michael@0: /*
michael@0: ******************************************************************************
michael@0: *
michael@0: *   Copyright (C) 1999-2013, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *
michael@0: ******************************************************************************
michael@0: *
michael@0: *
michael@0: *  ucnv_io.cpp:
michael@0: *  initializes global variables and defines functions pertaining to converter 
michael@0: *  name resolution aspect of the conversion code.
michael@0: *
michael@0: *   new implementation:
michael@0: *
michael@0: *   created on: 1999nov22
michael@0: *   created by: Markus W. Scherer
michael@0: *
michael@0: *   Use the binary cnvalias.icu (created from convrtrs.txt) to work
michael@0: *   with aliases for converter names.
michael@0: *
michael@0: *   Date        Name        Description
michael@0: *   11/22/1999  markus      Created
michael@0: *   06/28/2002  grhoten     Major overhaul of the converter alias design.
michael@0: *                           Now an alias can map to different converters
michael@0: *                           depending on the specified standard.
michael@0: *******************************************************************************
michael@0: */
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: 
michael@0: #if !UCONFIG_NO_CONVERSION
michael@0: 
michael@0: #include "unicode/ucnv.h"
michael@0: #include "unicode/udata.h"
michael@0: 
michael@0: #include "umutex.h"
michael@0: #include "uarrsort.h"
michael@0: #include "uassert.h"
michael@0: #include "udataswp.h"
michael@0: #include "cstring.h"
michael@0: #include "cmemory.h"
michael@0: #include "ucnv_io.h"
michael@0: #include "uenumimp.h"
michael@0: #include "ucln_cmn.h"
michael@0: 
michael@0: /* Format of cnvalias.icu -----------------------------------------------------
michael@0:  *
michael@0:  * cnvalias.icu is a binary, memory-mappable form of convrtrs.txt.
michael@0:  * This binary form contains several tables. All indexes are to uint16_t
michael@0:  * units, and not to the bytes (uint8_t units). Addressing everything on
michael@0:  * 16-bit boundaries allows us to store more information with small index
michael@0:  * numbers, which are also 16-bit in size. The majority of the table (except
michael@0:  * the string table) are 16-bit numbers.
michael@0:  *
michael@0:  * First there is the size of the Table of Contents (TOC). The TOC
michael@0:  * entries contain the size of each section. In order to find the offset
michael@0:  * you just need to sum up the previous offsets.
michael@0:  * The TOC length and entries are an array of uint32_t values.
michael@0:  * The first section after the TOC starts immediately after the TOC.
michael@0:  *
michael@0:  * 1) This section contains a list of converters. This list contains indexes
michael@0:  * into the string table for the converter name. The index of this list is
michael@0:  * also used by other sections, which are mentioned later on.
michael@0:  * This list is not sorted.
michael@0:  *
michael@0:  * 2) This section contains a list of tags. This list contains indexes
michael@0:  * into the string table for the tag name. The index of this list is
michael@0:  * also used by other sections, which are mentioned later on.
michael@0:  * This list is in priority order of standards.
michael@0:  *
michael@0:  * 3) This section contains a list of sorted unique aliases. This
michael@0:  * list contains indexes into the string table for the alias name. The
michael@0:  * index of this list is also used by other sections, like the 4th section.
michael@0:  * The index for the 3rd and 4th section is used to get the
michael@0:  * alias -> converter name mapping. Section 3 and 4 form a two column table.
michael@0:  * Some of the most significant bits of each index may contain other
michael@0:  * information (see findConverter for details).
michael@0:  *
michael@0:  * 4) This section contains a list of mapped converter names. Consider this
michael@0:  * as a table that maps the 3rd section to the 1st section. This list contains
michael@0:  * indexes into the 1st section. The index of this list is the same index in
michael@0:  * the 3rd section. There is also some extra information in the high bits of
michael@0:  * each converter index in this table. Currently it's only used to say that
michael@0:  * an alias mapped to this converter is ambiguous. See UCNV_CONVERTER_INDEX_MASK
michael@0:  * and UCNV_AMBIGUOUS_ALIAS_MAP_BIT for more information. This section is
michael@0:  * the predigested form of the 5th section so that an alias lookup can be fast.
michael@0:  *
michael@0:  * 5) This section contains a 2D array with indexes to the 6th section. This
michael@0:  * section is the full form of all alias mappings. The column index is the
michael@0:  * index into the converter list (column header). The row index is the index
michael@0:  * to tag list (row header). This 2D array is the top part a 3D array. The
michael@0:  * third dimension is in the 6th section.
michael@0:  *
michael@0:  * 6) This is blob of variable length arrays. Each array starts with a size,
michael@0:  * and is followed by indexes to alias names in the string table. This is
michael@0:  * the third dimension to the section 5. No other section should be referencing
michael@0:  * this section.
michael@0:  *
michael@0:  * 7) Starting in ICU 3.6, this can be a UConverterAliasOptions struct. Its
michael@0:  * presence indicates that a section 9 exists. UConverterAliasOptions specifies
michael@0:  * what type of string normalization is used among other potential things in the
michael@0:  * future.
michael@0:  *
michael@0:  * 8) This is the string table. All strings are indexed on an even address.
michael@0:  * There are two reasons for this. First many chip architectures locate strings
michael@0:  * faster on even address boundaries. Second, since all indexes are 16-bit
michael@0:  * numbers, this string table can be 128KB in size instead of 64KB when we
michael@0:  * only have strings starting on an even address.
michael@0:  *
michael@0:  * 9) When present this is a set of prenormalized strings from section 8. This
michael@0:  * table contains normalized strings with the dashes and spaces stripped out,
michael@0:  * and all strings lowercased. In the future, the options in section 7 may state
michael@0:  * other types of normalization.
michael@0:  *
michael@0:  * Here is the concept of section 5 and 6. It's a 3D cube. Each tag
michael@0:  * has a unique alias among all converters. That same alias can
michael@0:  * be mentioned in other standards on different converters,
michael@0:  * but only one alias per tag can be unique.
michael@0:  *
michael@0:  *
michael@0:  *              Converter Names (Usually in TR22 form)
michael@0:  *           -------------------------------------------.
michael@0:  *     T    /                                          /|
michael@0:  *     a   /                                          / |
michael@0:  *     g  /                                          /  |
michael@0:  *     s /                                          /   |
michael@0:  *      /                                          /    |
michael@0:  *      ------------------------------------------/     |
michael@0:  *    A |                                         |     |
michael@0:  *    l |                                         |     |
michael@0:  *    i |                                         |    /
michael@0:  *    a |                                         |   /
michael@0:  *    s |                                         |  /
michael@0:  *    e |                                         | /
michael@0:  *    s |                                         |/
michael@0:  *      -------------------------------------------
michael@0:  *
michael@0:  *
michael@0:  *
michael@0:  * Here is what it really looks like. It's like swiss cheese.
michael@0:  * There are holes. Some converters aren't recognized by
michael@0:  * a standard, or they are really old converters that the
michael@0:  * standard doesn't recognize anymore.
michael@0:  *
michael@0:  *              Converter Names (Usually in TR22 form)
michael@0:  *           -------------------------------------------.
michael@0:  *     T    /##########################################/|
michael@0:  *     a   /     #            #                       /#
michael@0:  *     g  /  #      ##     ##     ### # ### ### ### #/
michael@0:  *     s / #             #####  ####        ##  ## #/#
michael@0:  *      / ### # # ##  #  #   #          ### # #   #/##
michael@0:  *      ------------------------------------------/# #
michael@0:  *    A |### # # ##  #  #   #          ### # #   #|# #
michael@0:  *    l |# # #    #     #               ## #     #|# #
michael@0:  *    i |# # #    #     #                #       #|#
michael@0:  *    a |#                                       #|#
michael@0:  *    s |                                        #|#
michael@0:  *    e
michael@0:  *    s
michael@0:  *
michael@0:  */
michael@0: 
michael@0: /**
michael@0:  * Used by the UEnumeration API
michael@0:  */
michael@0: typedef struct UAliasContext {
michael@0:     uint32_t listOffset;
michael@0:     uint32_t listIdx;
michael@0: } UAliasContext;
michael@0: 
michael@0: static const char DATA_NAME[] = "cnvalias";
michael@0: static const char DATA_TYPE[] = "icu";
michael@0: 
michael@0: static UDataMemory *gAliasData=NULL;
michael@0: static icu::UInitOnce gAliasDataInitOnce = U_INITONCE_INITIALIZER;
michael@0: 
michael@0: enum {
michael@0:     tocLengthIndex=0,
michael@0:     converterListIndex=1,
michael@0:     tagListIndex=2,
michael@0:     aliasListIndex=3,
michael@0:     untaggedConvArrayIndex=4,
michael@0:     taggedAliasArrayIndex=5,
michael@0:     taggedAliasListsIndex=6,
michael@0:     tableOptionsIndex=7,
michael@0:     stringTableIndex=8,
michael@0:     normalizedStringTableIndex=9,
michael@0:     offsetsCount,    /* length of the swapper's temporary offsets[] */
michael@0:     minTocLength=8 /* min. tocLength in the file, does not count the tocLengthIndex! */
michael@0: };
michael@0: 
michael@0: static const UConverterAliasOptions defaultTableOptions = {
michael@0:     UCNV_IO_UNNORMALIZED,
michael@0:     0 /* containsCnvOptionInfo */
michael@0: };
michael@0: static UConverterAlias gMainTable;
michael@0: 
michael@0: #define GET_STRING(idx) (const char *)(gMainTable.stringTable + (idx))
michael@0: #define GET_NORMALIZED_STRING(idx) (const char *)(gMainTable.normalizedStringTable + (idx))
michael@0: 
michael@0: static UBool U_CALLCONV
michael@0: isAcceptable(void * /*context*/,
michael@0:              const char * /*type*/, const char * /*name*/,
michael@0:              const UDataInfo *pInfo) {
michael@0:     return (UBool)(
michael@0:         pInfo->size>=20 &&
michael@0:         pInfo->isBigEndian==U_IS_BIG_ENDIAN &&
michael@0:         pInfo->charsetFamily==U_CHARSET_FAMILY &&
michael@0:         pInfo->dataFormat[0]==0x43 &&   /* dataFormat="CvAl" */
michael@0:         pInfo->dataFormat[1]==0x76 &&
michael@0:         pInfo->dataFormat[2]==0x41 &&
michael@0:         pInfo->dataFormat[3]==0x6c &&
michael@0:         pInfo->formatVersion[0]==3);
michael@0: }
michael@0: 
michael@0: static UBool U_CALLCONV ucnv_io_cleanup(void)
michael@0: {
michael@0:     if (gAliasData) {
michael@0:         udata_close(gAliasData);
michael@0:         gAliasData = NULL;
michael@0:     }
michael@0:     gAliasDataInitOnce.reset();
michael@0: 
michael@0:     uprv_memset(&gMainTable, 0, sizeof(gMainTable));
michael@0: 
michael@0:     return TRUE;                   /* Everything was cleaned up */
michael@0: }
michael@0: 
michael@0: static void U_CALLCONV initAliasData(UErrorCode &errCode) {
michael@0:     UDataMemory *data;
michael@0:     const uint16_t *table;
michael@0:     const uint32_t *sectionSizes;
michael@0:     uint32_t tableStart;
michael@0:     uint32_t currOffset;
michael@0: 
michael@0:     ucln_common_registerCleanup(UCLN_COMMON_UCNV_IO, ucnv_io_cleanup);
michael@0: 
michael@0:     U_ASSERT(gAliasData == NULL);
michael@0:     data = udata_openChoice(NULL, DATA_TYPE, DATA_NAME, isAcceptable, NULL, &errCode);
michael@0:     if(U_FAILURE(errCode)) {
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     sectionSizes = (const uint32_t *)udata_getMemory(data);
michael@0:     table = (const uint16_t *)sectionSizes;
michael@0: 
michael@0:     tableStart      = sectionSizes[0];
michael@0:     if (tableStart < minTocLength) {
michael@0:         errCode = U_INVALID_FORMAT_ERROR;
michael@0:         udata_close(data);
michael@0:         return;
michael@0:     }
michael@0:     gAliasData = data;
michael@0: 
michael@0:     gMainTable.converterListSize      = sectionSizes[1];
michael@0:     gMainTable.tagListSize            = sectionSizes[2];
michael@0:     gMainTable.aliasListSize          = sectionSizes[3];
michael@0:     gMainTable.untaggedConvArraySize  = sectionSizes[4];
michael@0:     gMainTable.taggedAliasArraySize   = sectionSizes[5];
michael@0:     gMainTable.taggedAliasListsSize   = sectionSizes[6];
michael@0:     gMainTable.optionTableSize        = sectionSizes[7];
michael@0:     gMainTable.stringTableSize        = sectionSizes[8];
michael@0: 
michael@0:     if (tableStart > 8) {
michael@0:         gMainTable.normalizedStringTableSize = sectionSizes[9];
michael@0:     }
michael@0: 
michael@0:     currOffset = tableStart * (sizeof(uint32_t)/sizeof(uint16_t)) + (sizeof(uint32_t)/sizeof(uint16_t));
michael@0:     gMainTable.converterList = table + currOffset;
michael@0: 
michael@0:     currOffset += gMainTable.converterListSize;
michael@0:     gMainTable.tagList = table + currOffset;
michael@0: 
michael@0:     currOffset += gMainTable.tagListSize;
michael@0:     gMainTable.aliasList = table + currOffset;
michael@0: 
michael@0:     currOffset += gMainTable.aliasListSize;
michael@0:     gMainTable.untaggedConvArray = table + currOffset;
michael@0: 
michael@0:     currOffset += gMainTable.untaggedConvArraySize;
michael@0:     gMainTable.taggedAliasArray = table + currOffset;
michael@0: 
michael@0:     /* aliasLists is a 1's based array, but it has a padding character */
michael@0:     currOffset += gMainTable.taggedAliasArraySize;
michael@0:     gMainTable.taggedAliasLists = table + currOffset;
michael@0: 
michael@0:     currOffset += gMainTable.taggedAliasListsSize;
michael@0:     if (gMainTable.optionTableSize > 0
michael@0:         && ((const UConverterAliasOptions *)(table + currOffset))->stringNormalizationType < UCNV_IO_NORM_TYPE_COUNT)
michael@0:     {
michael@0:         /* Faster table */
michael@0:         gMainTable.optionTable = (const UConverterAliasOptions *)(table + currOffset);
michael@0:     }
michael@0:     else {
michael@0:         /* Smaller table, or I can't handle this normalization mode!
michael@0:         Use the original slower table lookup. */
michael@0:         gMainTable.optionTable = &defaultTableOptions;
michael@0:     }
michael@0: 
michael@0:     currOffset += gMainTable.optionTableSize;
michael@0:     gMainTable.stringTable = table + currOffset;
michael@0: 
michael@0:     currOffset += gMainTable.stringTableSize;
michael@0:     gMainTable.normalizedStringTable = ((gMainTable.optionTable->stringNormalizationType == UCNV_IO_UNNORMALIZED)
michael@0:         ? gMainTable.stringTable : (table + currOffset));
michael@0: }
michael@0: 
michael@0: 
michael@0: static UBool
michael@0: haveAliasData(UErrorCode *pErrorCode) {
michael@0:     umtx_initOnce(gAliasDataInitOnce, &initAliasData, *pErrorCode);
michael@0:     return U_SUCCESS(*pErrorCode);
michael@0: }
michael@0: 
michael@0: static inline UBool
michael@0: isAlias(const char *alias, UErrorCode *pErrorCode) {
michael@0:     if(alias==NULL) {
michael@0:         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return FALSE;
michael@0:     }
michael@0:     return (UBool)(*alias!=0);
michael@0: }
michael@0: 
michael@0: static uint32_t getTagNumber(const char *tagname) {
michael@0:     if (gMainTable.tagList) {
michael@0:         uint32_t tagNum;
michael@0:         for (tagNum = 0; tagNum < gMainTable.tagListSize; tagNum++) {
michael@0:             if (!uprv_stricmp(GET_STRING(gMainTable.tagList[tagNum]), tagname)) {
michael@0:                 return tagNum;
michael@0:             }
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     return UINT32_MAX;
michael@0: }
michael@0: 
michael@0: /* character types relevant for ucnv_compareNames() */
michael@0: enum {
michael@0:     UIGNORE,
michael@0:     ZERO,
michael@0:     NONZERO,
michael@0:     MINLETTER /* any values from here on are lowercase letter mappings */
michael@0: };
michael@0: 
michael@0: /* character types for ASCII 00..7F */
michael@0: static const uint8_t asciiTypes[128] = {
michael@0:     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
michael@0:     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
michael@0:     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
michael@0:     ZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, 0, 0, 0, 0, 0, 0,
michael@0:     0, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f,
michael@0:     0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0, 0, 0, 0, 0,
michael@0:     0, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f,
michael@0:     0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0, 0, 0, 0, 0
michael@0: };
michael@0: 
michael@0: #define GET_ASCII_TYPE(c) ((int8_t)(c) >= 0 ? asciiTypes[(uint8_t)c] : (uint8_t)UIGNORE)
michael@0: 
michael@0: /* character types for EBCDIC 80..FF */
michael@0: static const uint8_t ebcdicTypes[128] = {
michael@0:     0,    0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0, 0, 0, 0, 0, 0,
michael@0:     0,    0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0, 0, 0, 0, 0, 0,
michael@0:     0,    0,    0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0, 0, 0, 0, 0, 0,
michael@0:     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
michael@0:     0,    0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0, 0, 0, 0, 0, 0,
michael@0:     0,    0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0, 0, 0, 0, 0, 0,
michael@0:     0,    0,    0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0, 0, 0, 0, 0, 0,
michael@0:     ZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, NONZERO, 0, 0, 0, 0, 0, 0
michael@0: };
michael@0: 
michael@0: #define GET_EBCDIC_TYPE(c) ((int8_t)(c) < 0 ? ebcdicTypes[(c)&0x7f] : (uint8_t)UIGNORE)
michael@0: 
michael@0: #if U_CHARSET_FAMILY==U_ASCII_FAMILY
michael@0: #   define GET_CHAR_TYPE(c) GET_ASCII_TYPE(c)
michael@0: #elif U_CHARSET_FAMILY==U_EBCDIC_FAMILY
michael@0: #   define GET_CHAR_TYPE(c) GET_EBCDIC_TYPE(c)
michael@0: #else
michael@0: #   error U_CHARSET_FAMILY is not valid
michael@0: #endif
michael@0: 
michael@0: /* @see ucnv_compareNames */
michael@0: U_CFUNC char * U_EXPORT2
michael@0: ucnv_io_stripASCIIForCompare(char *dst, const char *name) {
michael@0:     char *dstItr = dst;
michael@0:     uint8_t type, nextType;
michael@0:     char c1;
michael@0:     UBool afterDigit = FALSE;
michael@0: 
michael@0:     while ((c1 = *name++) != 0) {
michael@0:         type = GET_ASCII_TYPE(c1);
michael@0:         switch (type) {
michael@0:         case UIGNORE:
michael@0:             afterDigit = FALSE;
michael@0:             continue; /* ignore all but letters and digits */
michael@0:         case ZERO:
michael@0:             if (!afterDigit) {
michael@0:                 nextType = GET_ASCII_TYPE(*name);
michael@0:                 if (nextType == ZERO || nextType == NONZERO) {
michael@0:                     continue; /* ignore leading zero before another digit */
michael@0:                 }
michael@0:             }
michael@0:             break;
michael@0:         case NONZERO:
michael@0:             afterDigit = TRUE;
michael@0:             break;
michael@0:         default:
michael@0:             c1 = (char)type; /* lowercased letter */
michael@0:             afterDigit = FALSE;
michael@0:             break;
michael@0:         }
michael@0:         *dstItr++ = c1;
michael@0:     }
michael@0:     *dstItr = 0;
michael@0:     return dst;
michael@0: }
michael@0: 
michael@0: U_CFUNC char * U_EXPORT2
michael@0: ucnv_io_stripEBCDICForCompare(char *dst, const char *name) {
michael@0:     char *dstItr = dst;
michael@0:     uint8_t type, nextType;
michael@0:     char c1;
michael@0:     UBool afterDigit = FALSE;
michael@0: 
michael@0:     while ((c1 = *name++) != 0) {
michael@0:         type = GET_EBCDIC_TYPE(c1);
michael@0:         switch (type) {
michael@0:         case UIGNORE:
michael@0:             afterDigit = FALSE;
michael@0:             continue; /* ignore all but letters and digits */
michael@0:         case ZERO:
michael@0:             if (!afterDigit) {
michael@0:                 nextType = GET_EBCDIC_TYPE(*name);
michael@0:                 if (nextType == ZERO || nextType == NONZERO) {
michael@0:                     continue; /* ignore leading zero before another digit */
michael@0:                 }
michael@0:             }
michael@0:             break;
michael@0:         case NONZERO:
michael@0:             afterDigit = TRUE;
michael@0:             break;
michael@0:         default:
michael@0:             c1 = (char)type; /* lowercased letter */
michael@0:             afterDigit = FALSE;
michael@0:             break;
michael@0:         }
michael@0:         *dstItr++ = c1;
michael@0:     }
michael@0:     *dstItr = 0;
michael@0:     return dst;
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Do a fuzzy compare of two converter/alias names.
michael@0:  * The comparison is case-insensitive, ignores leading zeroes if they are not
michael@0:  * followed by further digits, and ignores all but letters and digits.
michael@0:  * Thus the strings "UTF-8", "utf_8", "u*T@f08" and "Utf 8" are exactly equivalent.
michael@0:  * See section 1.4, Charset Alias Matching in Unicode Technical Standard #22
michael@0:  * at http://www.unicode.org/reports/tr22/
michael@0:  *
michael@0:  * This is a symmetrical (commutative) operation; order of arguments
michael@0:  * is insignificant.  This is an important property for sorting the
michael@0:  * list (when the list is preprocessed into binary form) and for
michael@0:  * performing binary searches on it at run time.
michael@0:  *
michael@0:  * @param name1 a converter name or alias, zero-terminated
michael@0:  * @param name2 a converter name or alias, zero-terminated
michael@0:  * @return 0 if the names match, or a negative value if the name1
michael@0:  * lexically precedes name2, or a positive value if the name1
michael@0:  * lexically follows name2.
michael@0:  *
michael@0:  * @see ucnv_io_stripForCompare
michael@0:  */
michael@0: U_CAPI int U_EXPORT2
michael@0: ucnv_compareNames(const char *name1, const char *name2) {
michael@0:     int rc;
michael@0:     uint8_t type, nextType;
michael@0:     char c1, c2;
michael@0:     UBool afterDigit1 = FALSE, afterDigit2 = FALSE;
michael@0: 
michael@0:     for (;;) {
michael@0:         while ((c1 = *name1++) != 0) {
michael@0:             type = GET_CHAR_TYPE(c1);
michael@0:             switch (type) {
michael@0:             case UIGNORE:
michael@0:                 afterDigit1 = FALSE;
michael@0:                 continue; /* ignore all but letters and digits */
michael@0:             case ZERO:
michael@0:                 if (!afterDigit1) {
michael@0:                     nextType = GET_CHAR_TYPE(*name1);
michael@0:                     if (nextType == ZERO || nextType == NONZERO) {
michael@0:                         continue; /* ignore leading zero before another digit */
michael@0:                     }
michael@0:                 }
michael@0:                 break;
michael@0:             case NONZERO:
michael@0:                 afterDigit1 = TRUE;
michael@0:                 break;
michael@0:             default:
michael@0:                 c1 = (char)type; /* lowercased letter */
michael@0:                 afterDigit1 = FALSE;
michael@0:                 break;
michael@0:             }
michael@0:             break; /* deliver c1 */
michael@0:         }
michael@0:         while ((c2 = *name2++) != 0) {
michael@0:             type = GET_CHAR_TYPE(c2);
michael@0:             switch (type) {
michael@0:             case UIGNORE:
michael@0:                 afterDigit2 = FALSE;
michael@0:                 continue; /* ignore all but letters and digits */
michael@0:             case ZERO:
michael@0:                 if (!afterDigit2) {
michael@0:                     nextType = GET_CHAR_TYPE(*name2);
michael@0:                     if (nextType == ZERO || nextType == NONZERO) {
michael@0:                         continue; /* ignore leading zero before another digit */
michael@0:                     }
michael@0:                 }
michael@0:                 break;
michael@0:             case NONZERO:
michael@0:                 afterDigit2 = TRUE;
michael@0:                 break;
michael@0:             default:
michael@0:                 c2 = (char)type; /* lowercased letter */
michael@0:                 afterDigit2 = FALSE;
michael@0:                 break;
michael@0:             }
michael@0:             break; /* deliver c2 */
michael@0:         }
michael@0: 
michael@0:         /* If we reach the ends of both strings then they match */
michael@0:         if ((c1|c2)==0) {
michael@0:             return 0;
michael@0:         }
michael@0: 
michael@0:         /* Case-insensitive comparison */
michael@0:         rc = (int)(unsigned char)c1 - (int)(unsigned char)c2;
michael@0:         if (rc != 0) {
michael@0:             return rc;
michael@0:         }
michael@0:     }
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * search for an alias
michael@0:  * return the converter number index for gConverterList
michael@0:  */
michael@0: static inline uint32_t
michael@0: findConverter(const char *alias, UBool *containsOption, UErrorCode *pErrorCode) {
michael@0:     uint32_t mid, start, limit;
michael@0:     uint32_t lastMid;
michael@0:     int result;
michael@0:     int isUnnormalized = (gMainTable.optionTable->stringNormalizationType == UCNV_IO_UNNORMALIZED);
michael@0:     char strippedName[UCNV_MAX_CONVERTER_NAME_LENGTH];
michael@0: 
michael@0:     if (!isUnnormalized) {
michael@0:         if (uprv_strlen(alias) >= UCNV_MAX_CONVERTER_NAME_LENGTH) {
michael@0:             *pErrorCode = U_BUFFER_OVERFLOW_ERROR;
michael@0:             return UINT32_MAX;
michael@0:         }
michael@0: 
michael@0:         /* Lower case and remove ignoreable characters. */
michael@0:         ucnv_io_stripForCompare(strippedName, alias);
michael@0:         alias = strippedName;
michael@0:     }
michael@0: 
michael@0:     /* do a binary search for the alias */
michael@0:     start = 0;
michael@0:     limit = gMainTable.untaggedConvArraySize;
michael@0:     mid = limit;
michael@0:     lastMid = UINT32_MAX;
michael@0: 
michael@0:     for (;;) {
michael@0:         mid = (uint32_t)((start + limit) / 2);
michael@0:         if (lastMid == mid) {   /* Have we moved? */
michael@0:             break;  /* We haven't moved, and it wasn't found. */
michael@0:         }
michael@0:         lastMid = mid;
michael@0:         if (isUnnormalized) {
michael@0:             result = ucnv_compareNames(alias, GET_STRING(gMainTable.aliasList[mid]));
michael@0:         }
michael@0:         else {
michael@0:             result = uprv_strcmp(alias, GET_NORMALIZED_STRING(gMainTable.aliasList[mid]));
michael@0:         }
michael@0: 
michael@0:         if (result < 0) {
michael@0:             limit = mid;
michael@0:         } else if (result > 0) {
michael@0:             start = mid;
michael@0:         } else {
michael@0:             /* Since the gencnval tool folds duplicates into one entry,
michael@0:              * this alias in gAliasList is unique, but different standards
michael@0:              * may map an alias to different converters.
michael@0:              */
michael@0:             if (gMainTable.untaggedConvArray[mid] & UCNV_AMBIGUOUS_ALIAS_MAP_BIT) {
michael@0:                 *pErrorCode = U_AMBIGUOUS_ALIAS_WARNING;
michael@0:             }
michael@0:             /* State whether the canonical converter name contains an option.
michael@0:             This information is contained in this list in order to maintain backward & forward compatibility. */
michael@0:             if (containsOption) {
michael@0:                 UBool containsCnvOptionInfo = (UBool)gMainTable.optionTable->containsCnvOptionInfo;
michael@0:                 *containsOption = (UBool)((containsCnvOptionInfo
michael@0:                     && ((gMainTable.untaggedConvArray[mid] & UCNV_CONTAINS_OPTION_BIT) != 0))
michael@0:                     || !containsCnvOptionInfo);
michael@0:             }
michael@0:             return gMainTable.untaggedConvArray[mid] & UCNV_CONVERTER_INDEX_MASK;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     return UINT32_MAX;
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * Is this alias in this list?
michael@0:  * alias and listOffset should be non-NULL.
michael@0:  */
michael@0: static inline UBool
michael@0: isAliasInList(const char *alias, uint32_t listOffset) {
michael@0:     if (listOffset) {
michael@0:         uint32_t currAlias;
michael@0:         uint32_t listCount = gMainTable.taggedAliasLists[listOffset];
michael@0:         /* +1 to skip listCount */
michael@0:         const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1;
michael@0:         for (currAlias = 0; currAlias < listCount; currAlias++) {
michael@0:             if (currList[currAlias]
michael@0:                 && ucnv_compareNames(alias, GET_STRING(currList[currAlias]))==0)
michael@0:             {
michael@0:                 return TRUE;
michael@0:             }
michael@0:         }
michael@0:     }
michael@0:     return FALSE;
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * Search for an standard name of an alias (what is the default name
michael@0:  * that this standard uses?)
michael@0:  * return the listOffset for gTaggedAliasLists. If it's 0,
michael@0:  * the it couldn't be found, but the parameters are valid.
michael@0:  */
michael@0: static uint32_t
michael@0: findTaggedAliasListsOffset(const char *alias, const char *standard, UErrorCode *pErrorCode) {
michael@0:     uint32_t idx;
michael@0:     uint32_t listOffset;
michael@0:     uint32_t convNum;
michael@0:     UErrorCode myErr = U_ZERO_ERROR;
michael@0:     uint32_t tagNum = getTagNumber(standard);
michael@0: 
michael@0:     /* Make a quick guess. Hopefully they used a TR22 canonical alias. */
michael@0:     convNum = findConverter(alias, NULL, &myErr);
michael@0:     if (myErr != U_ZERO_ERROR) {
michael@0:         *pErrorCode = myErr;
michael@0:     }
michael@0: 
michael@0:     if (tagNum < (gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) && convNum < gMainTable.converterListSize) {
michael@0:         listOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + convNum];
michael@0:         if (listOffset && gMainTable.taggedAliasLists[listOffset + 1]) {
michael@0:             return listOffset;
michael@0:         }
michael@0:         if (myErr == U_AMBIGUOUS_ALIAS_WARNING) {
michael@0:             /* Uh Oh! They used an ambiguous alias.
michael@0:                We have to search the whole swiss cheese starting
michael@0:                at the highest standard affinity.
michael@0:                This may take a while.
michael@0:             */
michael@0:             for (idx = 0; idx < gMainTable.taggedAliasArraySize; idx++) {
michael@0:                 listOffset = gMainTable.taggedAliasArray[idx];
michael@0:                 if (listOffset && isAliasInList(alias, listOffset)) {
michael@0:                     uint32_t currTagNum = idx/gMainTable.converterListSize;
michael@0:                     uint32_t currConvNum = (idx - currTagNum*gMainTable.converterListSize);
michael@0:                     uint32_t tempListOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + currConvNum];
michael@0:                     if (tempListOffset && gMainTable.taggedAliasLists[tempListOffset + 1]) {
michael@0:                         return tempListOffset;
michael@0:                     }
michael@0:                     /* else keep on looking */
michael@0:                     /* We could speed this up by starting on the next row
michael@0:                        because an alias is unique per row, right now.
michael@0:                        This would change if alias versioning appears. */
michael@0:                 }
michael@0:             }
michael@0:             /* The standard doesn't know about the alias */
michael@0:         }
michael@0:         /* else no default name */
michael@0:         return 0;
michael@0:     }
michael@0:     /* else converter or tag not found */
michael@0: 
michael@0:     return UINT32_MAX;
michael@0: }
michael@0: 
michael@0: /* Return the canonical name */
michael@0: static uint32_t
michael@0: findTaggedConverterNum(const char *alias, const char *standard, UErrorCode *pErrorCode) {
michael@0:     uint32_t idx;
michael@0:     uint32_t listOffset;
michael@0:     uint32_t convNum;
michael@0:     UErrorCode myErr = U_ZERO_ERROR;
michael@0:     uint32_t tagNum = getTagNumber(standard);
michael@0: 
michael@0:     /* Make a quick guess. Hopefully they used a TR22 canonical alias. */
michael@0:     convNum = findConverter(alias, NULL, &myErr);
michael@0:     if (myErr != U_ZERO_ERROR) {
michael@0:         *pErrorCode = myErr;
michael@0:     }
michael@0: 
michael@0:     if (tagNum < (gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) && convNum < gMainTable.converterListSize) {
michael@0:         listOffset = gMainTable.taggedAliasArray[tagNum*gMainTable.converterListSize + convNum];
michael@0:         if (listOffset && isAliasInList(alias, listOffset)) {
michael@0:             return convNum;
michael@0:         }
michael@0:         if (myErr == U_AMBIGUOUS_ALIAS_WARNING) {
michael@0:             /* Uh Oh! They used an ambiguous alias.
michael@0:                We have to search one slice of the swiss cheese.
michael@0:                We search only in the requested tag, not the whole thing.
michael@0:                This may take a while.
michael@0:             */
michael@0:             uint32_t convStart = (tagNum)*gMainTable.converterListSize;
michael@0:             uint32_t convLimit = (tagNum+1)*gMainTable.converterListSize;
michael@0:             for (idx = convStart; idx < convLimit; idx++) {
michael@0:                 listOffset = gMainTable.taggedAliasArray[idx];
michael@0:                 if (listOffset && isAliasInList(alias, listOffset)) {
michael@0:                     return idx-convStart;
michael@0:                 }
michael@0:             }
michael@0:             /* The standard doesn't know about the alias */
michael@0:         }
michael@0:         /* else no canonical name */
michael@0:     }
michael@0:     /* else converter or tag not found */
michael@0: 
michael@0:     return UINT32_MAX;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CFUNC const char *
michael@0: ucnv_io_getConverterName(const char *alias, UBool *containsOption, UErrorCode *pErrorCode) {
michael@0:     const char *aliasTmp = alias;
michael@0:     int32_t i = 0;
michael@0:     for (i = 0; i < 2; i++) {
michael@0:         if (i == 1) {
michael@0:             /*
michael@0:              * After the first unsuccess converter lookup, check to see if
michael@0:              * the name begins with 'x-'. If it does, strip it off and try
michael@0:              * again.  This behaviour is similar to how ICU4J does it.
michael@0:              */
michael@0:             if (aliasTmp[0] == 'x' || aliasTmp[1] == '-') {
michael@0:                 aliasTmp = aliasTmp+2;
michael@0:             } else {
michael@0:                 break;
michael@0:             }
michael@0:         }
michael@0:         if(haveAliasData(pErrorCode) && isAlias(aliasTmp, pErrorCode)) {
michael@0:             uint32_t convNum = findConverter(aliasTmp, containsOption, pErrorCode);
michael@0:             if (convNum < gMainTable.converterListSize) {
michael@0:                 return GET_STRING(gMainTable.converterList[convNum]);
michael@0:             }
michael@0:             /* else converter not found */
michael@0:         } else {
michael@0:             break;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     return NULL;
michael@0: }
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: ucnv_io_countStandardAliases(UEnumeration *enumerator, UErrorCode * /*pErrorCode*/) {
michael@0:     int32_t value = 0;
michael@0:     UAliasContext *myContext = (UAliasContext *)(enumerator->context);
michael@0:     uint32_t listOffset = myContext->listOffset;
michael@0: 
michael@0:     if (listOffset) {
michael@0:         value = gMainTable.taggedAliasLists[listOffset];
michael@0:     }
michael@0:     return value;
michael@0: }
michael@0: 
michael@0: static const char* U_CALLCONV
michael@0: ucnv_io_nextStandardAliases(UEnumeration *enumerator,
michael@0:                             int32_t* resultLength,
michael@0:                             UErrorCode * /*pErrorCode*/)
michael@0: {
michael@0:     UAliasContext *myContext = (UAliasContext *)(enumerator->context);
michael@0:     uint32_t listOffset = myContext->listOffset;
michael@0: 
michael@0:     if (listOffset) {
michael@0:         uint32_t listCount = gMainTable.taggedAliasLists[listOffset];
michael@0:         const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1;
michael@0: 
michael@0:         if (myContext->listIdx < listCount) {
michael@0:             const char *myStr = GET_STRING(currList[myContext->listIdx++]);
michael@0:             if (resultLength) {
michael@0:                 *resultLength = (int32_t)uprv_strlen(myStr);
michael@0:             }
michael@0:             return myStr;
michael@0:         }
michael@0:     }
michael@0:     /* Either we accessed a zero length list, or we enumerated too far. */
michael@0:     if (resultLength) {
michael@0:         *resultLength = 0;
michael@0:     }
michael@0:     return NULL;
michael@0: }
michael@0: 
michael@0: static void U_CALLCONV
michael@0: ucnv_io_resetStandardAliases(UEnumeration *enumerator, UErrorCode * /*pErrorCode*/) {
michael@0:     ((UAliasContext *)(enumerator->context))->listIdx = 0;
michael@0: }
michael@0: 
michael@0: static void U_CALLCONV
michael@0: ucnv_io_closeUEnumeration(UEnumeration *enumerator) {
michael@0:     uprv_free(enumerator->context);
michael@0:     uprv_free(enumerator);
michael@0: }
michael@0: 
michael@0: /* Enumerate the aliases for the specified converter and standard tag */
michael@0: static const UEnumeration gEnumAliases = {
michael@0:     NULL,
michael@0:     NULL,
michael@0:     ucnv_io_closeUEnumeration,
michael@0:     ucnv_io_countStandardAliases,
michael@0:     uenum_unextDefault,
michael@0:     ucnv_io_nextStandardAliases,
michael@0:     ucnv_io_resetStandardAliases
michael@0: };
michael@0: 
michael@0: U_CAPI UEnumeration * U_EXPORT2
michael@0: ucnv_openStandardNames(const char *convName,
michael@0:                        const char *standard,
michael@0:                        UErrorCode *pErrorCode)
michael@0: {
michael@0:     UEnumeration *myEnum = NULL;
michael@0:     if (haveAliasData(pErrorCode) && isAlias(convName, pErrorCode)) {
michael@0:         uint32_t listOffset = findTaggedAliasListsOffset(convName, standard, pErrorCode);
michael@0: 
michael@0:         /* When listOffset == 0, we want to acknowledge that the
michael@0:            converter name and standard are okay, but there
michael@0:            is nothing to enumerate. */
michael@0:         if (listOffset < gMainTable.taggedAliasListsSize) {
michael@0:             UAliasContext *myContext;
michael@0: 
michael@0:             myEnum = static_cast<UEnumeration *>(uprv_malloc(sizeof(UEnumeration)));
michael@0:             if (myEnum == NULL) {
michael@0:                 *pErrorCode = U_MEMORY_ALLOCATION_ERROR;
michael@0:                 return NULL;
michael@0:             }
michael@0:             uprv_memcpy(myEnum, &gEnumAliases, sizeof(UEnumeration));
michael@0:             myContext = static_cast<UAliasContext *>(uprv_malloc(sizeof(UAliasContext)));
michael@0:             if (myContext == NULL) {
michael@0:                 *pErrorCode = U_MEMORY_ALLOCATION_ERROR;
michael@0:                 uprv_free(myEnum);
michael@0:                 return NULL;
michael@0:             }
michael@0:             myContext->listOffset = listOffset;
michael@0:             myContext->listIdx = 0;
michael@0:             myEnum->context = myContext;
michael@0:         }
michael@0:         /* else converter or tag not found */
michael@0:     }
michael@0:     return myEnum;
michael@0: }
michael@0: 
michael@0: static uint16_t
michael@0: ucnv_io_countAliases(const char *alias, UErrorCode *pErrorCode) {
michael@0:     if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) {
michael@0:         uint32_t convNum = findConverter(alias, NULL, pErrorCode);
michael@0:         if (convNum < gMainTable.converterListSize) {
michael@0:             /* tagListNum - 1 is the ALL tag */
michael@0:             int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum];
michael@0: 
michael@0:             if (listOffset) {
michael@0:                 return gMainTable.taggedAliasLists[listOffset];
michael@0:             }
michael@0:             /* else this shouldn't happen. internal program error */
michael@0:         }
michael@0:         /* else converter not found */
michael@0:     }
michael@0:     return 0;
michael@0: }
michael@0: 
michael@0: static uint16_t
michael@0: ucnv_io_getAliases(const char *alias, uint16_t start, const char **aliases, UErrorCode *pErrorCode) {
michael@0:     if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) {
michael@0:         uint32_t currAlias;
michael@0:         uint32_t convNum = findConverter(alias, NULL, pErrorCode);
michael@0:         if (convNum < gMainTable.converterListSize) {
michael@0:             /* tagListNum - 1 is the ALL tag */
michael@0:             int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum];
michael@0: 
michael@0:             if (listOffset) {
michael@0:                 uint32_t listCount = gMainTable.taggedAliasLists[listOffset];
michael@0:                 /* +1 to skip listCount */
michael@0:                 const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1;
michael@0: 
michael@0:                 for (currAlias = start; currAlias < listCount; currAlias++) {
michael@0:                     aliases[currAlias] = GET_STRING(currList[currAlias]);
michael@0:                 }
michael@0:             }
michael@0:             /* else this shouldn't happen. internal program error */
michael@0:         }
michael@0:         /* else converter not found */
michael@0:     }
michael@0:     return 0;
michael@0: }
michael@0: 
michael@0: static const char *
michael@0: ucnv_io_getAlias(const char *alias, uint16_t n, UErrorCode *pErrorCode) {
michael@0:     if(haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) {
michael@0:         uint32_t convNum = findConverter(alias, NULL, pErrorCode);
michael@0:         if (convNum < gMainTable.converterListSize) {
michael@0:             /* tagListNum - 1 is the ALL tag */
michael@0:             int32_t listOffset = gMainTable.taggedAliasArray[(gMainTable.tagListSize - 1)*gMainTable.converterListSize + convNum];
michael@0: 
michael@0:             if (listOffset) {
michael@0:                 uint32_t listCount = gMainTable.taggedAliasLists[listOffset];
michael@0:                 /* +1 to skip listCount */
michael@0:                 const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1;
michael@0: 
michael@0:                 if (n < listCount)  {
michael@0:                     return GET_STRING(currList[n]);
michael@0:                 }
michael@0:                 *pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:             }
michael@0:             /* else this shouldn't happen. internal program error */
michael@0:         }
michael@0:         /* else converter not found */
michael@0:     }
michael@0:     return NULL;
michael@0: }
michael@0: 
michael@0: static uint16_t
michael@0: ucnv_io_countStandards(UErrorCode *pErrorCode) {
michael@0:     if (haveAliasData(pErrorCode)) {
michael@0:         /* Don't include the empty list */
michael@0:         return (uint16_t)(gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS);
michael@0:     }
michael@0: 
michael@0:     return 0;
michael@0: }
michael@0: 
michael@0: U_CAPI const char * U_EXPORT2
michael@0: ucnv_getStandard(uint16_t n, UErrorCode *pErrorCode) {
michael@0:     if (haveAliasData(pErrorCode)) {
michael@0:         if (n < gMainTable.tagListSize - UCNV_NUM_HIDDEN_TAGS) {
michael@0:             return GET_STRING(gMainTable.tagList[n]);
michael@0:         }
michael@0:         *pErrorCode = U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:     }
michael@0: 
michael@0:     return NULL;
michael@0: }
michael@0: 
michael@0: U_CAPI const char * U_EXPORT2
michael@0: ucnv_getStandardName(const char *alias, const char *standard, UErrorCode *pErrorCode) {
michael@0:     if (haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) {
michael@0:         uint32_t listOffset = findTaggedAliasListsOffset(alias, standard, pErrorCode);
michael@0: 
michael@0:         if (0 < listOffset && listOffset < gMainTable.taggedAliasListsSize) {
michael@0:             const uint16_t *currList = gMainTable.taggedAliasLists + listOffset + 1;
michael@0: 
michael@0:             /* Get the preferred name from this list */
michael@0:             if (currList[0]) {
michael@0:                 return GET_STRING(currList[0]);
michael@0:             }
michael@0:             /* else someone screwed up the alias table. */
michael@0:             /* *pErrorCode = U_INVALID_FORMAT_ERROR */
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     return NULL;
michael@0: }
michael@0: 
michael@0: U_CAPI uint16_t U_EXPORT2
michael@0: ucnv_countAliases(const char *alias, UErrorCode *pErrorCode)
michael@0: {
michael@0:     return ucnv_io_countAliases(alias, pErrorCode);
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI const char* U_EXPORT2
michael@0: ucnv_getAlias(const char *alias, uint16_t n, UErrorCode *pErrorCode)
michael@0: {
michael@0:     return ucnv_io_getAlias(alias, n, pErrorCode);
michael@0: }
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucnv_getAliases(const char *alias, const char **aliases, UErrorCode *pErrorCode)
michael@0: {
michael@0:     ucnv_io_getAliases(alias, 0, aliases, pErrorCode);
michael@0: }
michael@0: 
michael@0: U_CAPI uint16_t U_EXPORT2
michael@0: ucnv_countStandards(void)
michael@0: {
michael@0:     UErrorCode err = U_ZERO_ERROR;
michael@0:     return ucnv_io_countStandards(&err);
michael@0: }
michael@0: 
michael@0: U_CAPI const char * U_EXPORT2
michael@0: ucnv_getCanonicalName(const char *alias, const char *standard, UErrorCode *pErrorCode) {
michael@0:     if (haveAliasData(pErrorCode) && isAlias(alias, pErrorCode)) {
michael@0:         uint32_t convNum = findTaggedConverterNum(alias, standard, pErrorCode);
michael@0: 
michael@0:         if (convNum < gMainTable.converterListSize) {
michael@0:             return GET_STRING(gMainTable.converterList[convNum]);
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     return NULL;
michael@0: }
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: ucnv_io_countAllConverters(UEnumeration * /*enumerator*/, UErrorCode * /*pErrorCode*/) {
michael@0:     return gMainTable.converterListSize;
michael@0: }
michael@0: 
michael@0: static const char* U_CALLCONV
michael@0: ucnv_io_nextAllConverters(UEnumeration *enumerator,
michael@0:                             int32_t* resultLength,
michael@0:                             UErrorCode * /*pErrorCode*/)
michael@0: {
michael@0:     uint16_t *myContext = (uint16_t *)(enumerator->context);
michael@0: 
michael@0:     if (*myContext < gMainTable.converterListSize) {
michael@0:         const char *myStr = GET_STRING(gMainTable.converterList[(*myContext)++]);
michael@0:         if (resultLength) {
michael@0:             *resultLength = (int32_t)uprv_strlen(myStr);
michael@0:         }
michael@0:         return myStr;
michael@0:     }
michael@0:     /* Either we accessed a zero length list, or we enumerated too far. */
michael@0:     if (resultLength) {
michael@0:         *resultLength = 0;
michael@0:     }
michael@0:     return NULL;
michael@0: }
michael@0: 
michael@0: static void U_CALLCONV
michael@0: ucnv_io_resetAllConverters(UEnumeration *enumerator, UErrorCode * /*pErrorCode*/) {
michael@0:     *((uint16_t *)(enumerator->context)) = 0;
michael@0: }
michael@0: 
michael@0: static const UEnumeration gEnumAllConverters = {
michael@0:     NULL,
michael@0:     NULL,
michael@0:     ucnv_io_closeUEnumeration,
michael@0:     ucnv_io_countAllConverters,
michael@0:     uenum_unextDefault,
michael@0:     ucnv_io_nextAllConverters,
michael@0:     ucnv_io_resetAllConverters
michael@0: };
michael@0: 
michael@0: U_CAPI UEnumeration * U_EXPORT2
michael@0: ucnv_openAllNames(UErrorCode *pErrorCode) {
michael@0:     UEnumeration *myEnum = NULL;
michael@0:     if (haveAliasData(pErrorCode)) {
michael@0:         uint16_t *myContext;
michael@0: 
michael@0:         myEnum = static_cast<UEnumeration *>(uprv_malloc(sizeof(UEnumeration)));
michael@0:         if (myEnum == NULL) {
michael@0:             *pErrorCode = U_MEMORY_ALLOCATION_ERROR;
michael@0:             return NULL;
michael@0:         }
michael@0:         uprv_memcpy(myEnum, &gEnumAllConverters, sizeof(UEnumeration));
michael@0:         myContext = static_cast<uint16_t *>(uprv_malloc(sizeof(uint16_t)));
michael@0:         if (myContext == NULL) {
michael@0:             *pErrorCode = U_MEMORY_ALLOCATION_ERROR;
michael@0:             uprv_free(myEnum);
michael@0:             return NULL;
michael@0:         }
michael@0:         *myContext = 0;
michael@0:         myEnum->context = myContext;
michael@0:     }
michael@0:     return myEnum;
michael@0: }
michael@0: 
michael@0: U_CFUNC uint16_t
michael@0: ucnv_io_countKnownConverters(UErrorCode *pErrorCode) {
michael@0:     if (haveAliasData(pErrorCode)) {
michael@0:         return (uint16_t)gMainTable.converterListSize;
michael@0:     }
michael@0:     return 0;
michael@0: }
michael@0: 
michael@0: /* alias table swapping ----------------------------------------------------- */
michael@0: 
michael@0: typedef char * U_CALLCONV StripForCompareFn(char *dst, const char *name);
michael@0: 
michael@0: /*
michael@0:  * row of a temporary array
michael@0:  *
michael@0:  * gets platform-endian charset string indexes and sorting indexes;
michael@0:  * after sorting this array by strings, the actual arrays are permutated
michael@0:  * according to the sorting indexes
michael@0:  */
michael@0: typedef struct TempRow {
michael@0:     uint16_t strIndex, sortIndex;
michael@0: } TempRow;
michael@0: 
michael@0: typedef struct TempAliasTable {
michael@0:     const char *chars;
michael@0:     TempRow *rows;
michael@0:     uint16_t *resort;
michael@0:     StripForCompareFn *stripForCompare;
michael@0: } TempAliasTable;
michael@0: 
michael@0: enum {
michael@0:     STACK_ROW_CAPACITY=500
michael@0: };
michael@0: 
michael@0: static int32_t
michael@0: io_compareRows(const void *context, const void *left, const void *right) {
michael@0:     char strippedLeft[UCNV_MAX_CONVERTER_NAME_LENGTH],
michael@0:          strippedRight[UCNV_MAX_CONVERTER_NAME_LENGTH];
michael@0: 
michael@0:     TempAliasTable *tempTable=(TempAliasTable *)context;
michael@0:     const char *chars=tempTable->chars;
michael@0: 
michael@0:     return (int32_t)uprv_strcmp(tempTable->stripForCompare(strippedLeft, chars+2*((const TempRow *)left)->strIndex),
michael@0:                                 tempTable->stripForCompare(strippedRight, chars+2*((const TempRow *)right)->strIndex));
michael@0: }
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: ucnv_swapAliases(const UDataSwapper *ds,
michael@0:                  const void *inData, int32_t length, void *outData,
michael@0:                  UErrorCode *pErrorCode) {
michael@0:     const UDataInfo *pInfo;
michael@0:     int32_t headerSize;
michael@0: 
michael@0:     const uint16_t *inTable;
michael@0:     const uint32_t *inSectionSizes;
michael@0:     uint32_t toc[offsetsCount];
michael@0:     uint32_t offsets[offsetsCount]; /* 16-bit-addressed offsets from inTable/outTable */
michael@0:     uint32_t i, count, tocLength, topOffset;
michael@0: 
michael@0:     TempRow rows[STACK_ROW_CAPACITY];
michael@0:     uint16_t resort[STACK_ROW_CAPACITY];
michael@0:     TempAliasTable tempTable;
michael@0: 
michael@0:     /* udata_swapDataHeader checks the arguments */
michael@0:     headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode);
michael@0:     if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     /* check data format and format version */
michael@0:     pInfo=(const UDataInfo *)((const char *)inData+4);
michael@0:     if(!(
michael@0:         pInfo->dataFormat[0]==0x43 &&   /* dataFormat="CvAl" */
michael@0:         pInfo->dataFormat[1]==0x76 &&
michael@0:         pInfo->dataFormat[2]==0x41 &&
michael@0:         pInfo->dataFormat[3]==0x6c &&
michael@0:         pInfo->formatVersion[0]==3
michael@0:     )) {
michael@0:         udata_printError(ds, "ucnv_swapAliases(): data format %02x.%02x.%02x.%02x (format version %02x) is not an alias table\n",
michael@0:                          pInfo->dataFormat[0], pInfo->dataFormat[1],
michael@0:                          pInfo->dataFormat[2], pInfo->dataFormat[3],
michael@0:                          pInfo->formatVersion[0]);
michael@0:         *pErrorCode=U_UNSUPPORTED_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     /* an alias table must contain at least the table of contents array */
michael@0:     if(length>=0 && (length-headerSize)<4*(1+minTocLength)) {
michael@0:         udata_printError(ds, "ucnv_swapAliases(): too few bytes (%d after header) for an alias table\n",
michael@0:                          length-headerSize);
michael@0:         *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     inSectionSizes=(const uint32_t *)((const char *)inData+headerSize);
michael@0:     inTable=(const uint16_t *)inSectionSizes;
michael@0:     uprv_memset(toc, 0, sizeof(toc));
michael@0:     toc[tocLengthIndex]=tocLength=ds->readUInt32(inSectionSizes[tocLengthIndex]);
michael@0:     if(tocLength<minTocLength || offsetsCount<=tocLength) {
michael@0:         udata_printError(ds, "ucnv_swapAliases(): table of contents contains unsupported number of sections (%u sections)\n", tocLength);
michael@0:         *pErrorCode=U_INVALID_FORMAT_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     /* read the known part of the table of contents */
michael@0:     for(i=converterListIndex; i<=tocLength; ++i) {
michael@0:         toc[i]=ds->readUInt32(inSectionSizes[i]);
michael@0:     }
michael@0: 
michael@0:     /* compute offsets */
michael@0:     uprv_memset(offsets, 0, sizeof(offsets));
michael@0:     offsets[converterListIndex]=2*(1+tocLength); /* count two 16-bit units per toc entry */
michael@0:     for(i=tagListIndex; i<=tocLength; ++i) {
michael@0:         offsets[i]=offsets[i-1]+toc[i-1];
michael@0:     }
michael@0: 
michael@0:     /* compute the overall size of the after-header data, in numbers of 16-bit units */
michael@0:     topOffset=offsets[i-1]+toc[i-1];
michael@0: 
michael@0:     if(length>=0) {
michael@0:         uint16_t *outTable;
michael@0:         const uint16_t *p, *p2;
michael@0:         uint16_t *q, *q2;
michael@0:         uint16_t oldIndex;
michael@0: 
michael@0:         if((length-headerSize)<(2*(int32_t)topOffset)) {
michael@0:             udata_printError(ds, "ucnv_swapAliases(): too few bytes (%d after header) for an alias table\n",
michael@0:                              length-headerSize);
michael@0:             *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:             return 0;
michael@0:         }
michael@0: 
michael@0:         outTable=(uint16_t *)((char *)outData+headerSize);
michael@0: 
michael@0:         /* swap the entire table of contents */
michael@0:         ds->swapArray32(ds, inTable, 4*(1+tocLength), outTable, pErrorCode);
michael@0: 
michael@0:         /* swap unormalized strings & normalized strings */
michael@0:         ds->swapInvChars(ds, inTable+offsets[stringTableIndex], 2*(int32_t)(toc[stringTableIndex]+toc[normalizedStringTableIndex]),
michael@0:                              outTable+offsets[stringTableIndex], pErrorCode);
michael@0:         if(U_FAILURE(*pErrorCode)) {
michael@0:             udata_printError(ds, "ucnv_swapAliases().swapInvChars(charset names) failed\n");
michael@0:             return 0;
michael@0:         }
michael@0: 
michael@0:         if(ds->inCharset==ds->outCharset) {
michael@0:             /* no need to sort, just swap all 16-bit values together */
michael@0:             ds->swapArray16(ds,
michael@0:                             inTable+offsets[converterListIndex],
michael@0:                             2*(int32_t)(offsets[stringTableIndex]-offsets[converterListIndex]),
michael@0:                             outTable+offsets[converterListIndex],
michael@0:                             pErrorCode);
michael@0:         } else {
michael@0:             /* allocate the temporary table for sorting */
michael@0:             count=toc[aliasListIndex];
michael@0: 
michael@0:             tempTable.chars=(const char *)(outTable+offsets[stringTableIndex]); /* sort by outCharset */
michael@0: 
michael@0:             if(count<=STACK_ROW_CAPACITY) {
michael@0:                 tempTable.rows=rows;
michael@0:                 tempTable.resort=resort;
michael@0:             } else {
michael@0:                 tempTable.rows=(TempRow *)uprv_malloc(count*sizeof(TempRow)+count*2);
michael@0:                 if(tempTable.rows==NULL) {
michael@0:                     udata_printError(ds, "ucnv_swapAliases(): unable to allocate memory for sorting tables (max length: %u)\n",
michael@0:                                      count);
michael@0:                     *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
michael@0:                     return 0;
michael@0:                 }
michael@0:                 tempTable.resort=(uint16_t *)(tempTable.rows+count);
michael@0:             }
michael@0: 
michael@0:             if(ds->outCharset==U_ASCII_FAMILY) {
michael@0:                 tempTable.stripForCompare=ucnv_io_stripASCIIForCompare;
michael@0:             } else /* U_EBCDIC_FAMILY */ {
michael@0:                 tempTable.stripForCompare=ucnv_io_stripEBCDICForCompare;
michael@0:             }
michael@0: 
michael@0:             /*
michael@0:              * Sort unique aliases+mapped names.
michael@0:              *
michael@0:              * We need to sort the list again by outCharset strings because they
michael@0:              * sort differently for different charset families.
michael@0:              * First we set up a temporary table with the string indexes and
michael@0:              * sorting indexes and sort that.
michael@0:              * Then we permutate and copy/swap the actual values.
michael@0:              */
michael@0:             p=inTable+offsets[aliasListIndex];
michael@0:             q=outTable+offsets[aliasListIndex];
michael@0: 
michael@0:             p2=inTable+offsets[untaggedConvArrayIndex];
michael@0:             q2=outTable+offsets[untaggedConvArrayIndex];
michael@0: 
michael@0:             for(i=0; i<count; ++i) {
michael@0:                 tempTable.rows[i].strIndex=ds->readUInt16(p[i]);
michael@0:                 tempTable.rows[i].sortIndex=(uint16_t)i;
michael@0:             }
michael@0: 
michael@0:             uprv_sortArray(tempTable.rows, (int32_t)count, sizeof(TempRow),
michael@0:                            io_compareRows, &tempTable,
michael@0:                            FALSE, pErrorCode);
michael@0: 
michael@0:             if(U_SUCCESS(*pErrorCode)) {
michael@0:                 /* copy/swap/permutate items */
michael@0:                 if(p!=q) {
michael@0:                     for(i=0; i<count; ++i) {
michael@0:                         oldIndex=tempTable.rows[i].sortIndex;
michael@0:                         ds->swapArray16(ds, p+oldIndex, 2, q+i, pErrorCode);
michael@0:                         ds->swapArray16(ds, p2+oldIndex, 2, q2+i, pErrorCode);
michael@0:                     }
michael@0:                 } else {
michael@0:                     /*
michael@0:                      * If we swap in-place, then the permutation must use another
michael@0:                      * temporary array (tempTable.resort)
michael@0:                      * before the results are copied to the outBundle.
michael@0:                      */
michael@0:                     uint16_t *r=tempTable.resort;
michael@0: 
michael@0:                     for(i=0; i<count; ++i) {
michael@0:                         oldIndex=tempTable.rows[i].sortIndex;
michael@0:                         ds->swapArray16(ds, p+oldIndex, 2, r+i, pErrorCode);
michael@0:                     }
michael@0:                     uprv_memcpy(q, r, 2*count);
michael@0: 
michael@0:                     for(i=0; i<count; ++i) {
michael@0:                         oldIndex=tempTable.rows[i].sortIndex;
michael@0:                         ds->swapArray16(ds, p2+oldIndex, 2, r+i, pErrorCode);
michael@0:                     }
michael@0:                     uprv_memcpy(q2, r, 2*count);
michael@0:                 }
michael@0:             }
michael@0: 
michael@0:             if(tempTable.rows!=rows) {
michael@0:                 uprv_free(tempTable.rows);
michael@0:             }
michael@0: 
michael@0:             if(U_FAILURE(*pErrorCode)) {
michael@0:                 udata_printError(ds, "ucnv_swapAliases().uprv_sortArray(%u items) failed\n",
michael@0:                                  count);
michael@0:                 return 0;
michael@0:             }
michael@0: 
michael@0:             /* swap remaining 16-bit values */
michael@0:             ds->swapArray16(ds,
michael@0:                             inTable+offsets[converterListIndex],
michael@0:                             2*(int32_t)(offsets[aliasListIndex]-offsets[converterListIndex]),
michael@0:                             outTable+offsets[converterListIndex],
michael@0:                             pErrorCode);
michael@0:             ds->swapArray16(ds,
michael@0:                             inTable+offsets[taggedAliasArrayIndex],
michael@0:                             2*(int32_t)(offsets[stringTableIndex]-offsets[taggedAliasArrayIndex]),
michael@0:                             outTable+offsets[taggedAliasArrayIndex],
michael@0:                             pErrorCode);
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     return headerSize+2*(int32_t)topOffset;
michael@0: }
michael@0: 
michael@0: #endif
michael@0: 
michael@0: 
michael@0: /*
michael@0:  * Hey, Emacs, please set the following:
michael@0:  *
michael@0:  * Local Variables:
michael@0:  * indent-tabs-mode: nil
michael@0:  * End:
michael@0:  *
michael@0:  */