michael@0: /*
michael@0:  *******************************************************************************
michael@0:  *   Copyright (C) 2003-2013, International Business Machines
michael@0:  *   Corporation and others.  All Rights Reserved.
michael@0:  *******************************************************************************
michael@0:  *   file name:  ucm.h
michael@0:  *   encoding:   US-ASCII
michael@0:  *   tab size:   8 (not used)
michael@0:  *   indentation:4
michael@0:  *
michael@0:  *   created on: 2003jun20
michael@0:  *   created by: Markus W. Scherer
michael@0:  *
michael@0:  *   Definitions for the .ucm file parser and handler module ucm.c.
michael@0:  */
michael@0: 
michael@0: #ifndef __UCM_H__
michael@0: #define __UCM_H__
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: #include "ucnvmbcs.h"
michael@0: #include "ucnv_ext.h"
michael@0: #include "filestrm.h"
michael@0: #include <stdio.h>
michael@0: 
michael@0: #if !UCONFIG_NO_CONVERSION
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: 
michael@0: /* constants for UCMapping.moveFlag */
michael@0: enum {
michael@0:     UCM_MOVE_TO_EXT=1,
michael@0:     UCM_REMOVE_MAPPING=2
michael@0: };
michael@0: 
michael@0: /*
michael@0:  * Per-mapping data structure
michael@0:  *
michael@0:  * u if uLen==1: Unicode code point
michael@0:  *   else index to uLen code points
michael@0:  * b if bLen<=4: up to 4 bytes
michael@0:  *   else index to bLen bytes
michael@0:  * uLen number of code points
michael@0:  * bLen number of words containing left-justified bytes
michael@0:  * bIsMultipleChars indicates that the bytes contain more than one sequence
michael@0:  *                  according to the state table
michael@0:  * f flag for roundtrip (0), fallback (1), sub mapping (2), reverse fallback (3)
michael@0:  *   or "good one-way" mapping (4).
michael@0:  *   Same values as in the source file after |
michael@0:  */
michael@0: typedef struct UCMapping {
michael@0:     UChar32 u;
michael@0:     union {
michael@0:         uint32_t idx;
michael@0:         uint8_t bytes[4];
michael@0:     } b;
michael@0:     int8_t uLen, bLen, f, moveFlag;
michael@0: } UCMapping;
michael@0: 
michael@0: /* constants for UCMTable.flagsType */
michael@0: enum {
michael@0:     UCM_FLAGS_INITIAL,  /* no mappings parsed yet */
michael@0:     UCM_FLAGS_EXPLICIT, /* .ucm file has mappings with | fallback indicators */
michael@0:     UCM_FLAGS_IMPLICIT, /* .ucm file has mappings without | fallback indicators, later wins */
michael@0:     UCM_FLAGS_MIXED     /* both implicit and explicit */
michael@0: };
michael@0: 
michael@0: typedef struct UCMTable {
michael@0:     UCMapping *mappings;
michael@0:     int32_t mappingsCapacity, mappingsLength;
michael@0: 
michael@0:     UChar32 *codePoints;
michael@0:     int32_t codePointsCapacity, codePointsLength;
michael@0: 
michael@0:     uint8_t *bytes;
michael@0:     int32_t bytesCapacity, bytesLength;
michael@0: 
michael@0:     /* index map for mapping by bytes first */
michael@0:     int32_t *reverseMap;
michael@0: 
michael@0:     uint8_t unicodeMask;
michael@0:     int8_t flagsType; /* UCM_FLAGS_INITIAL etc. */
michael@0:     UBool isSorted;
michael@0: } UCMTable;
michael@0: 
michael@0: enum {
michael@0:     MBCS_STATE_FLAG_DIRECT=1,
michael@0:     MBCS_STATE_FLAG_SURROGATES,
michael@0: 
michael@0:     MBCS_STATE_FLAG_READY=16
michael@0: };
michael@0: 
michael@0: typedef struct UCMStates {
michael@0:     int32_t stateTable[MBCS_MAX_STATE_COUNT][256];
michael@0:     uint32_t stateFlags[MBCS_MAX_STATE_COUNT],
michael@0:              stateOffsetSum[MBCS_MAX_STATE_COUNT];
michael@0: 
michael@0:     int32_t countStates, minCharLength, maxCharLength, countToUCodeUnits;
michael@0:     int8_t conversionType, outputType;
michael@0: } UCMStates;
michael@0: 
michael@0: typedef struct UCMFile {
michael@0:     UCMTable *base, *ext;
michael@0:     UCMStates states;
michael@0: 
michael@0:     char baseName[UCNV_MAX_CONVERTER_NAME_LENGTH];
michael@0: } UCMFile;
michael@0: 
michael@0: /* simple accesses ---------------------------------------------------------- */
michael@0: 
michael@0: #define UCM_GET_CODE_POINTS(t, m) \
michael@0:     (((m)->uLen==1) ? &(m)->u : (t)->codePoints+(m)->u)
michael@0: 
michael@0: #define UCM_GET_BYTES(t, m) \
michael@0:     (((m)->bLen<=4) ? (m)->b.bytes : (t)->bytes+(m)->b.idx)
michael@0: 
michael@0: /* APIs --------------------------------------------------------------------- */
michael@0: 
michael@0: U_CAPI UCMFile * U_EXPORT2
michael@0: ucm_open(void);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_close(UCMFile *ucm);
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: ucm_parseHeaderLine(UCMFile *ucm,
michael@0:                     char *line, char **pKey, char **pValue);
michael@0: 
michael@0: /* @return -1 illegal bytes  0 suitable for base table  1 needs to go into extension table */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: ucm_mappingType(UCMStates *baseStates,
michael@0:                 UCMapping *m,
michael@0:                 UChar32 codePoints[UCNV_EXT_MAX_UCHARS],
michael@0:                 uint8_t bytes[UCNV_EXT_MAX_BYTES]);
michael@0: 
michael@0: /* add a mapping to the base or extension table as appropriate */
michael@0: U_CAPI UBool U_EXPORT2
michael@0: ucm_addMappingAuto(UCMFile *ucm, UBool forBase, UCMStates *baseStates,
michael@0:                    UCMapping *m,
michael@0:                    UChar32 codePoints[UCNV_EXT_MAX_UCHARS],
michael@0:                    uint8_t bytes[UCNV_EXT_MAX_BYTES]);
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: ucm_addMappingFromLine(UCMFile *ucm, const char *line, UBool forBase, UCMStates *baseStates);
michael@0: 
michael@0: 
michael@0: U_CAPI UCMTable * U_EXPORT2
michael@0: ucm_openTable(void);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_closeTable(UCMTable *table);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_resetTable(UCMTable *table);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_sortTable(UCMTable *t);
michael@0: 
michael@0: /*
michael@0:  * Remove mappings with their move flag set from the base table
michael@0:  * and move some of them (with UCM_MOVE_TO_EXT) to the extension table.
michael@0:  */
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_moveMappings(UCMTable *base, UCMTable *ext);
michael@0: 
michael@0: /**
michael@0:  * Read a table from a .ucm file, from after the CHARMAP line to
michael@0:  * including the END CHARMAP line.
michael@0:  */
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_readTable(UCMFile *ucm, FileStream* convFile,
michael@0:               UBool forBase, UCMStates *baseStates,
michael@0:               UErrorCode *pErrorCode);
michael@0: 
michael@0: /**
michael@0:  * Check the validity of mappings against a base table's states;
michael@0:  * necessary for extension-only tables that were read before their base tables.
michael@0:  */
michael@0: U_CAPI UBool U_EXPORT2
michael@0: ucm_checkValidity(UCMTable *ext, UCMStates *baseStates);
michael@0: 
michael@0: /**
michael@0:  * Check a base table against an extension table.
michael@0:  * Set the moveTarget!=NULL if it is possible to move mappings from the base.
michael@0:  * This is the case where base and extension tables are parsed from a single file
michael@0:  * (moveTarget==ext)
michael@0:  * or when delta file mappings are subtracted from a base table.
michael@0:  *
michael@0:  * When a base table cannot be modified because a delta file is parsed in makeconv,
michael@0:  * then set moveTarget=NULL.
michael@0:  *
michael@0:  * if(intersectBase) then mappings that exist in the base table but not in
michael@0:  * the extension table are moved to moveTarget instead of showing an error.
michael@0:  *
michael@0:  * Special mode:
michael@0:  * If intersectBase==2 for a DBCS extension table, then SBCS mappings are
michael@0:  * not moved out of the base unless their Unicode input requires it.
michael@0:  * This helps ucmkbase generate base tables for DBCS-only extension .cnv files.
michael@0:  *
michael@0:  * For both tables in the same file, the extension table is automatically
michael@0:  * built.
michael@0:  * For separate files, the extension file can use a complete mapping table (.ucm file),
michael@0:  * so that common mappings need not be stripped out manually.
michael@0:  *
michael@0:  *
michael@0:  * Sort both tables, and then for each mapping direction:
michael@0:  *
michael@0:  * If intersectBase is TRUE and the base table contains a mapping
michael@0:  * that does not exist in the extension table, then this mapping is moved
michael@0:  * to moveTarget.
michael@0:  *
michael@0:  * - otherwise -
michael@0:  *
michael@0:  * If the base table contains a mapping for which the input sequence is
michael@0:  * the same as the extension input, then
michael@0:  * - if the output is the same: remove the extension mapping
michael@0:  * - else: error
michael@0:  *
michael@0:  * If the base table contains a mapping for which the input sequence is
michael@0:  * a prefix of the extension input, then
michael@0:  * - if moveTarget!=NULL: move the base mapping to the moveTarget table
michael@0:  * - else: error
michael@0:  *
michael@0:  * @return FALSE in case of an irreparable error
michael@0:  */
michael@0: U_CAPI UBool U_EXPORT2
michael@0: ucm_checkBaseExt(UCMStates *baseStates, UCMTable *base, UCMTable *ext,
michael@0:                  UCMTable *moveTarget, UBool intersectBase);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_printTable(UCMTable *table, FILE *f, UBool byUnicode);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_printMapping(UCMTable *table, UCMapping *m, FILE *f);
michael@0: 
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_addState(UCMStates *states, const char *s);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_processStates(UCMStates *states, UBool ignoreSISOCheck);
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: ucm_countChars(UCMStates *states,
michael@0:                const uint8_t *bytes, int32_t length);
michael@0: 
michael@0: 
michael@0: U_CAPI int8_t U_EXPORT2
michael@0: ucm_parseBytes(uint8_t bytes[UCNV_EXT_MAX_BYTES], const char *line, const char **ps);
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: ucm_parseMappingLine(UCMapping *m,
michael@0:                      UChar32 codePoints[UCNV_EXT_MAX_UCHARS],
michael@0:                      uint8_t bytes[UCNV_EXT_MAX_BYTES],
michael@0:                      const char *line);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_addMapping(UCMTable *table,
michael@0:                UCMapping *m,
michael@0:                UChar32 codePoints[UCNV_EXT_MAX_UCHARS],
michael@0:                uint8_t bytes[UCNV_EXT_MAX_BYTES]);
michael@0: 
michael@0: /* very makeconv-specific functions ----------------------------------------- */
michael@0: 
michael@0: /* finalize and optimize states after the toUnicode mappings are processed */
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_optimizeStates(UCMStates *states,
michael@0:                    uint16_t **pUnicodeCodeUnits,
michael@0:                    _MBCSToUFallback *toUFallbacks, int32_t countToUFallbacks,
michael@0:                    UBool verbose);
michael@0: 
michael@0: /* moved here because it is used inside ucmstate.c */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: ucm_findFallback(_MBCSToUFallback *toUFallbacks, int32_t countToUFallbacks,
michael@0:                  uint32_t offset);
michael@0: 
michael@0: /* very rptp2ucm-specific functions ----------------------------------------- */
michael@0: 
michael@0: /*
michael@0:  * Input: Separate tables with mappings from/to Unicode,
michael@0:  * subchar and subchar1 (0 if none).
michael@0:  * All mappings must have flag 0.
michael@0:  *
michael@0:  * Output: fromUTable will contain the union of mappings with the correct
michael@0:  * precision flags, and be sorted.
michael@0:  */
michael@0: U_CAPI void U_EXPORT2
michael@0: ucm_mergeTables(UCMTable *fromUTable, UCMTable *toUTable,
michael@0:                 const uint8_t *subchar, int32_t subcharLength,
michael@0:                 uint8_t subchar1);
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: ucm_separateMappings(UCMFile *ucm, UBool isSISO);
michael@0: 
michael@0: U_CDECL_END
michael@0: 
michael@0: #endif
michael@0: 
michael@0: #endif
michael@0: