\n", filename
);
764 exit(U_FILE_ACCESS_ERROR
);
767 // prepare and swap the package name with a tree separator
768 // for prepending to item names
769 if(pkgPrefix
[0]==0) {
770 prefixLength
=(int32_t)strlen(prefix
);
772 prefixLength
=(int32_t)strlen(pkgPrefix
);
773 memcpy(prefix
, pkgPrefix
, prefixLength
);
774 if(prefixEndsWithType
) {
775 prefix
[prefixLength
-1]=outType
;
778 prefix
[prefixLength
++]=U_TREE_ENTRY_SEP_CHAR
;
779 prefix
[prefixLength
]=0;
780 if(dsLocalToOut
!=NULL
) {
781 dsLocalToOut
->swapInvChars(dsLocalToOut
, prefix
, prefixLength
, prefix
, &errorCode
);
782 if(U_FAILURE(errorCode
)) {
783 fprintf(stderr
, "icupkg: swapInvChars(output package name) failed - %s\n", u_errorName(errorCode
));
787 // swap and sort the item names (sorting needs to be done in the output charset)
788 dsLocalToOut
->swapInvChars(dsLocalToOut
, inStrings
, inStringTop
, inStrings
, &errorCode
);
789 if(U_FAILURE(errorCode
)) {
790 fprintf(stderr
, "icupkg: swapInvChars(item names) failed - %s\n", u_errorName(errorCode
));
796 // create the output item names in sorted order, with the package name prepended to each
797 for(i
=0; i
<itemCount
; ++i
) {
798 length
=(int32_t)strlen(items
[i
].name
);
799 name
=allocString(FALSE
, length
+prefixLength
);
800 memcpy(name
, prefix
, prefixLength
);
801 memcpy(name
+prefixLength
, items
[i
].name
, length
+1);
805 // calculate offsets for item names and items, pad to 16-align items
806 // align only the first item; each item's length is a multiple of 16
807 basenameOffset
=4+8*itemCount
;
808 offset
=basenameOffset
+outStringTop
;
809 if((length
=(offset
&15))!=0) {
811 memset(allocString(FALSE
, length
-1), 0xaa, length
);
815 // write the table of contents
816 // first the itemCount
818 if(dsLocalToOut
!=NULL
) {
819 dsLocalToOut
->swapArray32(dsLocalToOut
, &outInt32
, 4, &outInt32
, &errorCode
);
820 if(U_FAILURE(errorCode
)) {
821 fprintf(stderr
, "icupkg: swapArray32(item count) failed - %s\n", u_errorName(errorCode
));
825 length
=(int32_t)fwrite(&outInt32
, 1, 4, file
);
827 fprintf(stderr
, "icupkg: unable to write complete item count to file \"%s\"\n", filename
);
828 exit(U_FILE_ACCESS_ERROR
);
831 // then write the item entries (and collect the maxItemLength)
833 for(i
=0; i
<itemCount
; ++i
) {
834 entry
.nameOffset
=(uint32_t)(basenameOffset
+(items
[i
].name
-outStrings
));
835 entry
.dataOffset
=(uint32_t)offset
;
836 if(dsLocalToOut
!=NULL
) {
837 dsLocalToOut
->swapArray32(dsLocalToOut
, &entry
, 8, &entry
, &errorCode
);
838 if(U_FAILURE(errorCode
)) {
839 fprintf(stderr
, "icupkg: swapArray32(item entry %ld) failed - %s\n", (long)i
, u_errorName(errorCode
));
843 length
=(int32_t)fwrite(&entry
, 1, 8, file
);
845 fprintf(stderr
, "icupkg: unable to write complete item entry %ld to file \"%s\"\n", (long)i
, filename
);
846 exit(U_FILE_ACCESS_ERROR
);
849 length
=items
[i
].length
;
850 if(length
>maxItemLength
) {
851 maxItemLength
=length
;
856 // write the item names
857 length
=(int32_t)fwrite(outStrings
, 1, outStringTop
, file
);
858 if(length
!=outStringTop
) {
859 fprintf(stderr
, "icupkg: unable to write complete item names to file \"%s\"\n", filename
);
860 exit(U_FILE_ACCESS_ERROR
);
864 for(pItem
=items
, i
=0; i
<itemCount
; ++pItem
, ++i
) {
865 int32_t type
=makeTypeEnum(pItem
->type
);
867 // swap each item from its platform properties to the desired ones
870 pItem
->data
, pItem
->length
, pItem
->data
,
872 if(U_FAILURE(errorCode
)) {
873 fprintf(stderr
, "icupkg: udata_swap(item %ld) failed - %s\n", (long)i
, u_errorName(errorCode
));
877 length
=(int32_t)fwrite(pItem
->data
, 1, pItem
->length
, file
);
878 if(length
!=pItem
->length
) {
879 fprintf(stderr
, "icupkg: unable to write complete item %ld to file \"%s\"\n", (long)i
, filename
);
880 exit(U_FILE_ACCESS_ERROR
);
885 fprintf(stderr
, "icupkg: unable to write complete file \"%s\"\n", filename
);
886 exit(U_FILE_ACCESS_ERROR
);
890 for(i
=0; i
<TYPE_COUNT
; ++i
) {
891 udata_closeSwapper(ds
[i
]);
896 Package::findItem(const char *name
, int32_t length
) const {
897 int32_t i
, start
, limit
;
900 /* do a binary search for the string */
906 result
=strncmp(name
, items
[i
].name
, length
);
908 result
=strcmp(name
, items
[i
].name
);
915 * if we compared just prefixes, then we may need to back up
916 * to the first item with this prefix
918 while(i
>0 && 0==strncmp(name
, items
[i
-1].name
, length
)) {
923 } else if(result
<0) {
925 } else /* result>0 */ {
930 return ~start
; /* not found, return binary-not of the insertion point */
934 Package::findItems(const char *pattern
) {
937 if(pattern
==NULL
|| *pattern
==0) {
946 wild
=strchr(pattern
, '*');
949 findPrefixLength
=(int32_t)strlen(pattern
);
952 findPrefixLength
=(int32_t)(wild
-pattern
);
954 findSuffixLength
=(int32_t)strlen(findSuffix
);
955 if(NULL
!=strchr(findSuffix
, '*')) {
956 // two or more wildcards
957 fprintf(stderr
, "icupkg: syntax error (more than one '*') in item pattern \"%s\"\n", pattern
);
962 if(findPrefixLength
==0) {
965 findNextIndex
=findItem(findPrefix
, findPrefixLength
);
970 Package::findNextItem() {
971 const char *name
, *middle
, *treeSep
;
972 int32_t idx
, nameLength
, middleLength
;
974 if(findNextIndex
<0) {
978 while(findNextIndex
<itemCount
) {
980 name
=items
[idx
].name
;
981 nameLength
=(int32_t)strlen(name
);
982 if(nameLength
<(findPrefixLength
+findSuffixLength
)) {
983 // item name too short for prefix & suffix
986 if(findPrefixLength
>0 && 0!=memcmp(findPrefix
, name
, findPrefixLength
)) {
987 // left the range of names with this prefix
990 middle
=name
+findPrefixLength
;
991 middleLength
=nameLength
-findPrefixLength
-findSuffixLength
;
992 if(findSuffixLength
>0 && 0!=memcmp(findSuffix
, name
+(nameLength
-findSuffixLength
), findSuffixLength
)) {
993 // suffix does not match
996 // prefix & suffix match
998 if(matchMode
&MATCH_NOSLASH
) {
999 treeSep
=strchr(middle
, U_TREE_ENTRY_SEP_CHAR
);
1000 if(treeSep
!=NULL
&& (treeSep
-middle
)<middleLength
) {
1001 // the middle (matching the * wildcard) contains a tree separator /
1006 // found a matching item
1016 Package::setMatchMode(uint32_t mode
) {
1021 Package::addItem(const char *name
) {
1022 addItem(name
, NULL
, 0, FALSE
, U_ICUDATA_TYPE_LETTER
[0]);
1026 Package::addItem(const char *name
, uint8_t *data
, int32_t length
, UBool isDataOwned
, char type
) {
1031 // new item, make space at the insertion point
1032 ensureItemCapacity();
1033 // move the following items down
1036 memmove(items
+idx
+1, items
+idx
, (itemCount
-idx
)*sizeof(Item
));
1040 // reset this Item entry
1041 memset(items
+idx
, 0, sizeof(Item
));
1043 // copy the item's name
1044 items
[idx
].name
=allocString(TRUE
, strlen(name
));
1045 strcpy(items
[idx
].name
, name
);
1046 pathToTree(items
[idx
].name
);
1048 // same-name item found, replace it
1049 if(items
[idx
].isDataOwned
) {
1050 uprv_free(items
[idx
].data
);
1053 // keep the item's name since it is the same
1056 // set the item's data
1057 items
[idx
].data
=data
;
1058 items
[idx
].length
=length
;
1059 items
[idx
].isDataOwned
=isDataOwned
;
1060 items
[idx
].type
=type
;
1064 Package::addFile(const char *filesPath
, const char *name
) {
1069 data
=readFile(filesPath
, name
, length
, type
);
1070 // readFile() exits the tool if it fails
1071 addItem(name
, data
, length
, TRUE
, type
);
1075 Package::addItems(const Package
&listPkg
) {
1079 for(pItem
=listPkg
.items
, i
=0; i
<listPkg
.itemCount
; ++pItem
, ++i
) {
1080 addItem(pItem
->name
, pItem
->data
, pItem
->length
, FALSE
, pItem
->type
);
1085 Package::removeItem(int32_t idx
) {
1088 if(items
[idx
].isDataOwned
) {
1089 uprv_free(items
[idx
].data
);
1092 // move the following items up
1093 if((idx
+1)<itemCount
) {
1094 memmove(items
+idx
, items
+idx
+1, (itemCount
-(idx
+1))*sizeof(Item
));
1098 if(idx
<=findNextIndex
) {
1105 Package::removeItems(const char *pattern
) {
1109 while((idx
=findNextItem())>=0) {
1115 Package::removeItems(const Package
&listPkg
) {
1119 for(pItem
=listPkg
.items
, i
=0; i
<listPkg
.itemCount
; ++pItem
, ++i
) {
1120 removeItems(pItem
->name
);
1125 Package::extractItem(const char *filesPath
, const char *outName
, int32_t idx
, char outType
) {
1126 char filename
[1024];
1131 uint8_t itemCharset
, outCharset
;
1132 UBool itemIsBigEndian
, outIsBigEndian
;
1134 if(idx
<0 || itemCount
<=idx
) {
1139 // swap the data to the outType
1140 // outType==0: don't swap
1141 if(outType
!=0 && pItem
->type
!=outType
) {
1143 UErrorCode errorCode
=U_ZERO_ERROR
;
1144 makeTypeProps(pItem
->type
, itemCharset
, itemIsBigEndian
);
1145 makeTypeProps(outType
, outCharset
, outIsBigEndian
);
1146 ds
=udata_openSwapper(itemIsBigEndian
, itemCharset
, outIsBigEndian
, outCharset
, &errorCode
);
1147 if(U_FAILURE(errorCode
)) {
1148 fprintf(stderr
, "icupkg: udata_openSwapper(item %ld) failed - %s\n",
1149 (long)idx
, u_errorName(errorCode
));
1153 ds
->printError
=printPackageError
;
1154 ds
->printErrorContext
=stderr
;
1156 // swap the item from its platform properties to the desired ones
1157 udata_swap(ds
, pItem
->data
, pItem
->length
, pItem
->data
, &errorCode
);
1158 if(U_FAILURE(errorCode
)) {
1159 fprintf(stderr
, "icupkg: udata_swap(item %ld) failed - %s\n", (long)idx
, u_errorName(errorCode
));
1162 udata_closeSwapper(ds
);
1163 pItem
->type
=outType
;
1166 // create the file and write its contents
1167 makeFullFilenameAndDirs(filesPath
, outName
, filename
, (int32_t)sizeof(filename
));
1168 file
=fopen(filename
, "wb");
1170 fprintf(stderr
, "icupkg: unable to create file \"%s\"\n", filename
);
1171 exit(U_FILE_ACCESS_ERROR
);
1173 fileLength
=(int32_t)fwrite(pItem
->data
, 1, pItem
->length
, file
);
1175 if(ferror(file
) || fileLength
!=pItem
->length
) {
1176 fprintf(stderr
, "icupkg: unable to write complete file \"%s\"\n", filename
);
1177 exit(U_FILE_ACCESS_ERROR
);
1183 Package::extractItem(const char *filesPath
, int32_t idx
, char outType
) {
1184 extractItem(filesPath
, items
[idx
].name
, idx
, outType
);
1188 Package::extractItems(const char *filesPath
, const char *pattern
, char outType
) {
1192 while((idx
=findNextItem())>=0) {
1193 extractItem(filesPath
, idx
, outType
);
1198 Package::extractItems(const char *filesPath
, const Package
&listPkg
, char outType
) {
1202 for(pItem
=listPkg
.items
, i
=0; i
<listPkg
.itemCount
; ++pItem
, ++i
) {
1203 extractItems(filesPath
, pItem
->name
, outType
);
1208 Package::getItemCount() const {
1213 Package::getItem(int32_t idx
) const {
1214 if (0 <= idx
&& idx
< itemCount
) {
1221 Package::checkDependency(void *context
, const char *itemName
, const char *targetName
) {
1222 // check dependency: make sure the target item is in the package
1223 Package
*me
=(Package
*)context
;
1224 if(me
->findItem(targetName
)<0) {
1225 me
->isMissingItems
=TRUE
;
1226 fprintf(stderr
, "Item %s depends on missing item %s\n", itemName
, targetName
);
1231 Package::checkDependencies() {
1232 isMissingItems
=FALSE
;
1233 enumDependencies(this, checkDependency
);
1234 return (UBool
)!isMissingItems
;
1238 Package::enumDependencies(void *context
, CheckDependency check
) {
1241 for(i
=0; i
<itemCount
; ++i
) {
1242 enumDependencies(items
+i
, context
, check
);
1247 Package::allocString(UBool in
, int32_t length
) {
1260 if(top
>STRING_STORE_SIZE
) {
1261 fprintf(stderr
, "icupkg: string storage overflow\n");
1262 exit(U_BUFFER_OVERFLOW_ERROR
);
1273 Package::sortItems() {
1274 UErrorCode errorCode
=U_ZERO_ERROR
;
1275 uprv_sortArray(items
, itemCount
, (int32_t)sizeof(Item
), compareItems
, NULL
, FALSE
, &errorCode
);
1276 if(U_FAILURE(errorCode
)) {
1277 fprintf(stderr
, "icupkg: sorting item names failed - %s\n", u_errorName(errorCode
));
1282 void Package::setItemCapacity(int32_t max
)
1287 Item
*newItems
= (Item
*)uprv_malloc(max
* sizeof(items
[0]));
1288 Item
*oldItems
= items
;
1289 if(newItems
== NULL
) {
1290 fprintf(stderr
, "icupkg: Out of memory trying to allocate %lu bytes for %d items\n",
1291 (unsigned long)max
*sizeof(items
[0]), max
);
1292 exit(U_MEMORY_ALLOCATION_ERROR
);
1294 if(items
&& itemCount
>0) {
1295 uprv_memcpy(newItems
, items
, itemCount
*sizeof(items
[0]));
1299 uprv_free(oldItems
);
1302 void Package::ensureItemCapacity()
1304 if((itemCount
+1)>itemMax
) {
1305 setItemCapacity(itemCount
+kItemsChunk
);