/* Format of cnvalias.icu ----------------------------------------------------- * * cnvalias.icu is a binary, memory-mappable form of convrtrs.txt. * This binary form contains several tables. All indexes are to uint16_t * units, and not to the bytes (uint8_t units). Addressing everything on * 16-bit boundaries allows us to store more information with small index * numbers, which are also 16-bit in size. The majority of the table (except * the string table) are 16-bit numbers. * * First there is the size of the Table of Contents (TOC). The TOC * entries contain the size of each section. In order to find the offset * you just need to sum up the previous offsets. * The TOC length and entries are an array of uint32_t values. * The first section after the TOC starts immediately after the TOC. * * 1) This section contains a list of converters. This list contains indexes * into the string table for the converter name. The index of this list is * also used by other sections, which are mentioned later on. * This list is not sorted. * * 2) This section contains a list of tags. This list contains indexes * into the string table for the tag name. The index of this list is * also used by other sections, which are mentioned later on. * This list is in priority order of standards. * * 3) This section contains a list of sorted unique aliases. This * list contains indexes into the string table for the alias name. The * index of this list is also used by other sections, like the 4th section. * The index for the 3rd and 4th section is used to get the * alias -> converter name mapping. Section 3 and 4 form a two column table. * Some of the most significant bits of each index may contain other * information (see findConverter for details). * * 4) This section contains a list of mapped converter names. Consider this * as a table that maps the 3rd section to the 1st section. This list contains * indexes into the 1st section. The index of this list is the same index in * the 3rd section. There is also some extra information in the high bits of * each converter index in this table. Currently it's only used to say that * an alias mapped to this converter is ambiguous. See UCNV_CONVERTER_INDEX_MASK * and UCNV_AMBIGUOUS_ALIAS_MAP_BIT for more information. This section is * the predigested form of the 5th section so that an alias lookup can be fast. * * 5) This section contains a 2D array with indexes to the 6th section. This * section is the full form of all alias mappings. The column index is the * index into the converter list (column header). The row index is the index * to tag list (row header). This 2D array is the top part a 3D array. The * third dimension is in the 6th section. * * 6) This is blob of variable length arrays. Each array starts with a size, * and is followed by indexes to alias names in the string table. This is * the third dimension to the section 5. No other section should be referencing * this section. * * 7) Starting in ICU 3.6, this can be a UConverterAliasOptions struct. Its * presence indicates that a section 9 exists. UConverterAliasOptions specifies * what type of string normalization is used among other potential things in the * future. * * 8) This is the string table. All strings are indexed on an even address. * There are two reasons for this. First many chip architectures locate strings * faster on even address boundaries. Second, since all indexes are 16-bit * numbers, this string table can be 128KB in size instead of 64KB when we * only have strings starting on an even address. * * 9) When present this is a set of prenormalized strings from section 8. This * table contains normalized strings with the dashes and spaces stripped out, * and all strings lowercased. In the future, the options in section 7 may state * other types of normalization. * * Here is the concept of section 5 and 6. It's a 3D cube. Each tag * has a unique alias among all converters. That same alias can * be mentioned in other standards on different converters, * but only one alias per tag can be unique. * * * Converter Names (Usually in TR22 form) * -------------------------------------------. * T / /| * a / / | * g / / | * s / / | * / / | * ------------------------------------------/ | * A | | | * l | | | * i | | / * a | | / * s | | / * e | | / * s | |/ * ------------------------------------------- * * * * Here is what it really looks like. It's like swiss cheese. * There are holes. Some converters aren't recognized by * a standard, or they are really old converters that the * standard doesn't recognize anymore. * * Converter Names (Usually in TR22 form) * -------------------------------------------. * T /##########################################/| * a / # # /# * g / # ## ## ### # ### ### ### #/ * s / # ##### #### ## ## #/# * / ### # # ## # # # ### # # #/## * ------------------------------------------/# # * A |### # # ## # # # ### # # #|# # * l |# # # # # ## # #|# # * i |# # # # # # #|# * a |# #|# * s | #|# * e * s *
*/
/** * Used by the UEnumeration API
*/ typedefstruct UAliasContext {
uint32_t listOffset;
uint32_t listIdx;
} UAliasContext;
enum {
tocLengthIndex=0,
converterListIndex=1,
tagListIndex=2,
aliasListIndex=3,
untaggedConvArrayIndex=4,
taggedAliasArrayIndex=5,
taggedAliasListsIndex=6,
tableOptionsIndex=7,
stringTableIndex=8,
normalizedStringTableIndex=9,
offsetsCount, /* length of the swapper's temporary offsets[] */
minTocLength=8 /* min. tocLength in the file, does not count the tocLengthIndex! */
};
/* aliasLists is a 1's based array, but it has a padding character */
currOffset += gMainTable.taggedAliasArraySize;
gMainTable.taggedAliasLists = table + currOffset;
currOffset += gMainTable.taggedAliasListsSize; if (gMainTable.optionTableSize > 0
&& reinterpret_cast<const UConverterAliasOptions*>(table + currOffset)->stringNormalizationType < UCNV_IO_NORM_TYPE_COUNT)
{ /* Faster table */
gMainTable.optionTable = reinterpret_cast<const UConverterAliasOptions*>(table + currOffset);
} else { /* Smaller table, or I can't handle this normalization mode!
Use the original slower table lookup. */
gMainTable.optionTable = &defaultTableOptions;
}
static uint32_t getTagNumber(constchar *tagname) { if (gMainTable.tagList) {
uint32_t tagNum; for (tagNum = 0; tagNum < gMainTable.tagListSize; tagNum++) { if (!uprv_stricmp(GET_STRING(gMainTable.tagList[tagNum]), tagname)) { return tagNum;
}
}
}
return UINT32_MAX;
}
/* character types relevant for ucnv_compareNames() */ enum {
UIGNORE,
ZERO,
NONZERO,
MINLETTER /* any values from here on are lowercase letter mappings */
};
while ((c1 = *name++) != 0) {
type = GET_EBCDIC_TYPE(c1); switch (type) { case UIGNORE:
afterDigit = false; continue; /* ignore all but letters and digits */ case ZERO: if (!afterDigit) {
nextType = GET_EBCDIC_TYPE(*name); if (nextType == ZERO || nextType == NONZERO) { continue; /* ignore leading zero before another digit */
}
} break; case NONZERO:
afterDigit = true; break; default:
c1 = (char)type; /* lowercased letter */
afterDigit = false; break;
}
*dstItr++ = c1;
}
*dstItr = 0; return dst;
}
/** * Do a fuzzy compare of two converter/alias names. * The comparison is case-insensitive, ignores leading zeroes if they are not * followed by further digits, and ignores all but letters and digits. * Thus the strings "UTF-8", "utf_8", "u*T@f08" and "Utf 8" are exactly equivalent. * See section 1.4, Charset Alias Matching in Unicode Technical Standard #22 * at http://www.unicode.org/reports/tr22/ * * This is a symmetrical (commutative) operation; order of arguments * is insignificant. This is an important property for sorting the * list (when the list is preprocessed into binary form) and for * performing binary searches on it at run time. * * @param name1 a converter name or alias, zero-terminated * @param name2 a converter name or alias, zero-terminated * @return 0 if the names match, or a negative value if the name1 * lexically precedes name2, or a positive value if the name1 * lexically follows name2. * * @see ucnv_io_stripForCompare
*/
U_CAPI int U_EXPORT2
ucnv_compareNames(constchar *name1, constchar *name2) { int rc;
uint8_t type, nextType; char c1, c2;
UBool afterDigit1 = false, afterDigit2 = false;
for (;;) { while ((c1 = *name1++) != 0) {
type = GET_CHAR_TYPE(c1); switch (type) { case UIGNORE:
afterDigit1 = false; continue; /* ignore all but letters and digits */ case ZERO: if (!afterDigit1) {
nextType = GET_CHAR_TYPE(*name1); if (nextType == ZERO || nextType == NONZERO) { continue; /* ignore leading zero before another digit */
}
} break; case NONZERO:
afterDigit1 = true; break; default:
c1 = (char)type; /* lowercased letter */
afterDigit1 = false; break;
} break; /* deliver c1 */
} while ((c2 = *name2++) != 0) {
type = GET_CHAR_TYPE(c2); switch (type) { case UIGNORE:
afterDigit2 = false; continue; /* ignore all but letters and digits */ case ZERO: if (!afterDigit2) {
nextType = GET_CHAR_TYPE(*name2); if (nextType == ZERO || nextType == NONZERO) { continue; /* ignore leading zero before another digit */
}
} break; case NONZERO:
afterDigit2 = true; break; default:
c2 = (char)type; /* lowercased letter */
afterDigit2 = false; break;
} break; /* deliver c2 */
}
/* If we reach the ends of both strings then they match */ if ((c1|c2)==0) { return 0;
}
/* * search for an alias * return the converter number index for gConverterList
*/ staticinline uint32_t
findConverter(constchar *alias, UBool *containsOption, UErrorCode *pErrorCode) {
uint32_t mid, start, limit;
uint32_t lastMid; int result; int isUnnormalized = (gMainTable.optionTable->stringNormalizationType == UCNV_IO_UNNORMALIZED); char strippedName[UCNV_MAX_CONVERTER_NAME_LENGTH];
if (!isUnnormalized) { if (uprv_strlen(alias) >= UCNV_MAX_CONVERTER_NAME_LENGTH) {
*pErrorCode = U_BUFFER_OVERFLOW_ERROR; return UINT32_MAX;
}
/* Lower case and remove ignoreable characters. */
ucnv_io_stripForCompare(strippedName, alias);
alias = strippedName;
}
/* do a binary search for the alias */
start = 0;
limit = gMainTable.untaggedConvArraySize;
mid = limit;
lastMid = UINT32_MAX;
for (;;) {
mid = (start + limit) / 2; if (lastMid == mid) { /* Have we moved? */ break; /* We haven't moved, and it wasn't found. */
}
lastMid = mid; if (isUnnormalized) {
result = ucnv_compareNames(alias, GET_STRING(gMainTable.aliasList[mid]));
} else {
result = uprv_strcmp(alias, GET_NORMALIZED_STRING(gMainTable.aliasList[mid]));
}
if (result < 0) {
limit = mid;
} elseif (result > 0) {
start = mid;
} else { /* Since the gencnval tool folds duplicates into one entry, * this alias in gAliasList is unique, but different standards * may map an alias to different converters.
*/ if (gMainTable.untaggedConvArray[mid] & UCNV_AMBIGUOUS_ALIAS_MAP_BIT) {
*pErrorCode = U_AMBIGUOUS_ALIAS_WARNING;
} /* State whether the canonical converter name contains an option.
This information is contained in this list in order to maintain backward & forward compatibility. */ if (containsOption) {
UBool containsCnvOptionInfo = static_cast<UBool>(gMainTable.optionTable->containsCnvOptionInfo);
*containsOption = static_cast<UBool>((containsCnvOptionInfo
&& ((gMainTable.untaggedConvArray[mid] & UCNV_CONTAINS_OPTION_BIT) != 0))
|| !containsCnvOptionInfo);
} return gMainTable.untaggedConvArray[mid] & UCNV_CONVERTER_INDEX_MASK;
}
}
return UINT32_MAX;
}
/* * Is this alias in this list? * alias and listOffset should be non-nullptr.
*/ staticinline UBool
isAliasInList(constchar *alias, uint32_t listOffset) { if (listOffset) {
uint32_t currAlias;
uint32_t listCount = gMainTable.taggedAliasLists[listOffset]; /* +1 to skip listCount */ const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1; for (currAlias = 0; currAlias < listCount; currAlias++) { if (currList[currAlias]
&& ucnv_compareNames(alias, GET_STRING(currList[currAlias]))==0)
{ returntrue;
}
}
} returnfalse;
}
/* * Search for an standard name of an alias (what is the default name * that this standard uses?) * return the listOffset for gTaggedAliasLists. If it's 0, * the it couldn't be found, but the parameters are valid.
*/ static uint32_t
findTaggedAliasListsOffset(constchar *alias, constchar *standard, UErrorCode *pErrorCode) {
uint32_t idx;
uint32_t listOffset;
uint32_t convNum;
UErrorCode myErr = U_ZERO_ERROR;
uint32_t tagNum = getTagNumber(standard);
/* Make a quick guess. Hopefully they used a TR22 canonical alias. */
convNum = findConverter(alias, nullptr, &myErr); if (myErr != U_ZERO_ERROR) {
*pErrorCode = myErr;
}
if (tagNum < (gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) && convNum < gMainTable.converterListSize) {
listOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + convNum]; if (listOffset && gMainTable.taggedAliasLists[listOffset + 1]) { return listOffset;
} if (myErr == U_AMBIGUOUS_ALIAS_WARNING) { /* Uh Oh! They used an ambiguous alias. We have to search the whole swiss cheese starting at the highest standard affinity. This may take a while.
*/ for (idx = 0; idx < gMainTable.taggedAliasArraySize; idx++) {
listOffset = gMainTable.taggedAliasArray[idx]; if (listOffset && isAliasInList(alias, listOffset)) {
uint32_t currTagNum = idx/gMainTable.converterListSize;
uint32_t currConvNum = (idx - currTagNum*gMainTable.converterListSize);
uint32_t tempListOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + currConvNum]; if (tempListOffset && gMainTable.taggedAliasLists[tempListOffset + 1]) { return tempListOffset;
} /* else keep on looking */ /* We could speed this up by starting on the next row because an alias is unique per row, right now.
This would change if alias versioning appears. */
}
} /* The standard doesn't know about the alias */
} /* else no default name */ return 0;
} /* else converter or tag not found */
/* Make a quick guess. Hopefully they used a TR22 canonical alias. */
convNum = findConverter(alias, nullptr, &myErr); if (myErr != U_ZERO_ERROR) {
*pErrorCode = myErr;
}
if (tagNum < (gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) && convNum < gMainTable.converterListSize) {
listOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + convNum]; if (listOffset && isAliasInList(alias, listOffset)) { return convNum;
} if (myErr == U_AMBIGUOUS_ALIAS_WARNING) { /* Uh Oh! They used an ambiguous alias. We have to search one slice of the swiss cheese. We search only in the requested tag, not the whole thing. This may take a while.
*/
uint32_t convStart = (tagNum)*gMainTable.converterListSize;
uint32_t convLimit = (tagNum+1)*gMainTable.converterListSize; for (idx = convStart; idx < convLimit; idx++) {
listOffset = gMainTable.taggedAliasArray[idx]; if (listOffset && isAliasInList(alias, listOffset)) { return idx-convStart;
}
} /* The standard doesn't know about the alias */
} /* else no canonical name */
} /* else converter or tag not found */
return UINT32_MAX;
}
U_CAPI constchar *
ucnv_io_getConverterName(constchar *alias, UBool *containsOption, UErrorCode *pErrorCode) { constchar *aliasTmp = alias;
int32_t i = 0; for (i = 0; i < 2; i++) { if (i == 1) { /* * After the first unsuccess converter lookup, check to see if * the name begins with 'x-'. If it does, strip it off and try * again. This behaviour is similar to how ICU4J does it.
*/ if (aliasTmp[0] == 'x' && aliasTmp[1] == '-') {
aliasTmp = aliasTmp+2;
} else { break;
}
} if(haveAliasData(pErrorCode) && isAlias(aliasTmp, pErrorCode)) {
uint32_t convNum = findConverter(aliasTmp, containsOption, pErrorCode); if (convNum < gMainTable.converterListSize) { return GET_STRING(gMainTable.converterList[convNum]);
} /* else converter not found */
} else { break;
}
}
if (myContext->listIdx < listCount) { constchar *myStr = GET_STRING(currList[myContext->listIdx++]); if (resultLength) {
*resultLength = (int32_t)uprv_strlen(myStr);
} return myStr;
}
} /* Either we accessed a zero length list, or we enumerated too far. */ if (resultLength) {
*resultLength = 0;
} return nullptr;
}
/* Enumerate the aliases for the specified converter and standard tag */ staticconst UEnumeration gEnumAliases = {
nullptr,
nullptr,
ucnv_io_closeUEnumeration,
ucnv_io_countStandardAliases,
uenum_unextDefault,
ucnv_io_nextStandardAliases,
ucnv_io_resetStandardAliases
};
/* When listOffset == 0, we want to acknowledge that the converter name and standard are okay, but there
is nothing to enumerate. */ if (listOffset < gMainTable.taggedAliasListsSize) {
UAliasContext *myContext;
myEnum = static_cast<UEnumeration *>(uprv_malloc(sizeof(UEnumeration))); if (myEnum == nullptr) {
*pErrorCode = U_MEMORY_ALLOCATION_ERROR; return nullptr;
}
uprv_memcpy(myEnum, &gEnumAliases, sizeof(UEnumeration));
myContext = static_cast<UAliasContext *>(uprv_malloc(sizeof(UAliasContext))); if (myContext == nullptr) {
*pErrorCode = U_MEMORY_ALLOCATION_ERROR;
uprv_free(myEnum); return nullptr;
}
myContext->listOffset = listOffset;
myContext->listIdx = 0;
myEnum->context = myContext;
} /* else converter or tag not found */
} return myEnum;
}
static uint16_t
ucnv_io_countAliases(constchar *alias, UErrorCode *pErrorCode) { if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) {
uint32_t convNum = findConverter(alias, nullptr, pErrorCode); if (convNum < gMainTable.converterListSize) { /* tagListNum - 1 is the ALL tag */
int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum];
if (listOffset) { return gMainTable.taggedAliasLists[listOffset];
} /* else this shouldn't happen. internal program error */
} /* else converter not found */
} return 0;
}
static uint16_t
ucnv_io_getAliases(constchar *alias, uint16_t start, constchar **aliases, UErrorCode *pErrorCode) { if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) {
uint32_t currAlias;
uint32_t convNum = findConverter(alias, nullptr, pErrorCode); if (convNum < gMainTable.converterListSize) { /* tagListNum - 1 is the ALL tag */
int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum];
if (n < listCount) { return GET_STRING(currList[n]);
}
*pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR;
} /* else this shouldn't happen. internal program error */
} /* else converter not found */
} return nullptr;
}
static uint16_t
ucnv_io_countStandards(UErrorCode *pErrorCode) { if (haveAliasData(pErrorCode)) { /* Don't include the empty list */ returnstatic_cast<uint16_t>(gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS);
}
return 0;
}
U_CAPI constchar * U_EXPORT2
ucnv_getStandard(uint16_t n, UErrorCode *pErrorCode) { if (haveAliasData(pErrorCode)) { if (n < gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) { return GET_STRING(gMainTable.tagList[n]);
}
*pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR;
}
/* Get the preferred name from this list */ if (currList[0]) { return GET_STRING(currList[0]);
} /* else someone screwed up the alias table. */ /* *pErrorCode = U_INVALID_FORMAT_ERROR */
}
}
if (*myContext < gMainTable.converterListSize) { constchar *myStr = GET_STRING(gMainTable.converterList[(*myContext)++]); if (resultLength) {
*resultLength = (int32_t)uprv_strlen(myStr);
} return myStr;
} /* Either we accessed a zero length list, or we enumerated too far. */ if (resultLength) {
*resultLength = 0;
} return nullptr;
}
/* * row of a temporary array * * gets platform-endian charset string indexes and sorting indexes; * after sorting this array by strings, the actual arrays are permutated * according to the sorting indexes
*/ typedefstruct TempRow {
uint16_t strIndex, sortIndex;
} TempRow;
/* check data format and format version */
pInfo=(const UDataInfo *)((constchar *)inData+4); if(!(
pInfo->dataFormat[0]==0x43 && /* dataFormat="CvAl" */
pInfo->dataFormat[1]==0x76 &&
pInfo->dataFormat[2]==0x41 &&
pInfo->dataFormat[3]==0x6c &&
pInfo->formatVersion[0]==3
)) {
udata_printError(ds, "ucnv_swapAliases(): data format %02x.%02x.%02x.%02x (format version %02x) is not an alias table\n",
pInfo->dataFormat[0], pInfo->dataFormat[1],
pInfo->dataFormat[2], pInfo->dataFormat[3],
pInfo->formatVersion[0]);
*pErrorCode=U_UNSUPPORTED_ERROR; return 0;
}
/* an alias table must contain at least the table of contents array */ if(length>=0 && (length-headerSize)<4*(1+minTocLength)) {
udata_printError(ds, "ucnv_swapAliases(): too few bytes (%d after header) for an alias table\n",
length-headerSize);
*pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; return 0;
}
if((length-headerSize)<(2*(int32_t)topOffset)) {
udata_printError(ds, "ucnv_swapAliases(): too few bytes (%d after header) for an alias table\n",
length-headerSize);
*pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; return 0;
}
if(ds->inCharset==ds->outCharset) { /* no need to sort, just swap all 16-bit values together */
ds->swapArray16(ds,
inTable+offsets[converterListIndex],
2*(int32_t)(offsets[stringTableIndex]-offsets[converterListIndex]),
outTable+offsets[converterListIndex],
pErrorCode);
} else { /* allocate the temporary table for sorting */
count=toc[aliasListIndex];
tempTable.chars=(constchar *)(outTable+offsets[stringTableIndex]); /* sort by outCharset */
/* * Sort unique aliases+mapped names. * * We need to sort the list again by outCharset strings because they * sort differently for different charset families. * First we set up a temporary table with the string indexes and * sorting indexes and sort that. * Then we permutate and copy/swap the actual values.
*/
p=inTable+offsets[aliasListIndex];
q=outTable+offsets[aliasListIndex];
if(U_SUCCESS(*pErrorCode)) { /* copy/swap/permutate items */ if(p!=q) { for(i=0; i<count; ++i) {
oldIndex=tempTable.rows[i].sortIndex;
ds->swapArray16(ds, p+oldIndex, 2, q+i, pErrorCode);
ds->swapArray16(ds, p2+oldIndex, 2, q2+i, pErrorCode);
}
} else { /* * If we swap in-place, then the permutation must use another * temporary array (tempTable.resort) * before the results are copied to the outBundle.
*/
uint16_t *r=tempTable.resort;
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.