// © 2016 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html
/*
**********************************************************************
*   Copyright (C) 2000-2016, International Business Machines
*   Corporation and others.  All Rights Reserved.
**********************************************************************
*   file name:  ucnv2022.cpp
*   encoding:   UTF-8
*   tab size:   8 (not used)
*   indentation:4
*
*   created on: 2000feb03
*   created by: Markus W. Scherer
*
*   Change history:
*
*   06/29/2000  helena  Major rewrite of the callback APIs.
*   08/08/2000  Ram     Included support for ISO-2022-JP-2
*                       Changed implementation of toUnicode
*                       function
*   08/21/2000  Ram     Added support for ISO-2022-KR
*   08/29/2000  Ram     Seperated implementation of EBCDIC to
*                       ucnvebdc.c
*   09/20/2000  Ram     Added support for ISO-2022-CN
*                       Added implementations for getNextUChar()
*                       for specific 2022 country variants.
*   10/31/2000  Ram     Implemented offsets logic functions
*/

#include "unicode/utypes.h"

#if !UCONFIG_NO_CONVERSION && !UCONFIG_NO_LEGACY_CONVERSION

#include "unicode/ucnv.h"
#include "unicode/uset.h"
#include "unicode/ucnv_err.h"
#include "unicode/ucnv_cb.h"
#include "unicode/utf16.h"
#include "ucnv_imp.h"
#include "ucnv_bld.h"
#include "ucnv_cnv.h"
#include "ucnvmbcs.h"
#include "cstring.h"
#include "cmemory.h"
#include "uassert.h"

#ifdef U_ENABLE_GENERIC_ISO_2022
/*
 * I am disabling the generic ISO-2022 converter after proposing to do so on
 * the icu mailing list two days ago.
 *
 * Reasons:
 * 1. It does not fully support the ISO-2022/ECMA-35 specification with all of
 *    its designation sequences, single shifts with return to the previous state,
 *    switch-with-no-return to UTF-16BE or similar, etc.
 *    This is unlike the language-specific variants like ISO-2022-JP which
 *    require a much smaller repertoire of ISO-2022 features.
 *    These variants continue to be supported.
 * 2. I believe that no one is really using the generic ISO-2022 converter
 *    but rather always one of the language-specific variants.
 *    Note that ICU's generic ISO-2022 converter has always output one escape
 *    sequence followed by UTF-8 for the whole stream.
 * 3. Switching between subcharsets is extremely slow, because each time
 *    the previous converter is closed and a new one opened,
 *    without any kind of caching, least-recently-used list, etc.
 * 4. The code is currently buggy, and given the above it does not seem
 *    reasonable to spend the time on maintenance.
 * 5. ISO-2022 subcharsets should normally be used with 7-bit byte encodings.
 *    This means, for example, that when ISO-8859-7 is designated, the following
 *    ISO-2022 bytes 00..7f should be interpreted as ISO-8859-7 bytes 80..ff.
 *    The ICU ISO-2022 converter does not handle this - and has no information
 *    about which subconverter would have to be shifted vs. which is designed
 *    for 7-bit ISO-2022.
 *
 * Markus Scherer 2003-dec-03
 */
#endif

#if !UCONFIG_ONLY_HTML_CONVERSION
static const char SHIFT_IN_STR[]  = "\x0F";
// static const char SHIFT_OUT_STR[] = "\x0E";
#endif

#define CR      0x0D
#define LF      0x0A
#define H_TAB   0x09
#define V_TAB   0x0B
#define SPACE   0x20

enum {
    HWKANA_START=0xff61,
    HWKANA_END=0xff9f
};

/*
 * 94-character sets with native byte values A1..FE are encoded in ISO 2022
 * as bytes 21..7E. (Subtract 0x80.)
 * 96-character sets with native byte values A0..FF are encoded in ISO 2022
 * as bytes 20..7F. (Subtract 0x80.)
 * Do not encode C1 control codes with native bytes 80..9F
 * as bytes 00..1F (C0 control codes).
 */
enum {
    GR94_START=0xa1,
    GR94_END=0xfe,
    GR96_START=0xa0,
    GR96_END=0xff
};

/*
 * ISO 2022 control codes must not be converted from Unicode
 * because they would mess up the byte stream.
 * The bit mask 0x0800c000 has bits set at bit positions 0xe, 0xf, 0x1b
 * corresponding to SO, SI, and ESC.
 */
#define IS_2022_CONTROL(c) (((c)<0x20) && (((uint32_t)1<<(c))&0x0800c000)!=0)

/* for ISO-2022-JP and -CN implementations */
typedef enum  {
        /* shared values */
        INVALID_STATE=-1,
        ASCII = 0,

        SS2_STATE=0x10,
        SS3_STATE,

        /* JP */
        ISO8859_1 = 1 ,
        ISO8859_7 = 2 ,
        JISX201  = 3,
        JISX208 = 4,
        JISX212 = 5,
        GB2312  =6,
        KSC5601 =7,
        HWKANA_7BIT=8,    /* Halfwidth Katakana 7 bit */

        /* CN */
        /* the first few enum constants must keep their values because they correspond to myConverterArray[] */
        GB2312_1=1,
        ISO_IR_165=2,
        CNS_11643=3,

        /*
         * these are used in StateEnum and ISO2022State variables,
         * but CNS_11643 must be used to index into myConverterArray[]
         */
        CNS_11643_0=0x20,
        CNS_11643_1,
        CNS_11643_2,
        CNS_11643_3,
        CNS_11643_4,
        CNS_11643_5,
        CNS_11643_6,
        CNS_11643_7
} StateEnum;

/* is the StateEnum charset value for a DBCS charset? */
#if UCONFIG_ONLY_HTML_CONVERSION
#define IS_JP_DBCS(cs) (JISX208==(cs))
#else
#define IS_JP_DBCS(cs) (JISX208<=(cs) && (cs)<=KSC5601)
#endif

#define CSM(cs) ((uint16_t)1<<(cs))

/*
 * Each of these charset masks (with index x) contains a bit for a charset in exact correspondence
 * to whether that charset is used in the corresponding version x of ISO_2022,locale=ja,version=x
 *
 * Note: The converter uses some leniency:
 * - The escape sequence ESC ( I for half-width 7-bit Katakana is recognized in
 *   all versions, not just JIS7 and JIS8.
 * - ICU does not distinguish between different versions of JIS X 0208.
 */
#if UCONFIG_ONLY_HTML_CONVERSION
enum { MAX_JA_VERSION=0 };
#else
enum { MAX_JA_VERSION=4 };
#endif
static const uint16_t jpCharsetMasks[MAX_JA_VERSION+1]={
    CSM(ASCII)|CSM(JISX201)|CSM(JISX208)|CSM(HWKANA_7BIT),
#if !UCONFIG_ONLY_HTML_CONVERSION
    CSM(ASCII)|CSM(JISX201)|CSM(JISX208)|CSM(HWKANA_7BIT)|CSM(JISX212),
    CSM(ASCII)|CSM(JISX201)|CSM(JISX208)|CSM(HWKANA_7BIT)|CSM(JISX212)|CSM(GB2312)|CSM(KSC5601)|CSM(ISO8859_1)|CSM(ISO8859_7),
    CSM(ASCII)|CSM(JISX201)|CSM(JISX208)|CSM(HWKANA_7BIT)|CSM(JISX212)|CSM(GB2312)|CSM(KSC5601)|CSM(ISO8859_1)|CSM(ISO8859_7),
    CSM(ASCII)|CSM(JISX201)|CSM(JISX208)|CSM(HWKANA_7BIT)|CSM(JISX212)|CSM(GB2312)|CSM(KSC5601)|CSM(ISO8859_1)|CSM(ISO8859_7)
#endif
};

typedef enum {
        ASCII1=0,
        LATIN1,
        SBCS,
        DBCS,
        MBCS,
        HWKANA
}Cnv2022Type;

typedef struct ISO2022State {
    int8_t cs[4];       /* charset number for SI (G0)/SO (G1)/SS2 (G2)/SS3 (G3) */
    int8_t g;           /* 0..3 for G0..G3 (SI/SO/SS2/SS3) */
    int8_t prevG;       /* g before single shift (SS2 or SS3) */
} ISO2022State;

#define UCNV_OPTIONS_VERSION_MASK 0xf
#define UCNV_2022_MAX_CONVERTERS 10

typedef struct{
    UConverterSharedData *myConverterArray[UCNV_2022_MAX_CONVERTERS];
    UConverter *currentConverter;
    Cnv2022Type currentType;
    ISO2022State toU2022State, fromU2022State;
    uint32_t key;
    uint32_t version;
#ifdef U_ENABLE_GENERIC_ISO_2022
    UBool isFirstBuffer;
#endif
    UBool isEmptySegment;
    char name[30];
    char locale[3];
}UConverterDataISO2022;

/* Protos */
/* ISO-2022 ----------------------------------------------------------------- */

/*Forward declaration */
U_CFUNC void U_CALLCONV
ucnv_fromUnicode_UTF8(UConverterFromUnicodeArgs * args,
                      UErrorCode * err);
U_CFUNC void U_CALLCONV
ucnv_fromUnicode_UTF8_OFFSETS_LOGIC(UConverterFromUnicodeArgs * args,
                                    UErrorCode * err);

#define ESC_2022 0x1B /*ESC*/

typedef enum
{
        INVALID_2022 = -1, /*Doesn't correspond to a valid iso 2022 escape sequence*/
        VALID_NON_TERMINAL_2022 = 0, /*so far corresponds to a valid iso 2022 escape sequence*/
        VALID_TERMINAL_2022 = 1, /*corresponds to a valid iso 2022 escape sequence*/
        VALID_MAYBE_TERMINAL_2022 = 2 /*so far matches one iso 2022 escape sequence, but by adding more characters might match another escape sequence*/
} UCNV_TableStates_2022;

/*
* The way these state transition arrays work is:
* ex : ESC$B is the sequence for JISX208
*      a) First Iteration: char is ESC
*          i) Get the value of ESC from normalize_esq_chars_2022[] with int value of ESC as index
*             int x = normalize_esq_chars_2022[27] which is equal to 1
*         ii) Search for this value in escSeqStateTable_Key_2022[]
*             value of x is stored at escSeqStateTable_Key_2022[0]
*        iii) Save this index as offset
*         iv) Get state of this sequence from escSeqStateTable_Value_2022[]
*             escSeqStateTable_Value_2022[offset], which is VALID_NON_TERMINAL_2022
*     b) Switch on this state and continue to next char
*          i) Get the value of $ from normalize_esq_chars_2022[] with int value of $ as index
*             which is normalize_esq_chars_2022[36] == 4
*         ii) x is currently 1(from above)
*               x<<=5 -- x is now 32
*               x+=normalize_esq_chars_2022[36]
*               now x is 36
*        iii) Search for this value in escSeqStateTable_Key_2022[]
*             value of x is stored at escSeqStateTable_Key_2022[2], so offset is 2
*         iv) Get state of this sequence from escSeqStateTable_Value_2022[]
*             escSeqStateTable_Value_2022[offset], which is VALID_NON_TERMINAL_2022
*     c) Switch on this state and continue to next char
*        i)  Get the value of B from normalize_esq_chars_2022[] with int value of B as index
*        ii) x is currently 36 (from above)
*            x<<=5 -- x is now 1152
*            x+=normalize_esq_chars_2022[66]
*            now x is 1161
*       iii) Search for this value in escSeqStateTable_Key_2022[]
*            value of x is stored at escSeqStateTable_Key_2022[21], so offset is 21
*        iv) Get state of this sequence from escSeqStateTable_Value_2022[21]
*            escSeqStateTable_Value_2022[offset], which is VALID_TERMINAL_2022
*         v) Get the converter name form escSeqStateTable_Result_2022[21] which is JISX208
*/


/*Below are the 3 arrays depicting a state transition table*/
static const int8_t normalize_esq_chars_2022[256] = {
/*       0      1       2       3       4      5       6        7       8       9           */

         0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,1      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,4      ,7      ,29      ,0
        ,2     ,24     ,26     ,27     ,0      ,3      ,23     ,6      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,5      ,8      ,9      ,10     ,11     ,12
        ,13    ,14     ,15     ,16     ,17     ,18     ,19     ,20     ,25     ,28
        ,0     ,0      ,21     ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,22    ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0      ,0
        ,0     ,0      ,0      ,0      ,0      ,0
};

#ifdef U_ENABLE_GENERIC_ISO_2022
/*
 * When the generic ISO-2022 converter is completely removed, not just disabled
 * per #ifdef, then the following state table and the associated tables that are
 * dimensioned with MAX_STATES_2022 should be trimmed.
 *
 * Especially, VALID_MAYBE_TERMINAL_2022 will not be used any more, and all of
 * the associated escape sequences starting with ESC ( B should be removed.
 * This includes the ones with key values 1097 and all of the ones above 1000000.
 *
 * For the latter, the tables can simply be truncated.
 * For the former, since the tables must be kept parallel, it is probably best
 * to simply duplicate an adjacent table cell, parallel in all tables.
 *
 * It may make sense to restructure the tables, especially by using small search
 * tables for the variants instead of indexing them parallel to the table here.
 */
#endif

#define MAX_STATES_2022 74
static const int32_t escSeqStateTable_Key_2022[MAX_STATES_2022] = {
/*   0           1           2           3           4           5           6           7           8           9           */

     1          ,34         ,36         ,39         ,55         ,57         ,60         ,61         ,1093       ,1096
    ,1097       ,1098       ,1099       ,1100       ,1101       ,1102       ,1103       ,1104       ,1105       ,1106
    ,1109       ,1154       ,1157       ,1160       ,1161       ,1176       ,1178       ,1179       ,1254       ,1257
    ,1768       ,1773       ,1957       ,35105      ,36933      ,36936      ,36937      ,36938      ,36939      ,36940
    ,36942      ,36943      ,36944      ,36945      ,36946      ,36947      ,36948      ,37640      ,37642      ,37644
    ,37646      ,37711      ,37744      ,37745      ,37746      ,37747      ,37748      ,40133      ,40136      ,40138
    ,40139      ,40140      ,40141      ,1123363    ,35947624   ,35947625   ,35947626   ,35947627   ,35947629   ,35947630
    ,35947631   ,35947635   ,35947636   ,35947638
};

#ifdef U_ENABLE_GENERIC_ISO_2022

static const char* const escSeqStateTable_Result_2022[MAX_STATES_2022] = {
 /*  0                      1                        2                      3                   4                   5                        6                      7                       8                       9    */

     nullptr                   ,nullptr                   ,nullptr                   ,nullptr               ,nullptr               ,nullptr                   ,nullptr                   ,nullptr                   ,"latin1"               ,"latin1"
    ,"latin1"               ,"ibm-865"              ,"ibm-865"              ,"ibm-865"          ,"ibm-865"          ,"ibm-865"              ,"ibm-865"              ,"JISX0201"             ,"JISX0201"             ,"latin1"
    ,"latin1"               ,nullptr                   ,"JISX-208"             ,"ibm-5478"         ,"JISX-208"         ,nullptr                   ,nullptr                   ,nullptr                   ,nullptr                   ,"UTF8"
    ,"ISO-8859-1"           ,"ISO-8859-7"           ,"JIS-X-208"            ,nullptr               ,"ibm-955"          ,"ibm-367"              ,"ibm-952"              ,"ibm-949"              ,"JISX-212"             ,"ibm-1383"
    ,"ibm-952"              ,"ibm-964"              ,"ibm-964"              ,"ibm-964"          ,"ibm-964"          ,"ibm-964"              ,"ibm-964"              ,"ibm-5478"         ,"ibm-949"              ,"ISO-IR-165"
    ,"CNS-11643-1992,1"     ,"CNS-11643-1992,2"     ,"CNS-11643-1992,3"     ,"CNS-11643-1992,4" ,"CNS-11643-1992,5" ,"CNS-11643-1992,6"     ,"CNS-11643-1992,7"     ,"UTF16_PlatformEndian" ,"UTF16_PlatformEndian" ,"UTF16_PlatformEndian"
    ,"UTF16_PlatformEndian" ,"UTF16_PlatformEndian" ,"UTF16_PlatformEndian" ,nullptr               ,"latin1"           ,"ibm-912"              ,"ibm-913"              ,"ibm-914"              ,"ibm-813"              ,"ibm-1089"
    ,"ibm-920"              ,"ibm-915"              ,"ibm-915"              ,"latin1"
};

#endif

static const int8_t escSeqStateTable_Value_2022[MAX_STATES_2022] = {
/*          0                           1                         2                             3                           4                           5                               6                        7                          8                           9       */
     VALID_NON_TERMINAL_2022    ,VALID_NON_TERMINAL_2022    ,VALID_NON_TERMINAL_2022    ,VALID_NON_TERMINAL_2022     ,VALID_NON_TERMINAL_2022   ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_NON_TERMINAL_2022    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022
    ,VALID_MAYBE_TERMINAL_2022  ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022
    ,VALID_TERMINAL_2022        ,VALID_NON_TERMINAL_2022    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_NON_TERMINAL_2022    ,VALID_NON_TERMINAL_2022    ,VALID_NON_TERMINAL_2022    ,VALID_NON_TERMINAL_2022    ,VALID_TERMINAL_2022
    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_NON_TERMINAL_2022    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022
    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022
    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022
    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_NON_TERMINAL_2022    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022
    ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022        ,VALID_TERMINAL_2022
};

/* Type def for refactoring changeState_2022 code*/
typedef enum{
#ifdef U_ENABLE_GENERIC_ISO_2022
    ISO_2022=0,
#endif
    ISO_2022_JP=1,
#if !UCONFIG_ONLY_HTML_CONVERSION
    ISO_2022_KR=2,
    ISO_2022_CN=3
#endif
} Variant2022;

/*********** ISO 2022 Converter Protos ***********/
static void U_CALLCONV
_ISO2022Open(UConverter *cnv, UConverterLoadArgs *pArgs, UErrorCode *errorCode);

static void U_CALLCONV
 _ISO2022Close(UConverter *converter);

static void U_CALLCONV
_ISO2022Reset(UConverter *converter, UConverterResetChoice choice);

U_CDECL_BEGIN
static const char * U_CALLCONV
_ISO2022getName(const UConverter* cnv);
U_CDECL_END

static void  U_CALLCONV
_ISO_2022_WriteSub(UConverterFromUnicodeArgs *args, int32_t offsetIndex, UErrorCode *err);

U_CDECL_BEGIN
static UConverter * U_CALLCONV
_ISO_2022_SafeClone(const UConverter *cnv, void *stackBuffer, int32_t *pBufferSize, UErrorCode *status);

U_CDECL_END

#ifdef U_ENABLE_GENERIC_ISO_2022
static void U_CALLCONV
T_UConverter_toUnicode_ISO_2022_OFFSETS_LOGIC(UConverterToUnicodeArgs* args, UErrorCode* err);
#endif

namespace {

/*const UConverterSharedData _ISO2022Data;*/
extern const UConverterSharedData _ISO2022JPData;

#if !UCONFIG_ONLY_HTML_CONVERSION
extern const UConverterSharedData _ISO2022KRData;
extern const UConverterSharedData _ISO2022CNData;
#endif

}  // namespace

/*************** Converter implementations ******************/

/* The purpose of this function is to get around gcc compiler warnings. */
static inline void
fromUWriteUInt8(UConverter *cnv,
                 const char *bytes, int32_t length,
                 uint8_t **target, const char *targetLimit,
                 int32_t **offsets,
                 int32_t sourceIndex,
                 UErrorCode *pErrorCode)
{
    char* targetChars = reinterpret_cast<char*>(*target);
    ucnv_fromUWriteBytes(cnv, bytes, length, &targetChars, targetLimit,
                         offsets, sourceIndex, pErrorCode);
    *target = reinterpret_cast<uint8_t*>(targetChars);

}

static inline void
setInitialStateToUnicodeKR(UConverter* /*converter*/, UConverterDataISO2022 *myConverterData){
    if(myConverterData->version == 1) {
        UConverter *cnv = myConverterData->currentConverter;

        cnv->toUnicodeStatus=0;     /* offset */
        cnv->mode=0;                /* state */
        cnv->toULength=0;           /* byteIndex */
    }
}

static inline void
setInitialStateFromUnicodeKR(UConverter* converter,UConverterDataISO2022 *myConverterData){
   /* in ISO-2022-KR the designator sequence appears only once
    * in a file so we append it only once
    */
    if( converter->charErrorBufferLength==0){

        converter->charErrorBufferLength = 4;
        converter->charErrorBuffer[0] = 0x1b;
        converter->charErrorBuffer[1] = 0x24;
        converter->charErrorBuffer[2] = 0x29;
        converter->charErrorBuffer[3] = 0x43;
    }
    if(myConverterData->version == 1) {
        UConverter *cnv = myConverterData->currentConverter;

        cnv->fromUChar32=0;
        cnv->fromUnicodeStatus=1;   /* prevLength */
    }
}

static void U_CALLCONV
_ISO2022Open(UConverter *cnv, UConverterLoadArgs *pArgs, UErrorCode *errorCode){

    char myLocale[7]={' ',' ',' ',' ',' ',' ', '\0'};

    cnv->extraInfo = uprv_malloc (sizeof (UConverterDataISO2022));
    if(cnv->extraInfo != nullptr) {
        UConverterNamePieces stackPieces;
        UConverterLoadArgs stackArgs=UCNV_LOAD_ARGS_INITIALIZER;
        UConverterDataISO2022* myConverterData = static_cast<UConverterDataISO2022*>(cnv->extraInfo);
        uint32_t version;

        stackArgs.onlyTestIsLoadable = pArgs->onlyTestIsLoadable;

        uprv_memset(myConverterData, 0, sizeof(UConverterDataISO2022));
        myConverterData->currentType = ASCII1;
        cnv->fromUnicodeStatus =false;
        if(pArgs->locale){
            uprv_strncpy(myLocale, pArgs->locale, sizeof(myLocale)-1);
        }
        version = pArgs->options & UCNV_OPTIONS_VERSION_MASK;
        myConverterData->version = version;
        if(myLocale[0]=='j' && (myLocale[1]=='a'|| myLocale[1]=='p') &&
            (myLocale[2]=='_' || myLocale[2]=='\0'))
        {
            /* open the required converters and cache them */
            if(version>MAX_JA_VERSION) {
                // ICU 55 fails to open a converter for an unsupported version.
                // Previously, it fell back to version 0, but that would yield
                // unexpected behavior.
                *errorCode = U_MISSING_RESOURCE_ERROR;
                return;
            }
            if(jpCharsetMasks[version]&CSM(ISO8859_7)) {
                myConverterData->myConverterArray[ISO8859_7] =
                    ucnv_loadSharedData("ISO8859_7", &stackPieces, &stackArgs, errorCode);
            }
            myConverterData->myConverterArray[JISX208] =
                ucnv_loadSharedData("Shift-JIS", &stackPieces, &stackArgs, errorCode);
            if(jpCharsetMasks[version]&CSM(JISX212)) {
                myConverterData->myConverterArray[JISX212] =
                    ucnv_loadSharedData("jisx-212", &stackPieces, &stackArgs, errorCode);
            }
            if(jpCharsetMasks[version]&CSM(GB2312)) {
                myConverterData->myConverterArray[GB2312] =
                    ucnv_loadSharedData("ibm-5478", &stackPieces, &stackArgs, errorCode);   /* gb_2312_80-1 */
            }
            if(jpCharsetMasks[version]&CSM(KSC5601)) {
                myConverterData->myConverterArray[KSC5601] =
                    ucnv_loadSharedData("ksc_5601", &stackPieces, &stackArgs, errorCode);
            }

            /* set the function pointers to appropriate functions */
            cnv->sharedData = const_cast<UConverterSharedData*>(&_ISO2022JPData);
            uprv_strcpy(myConverterData->locale,"ja");

            (void)uprv_strcpy(myConverterData->name,"ISO_2022,locale=ja,version=");
            size_t len = uprv_strlen(myConverterData->name);
            myConverterData->name[len] = static_cast<char>(myConverterData->version + static_cast<int>('0'));
            myConverterData->name[len+1]='\0';
        }
#if !UCONFIG_ONLY_HTML_CONVERSION
        else if(myLocale[0]=='k' && (myLocale[1]=='o'|| myLocale[1]=='r') &&
            (myLocale[2]=='_' || myLocale[2]=='\0'))
        {
            if(version>1) {
                // ICU 55 fails to open a converter for an unsupported version.
                // Previously, it fell back to version 0, but that would yield
                // unexpected behavior.
                *errorCode = U_MISSING_RESOURCE_ERROR;
                return;
            }
            const char *cnvName;
            if(version==1) {
                cnvName="icu-internal-25546";
            } else {
                cnvName="ibm-949";
                myConverterData->version=version=0;
            }
            if(pArgs->onlyTestIsLoadable) {
                ucnv_canCreateConverter(cnvName, errorCode);  /* errorCode carries result */
                uprv_free(cnv->extraInfo);
                cnv->extraInfo=nullptr;
                return;
            } else {
                myConverterData->currentConverter=ucnv_open(cnvName, errorCode);
                if (U_FAILURE(*errorCode)) {
                    _ISO2022Close(cnv);
                    return;
                }

                if(version==1) {
                    (void)uprv_strcpy(myConverterData->name,"ISO_2022,locale=ko,version=1");
                    uprv_memcpy(cnv->subChars, myConverterData->currentConverter->subChars, 4);
                    cnv->subCharLen = myConverterData->currentConverter->subCharLen;
                }else{
                    (void)uprv_strcpy(myConverterData->name,"ISO_2022,locale=ko,version=0");
                }

                /* initialize the state variables */
                setInitialStateToUnicodeKR(cnv, myConverterData);
                setInitialStateFromUnicodeKR(cnv, myConverterData);

                /* set the function pointers to appropriate functions */
                cnv->sharedData = const_cast<UConverterSharedData*>(&_ISO2022KRData);
                uprv_strcpy(myConverterData->locale,"ko");
            }
        }
        else if(((myLocale[0]=='z' && myLocale[1]=='h') || (myLocale[0]=='c'&& myLocale[1]=='n'))&&
            (myLocale[2]=='_' || myLocale[2]=='\0'))
        {
            if(version>2) {
                // ICU 55 fails to open a converter for an unsupported version.
                // Previously, it fell back to version 0, but that would yield
                // unexpected behavior.
                *errorCode = U_MISSING_RESOURCE_ERROR;
                return;
            }

            /* open the required converters and cache them */
            myConverterData->myConverterArray[GB2312_1] =
                ucnv_loadSharedData("ibm-5478", &stackPieces, &stackArgs, errorCode);
            if(version>=1) {
                myConverterData->myConverterArray[ISO_IR_165] =
                    ucnv_loadSharedData("iso-ir-165", &stackPieces, &stackArgs, errorCode);
            }
            myConverterData->myConverterArray[CNS_11643] =
                ucnv_loadSharedData("cns-11643-1992", &stackPieces, &stackArgs, errorCode);


            /* set the function pointers to appropriate functions */
            cnv->sharedData = const_cast<UConverterSharedData*>(&_ISO2022CNData);
            uprv_strcpy(myConverterData->locale,"cn");

            if (version==0){
                myConverterData->version = 0;
                (void)uprv_strcpy(myConverterData->name,"ISO_2022,locale=zh,version=0");
            }else if (version==1){
                myConverterData->version = 1;
                (void)uprv_strcpy(myConverterData->name,"ISO_2022,locale=zh,version=1");
            }else {
                myConverterData->version = 2;
                (void)uprv_strcpy(myConverterData->name,"ISO_2022,locale=zh,version=2");
            }
        }
#endif  // !UCONFIG_ONLY_HTML_CONVERSION
        else{
#ifdef U_ENABLE_GENERIC_ISO_2022
            myConverterData->isFirstBuffer = true;

            /* append the UTF-8 escape sequence */
            cnv->charErrorBufferLength = 3;
            cnv->charErrorBuffer[0] = 0x1b;
            cnv->charErrorBuffer[1] = 0x25;
            cnv->charErrorBuffer[2] = 0x42;

            cnv->sharedData=(UConverterSharedData*)&_ISO2022Data;
            /* initialize the state variables */
            uprv_strcpy(myConverterData->name,"ISO_2022");
#else
            *errorCode = U_MISSING_RESOURCE_ERROR;
            // Was U_UNSUPPORTED_ERROR but changed in ICU 55 to a more standard
            // data loading error code.
            return;
#endif
        }

        cnv->maxBytesPerUChar=cnv->sharedData->staticData->maxBytesPerChar;

        if(U_FAILURE(*errorCode) || pArgs->onlyTestIsLoadable) {
            _ISO2022Close(cnv);
        }
    } else {
        *errorCode = U_MEMORY_ALLOCATION_ERROR;
    }
}


static void U_CALLCONV
_ISO2022Close(UConverter *converter) {
    UConverterDataISO2022* myData = static_cast<UConverterDataISO2022*>(converter->extraInfo);
    UConverterSharedData **array = myData->myConverterArray;
    int32_t i;

    if (converter->extraInfo != nullptr) {
        /*close the array of converter pointers and free the memory*/
        for (i=0; i<UCNV_2022_MAX_CONVERTERS; i++) {
            if(array[i]!=nullptr) {
                ucnv_unloadSharedDataIfReady(array[i]);
            }
        }

        ucnv_close(myData->currentConverter);

        if(!converter->isExtraLocal){
            uprv_free (converter->extraInfo);
            converter->extraInfo = nullptr;
        }
    }
}

static void U_CALLCONV
_ISO2022Reset(UConverter *converter, UConverterResetChoice choice) {
    UConverterDataISO2022* myConverterData = static_cast<UConverterDataISO2022*>(converter->extraInfo);
    if(choice<=UCNV_RESET_TO_UNICODE) {
        uprv_memset(&myConverterData->toU2022State, 0, sizeof(ISO2022State));
        myConverterData->key = 0;
        myConverterData->isEmptySegment = false;
    }
    if(choice!=UCNV_RESET_TO_UNICODE) {
        uprv_memset(&myConverterData->fromU2022State, 0, sizeof(ISO2022State));
    }
#ifdef U_ENABLE_GENERIC_ISO_2022
    if(myConverterData->locale[0] == 0){
        if(choice<=UCNV_RESET_TO_UNICODE) {
            myConverterData->isFirstBuffer = true;
            myConverterData->key = 0;
            if (converter->mode == UCNV_SO){
                ucnv_close (myConverterData->currentConverter);
                myConverterData->currentConverter=nullptr;
            }
            converter->mode = UCNV_SI;
        }
        if(choice!=UCNV_RESET_TO_UNICODE) {
            /* re-append UTF-8 escape sequence */
            converter->charErrorBufferLength = 3;
            converter->charErrorBuffer[0] = 0x1b;
            converter->charErrorBuffer[1] = 0x28;
            converter->charErrorBuffer[2] = 0x42;
        }
    }
    else
#endif
    {
        /* reset the state variables */
        if(myConverterData->locale[0] == 'k'){
            if(choice<=UCNV_RESET_TO_UNICODE) {
                setInitialStateToUnicodeKR(converter, myConverterData);
            }
            if(choice!=UCNV_RESET_TO_UNICODE) {
                setInitialStateFromUnicodeKR(converter, myConverterData);
            }
        }
    }
}

U_CDECL_BEGIN

static const char * U_CALLCONV
_ISO2022getName(const UConverter* cnv){
    if(cnv->extraInfo){
        UConverterDataISO2022* myData= (UConverterDataISO2022*)cnv->extraInfo;
        return myData->name;
    }
    return nullptr;
}

U_CDECL_END


/*************** to unicode *******************/
/****************************************************************************
 * Recognized escape sequences are
 * <ESC>(B  ASCII
 * <ESC>.A  ISO-8859-1
 * <ESC>.F  ISO-8859-7
 * <ESC>(J  JISX-201
 * <ESC>(I  JISX-201
 * <ESC>$B  JISX-208
 * <ESC>$@  JISX-208
 * <ESC>$(D JISX-212
 * <ESC>$A  GB2312
 * <ESC>$(C KSC5601
 */
static const int8_t nextStateToUnicodeJP[MAX_STATES_2022]= {
/*      0                1               2               3               4               5               6               7               8               9    */
    INVALID_STATE   ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,SS2_STATE      ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,ASCII          ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,JISX201        ,HWKANA_7BIT    ,JISX201        ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,JISX208        ,GB2312         ,JISX208        ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,ISO8859_1      ,ISO8859_7      ,JISX208        ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,KSC5601        ,JISX212        ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
};

#if !UCONFIG_ONLY_HTML_CONVERSION
/*************** to unicode *******************/
static const int8_t nextStateToUnicodeCN[MAX_STATES_2022]= {
/*      0                1               2               3               4               5               6               7               8               9    */
     INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,SS2_STATE      ,SS3_STATE      ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,GB2312_1       ,INVALID_STATE  ,ISO_IR_165
    ,CNS_11643_1    ,CNS_11643_2    ,CNS_11643_3    ,CNS_11643_4    ,CNS_11643_5    ,CNS_11643_6    ,CNS_11643_7    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
    ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE  ,INVALID_STATE
};
#endif


static UCNV_TableStates_2022
getKey_2022(char c,int32_t* key,int32_t* offset){
    int32_t togo;
    int32_t low = 0;
    int32_t hi = MAX_STATES_2022;
    int32_t oldmid=0;

    togo = normalize_esq_chars_2022[static_cast<uint8_t>(c)];
    if(togo == 0) {
        /* not a valid character anywhere in an escape sequence */
        *key = 0;
        *offset = 0;
        return INVALID_2022;
    }
    togo = (*key << 5) + togo;

    while (hi != low)  /*binary search*/{

        int32_t mid = (hi+low) >> 1; /*Finds median*/

        if (mid == oldmid)
            break;

        if (escSeqStateTable_Key_2022[mid] > togo){
            hi = mid;
        }
        else if (escSeqStateTable_Key_2022[mid] < togo){
            low = mid;
        }
        else /*we found it*/{
            *key = togo;
            *offset = mid;
            return static_cast<UCNV_TableStates_2022>(escSeqStateTable_Value_2022[mid]);
        }
        oldmid = mid;

    }

    *key = 0;
    *offset = 0;
    return INVALID_2022;
}

/*runs through a state machine to determine the escape sequence - codepage correspondence
 */
static void
changeState_2022(UConverter* _this,
                const char** source,
                const char* sourceLimit,
                Variant2022 var,
                UErrorCode* err){
    UCNV_TableStates_2022 value;
    UConverterDataISO2022* myData2022 = static_cast<UConverterDataISO2022*>(_this->extraInfo);
    uint32_t key = myData2022->key;
    int32_t offset = 0;
    int8_t initialToULength = _this->toULength;
    char c;

    value = VALID_NON_TERMINAL_2022;
    while (*source < sourceLimit) {
        c = *(*source)++;
        _this->toUBytes[_this->toULength++] = static_cast<uint8_t>(c);
        value = getKey_2022(c, reinterpret_cast<int32_t*>(&key), &offset);

        switch (value){

        case VALID_NON_TERMINAL_2022 :
            /* continue with the loop */
            break;

        case VALID_TERMINAL_2022:
            key = 0;
            goto DONE;

        case INVALID_2022:
            goto DONE;

        case VALID_MAYBE_TERMINAL_2022:
#ifdef U_ENABLE_GENERIC_ISO_2022
            /* ESC ( B is ambiguous only for ISO_2022 itself */
            if(var == ISO_2022) {
                /* discard toUBytes[] for ESC ( B because this sequence is correct and complete */
                _this->toULength = 0;

                /* TODO need to indicate that ESC ( B was seen; if failure, then need to replay from source or from MBCS-style replay */

                /* continue with the loop */
                value = VALID_NON_TERMINAL_2022;
                break;
            } else
#endif
            {
                /* not ISO_2022 itself, finish here */
                value = VALID_TERMINAL_2022;
                key = 0;
                goto DONE;
            }
        }
    }

DONE:
    myData2022->key = key;

    if (value == VALID_NON_TERMINAL_2022) {
        /* indicate that the escape sequence is incomplete: key!=0 */
        return;
    } else if (value == INVALID_2022 ) {
        *err = U_ILLEGAL_ESCAPE_SEQUENCE;
    } else /* value == VALID_TERMINAL_2022 */ {
        switch(var){
#ifdef U_ENABLE_GENERIC_ISO_2022
        case ISO_2022:
        {
            const char *chosenConverterName = escSeqStateTable_Result_2022[offset];
            if(chosenConverterName == nullptr) {
                /* SS2 or SS3 */
                *err = U_UNSUPPORTED_ESCAPE_SEQUENCE;
                _this->toUCallbackReason = UCNV_UNASSIGNED;
                return;
            }

            _this->mode = UCNV_SI;
            ucnv_close(myData2022->currentConverter);
            myData2022->currentConverter = myUConverter = ucnv_open(chosenConverterName, err);
            if(U_SUCCESS(*err)) {
                myUConverter->fromCharErrorBehaviour = UCNV_TO_U_CALLBACK_STOP;
                _this->mode = UCNV_SO;
            }
            break;
        }
#endif
        case ISO_2022_JP:
            {
                StateEnum tempState = static_cast<StateEnum>(nextStateToUnicodeJP[offset]);
                switch(tempState) {
                case INVALID_STATE:
                    *err = U_UNSUPPORTED_ESCAPE_SEQUENCE;
                    break;
                case SS2_STATE:
                    if(myData2022->toU2022State.cs[2]!=0) {
                        if(myData2022->toU2022State.g<2) {
                            myData2022->toU2022State.prevG=myData2022->toU2022State.g;
                        }
                        myData2022->toU2022State.g=2;
                    } else {
                        /* illegal to have SS2 before a matching designator */
                        *err = U_ILLEGAL_ESCAPE_SEQUENCE;
                    }
                    break;
                /* case SS3_STATE: not used in ISO-2022-JP-x */
                case ISO8859_1:
                case ISO8859_7:
                    if((jpCharsetMasks[myData2022->version] & CSM(tempState)) == 0) {
                        *err = U_UNSUPPORTED_ESCAPE_SEQUENCE;
                    } else {
                        /* G2 charset for SS2 */
                        myData2022->toU2022State.cs[2] = static_cast<int8_t>(tempState);
                    }
                    break;
                default:
                    if((jpCharsetMasks[myData2022->version] & CSM(tempState)) == 0) {
                        *err = U_UNSUPPORTED_ESCAPE_SEQUENCE;
                    } else {
                        /* G0 charset */
                        myData2022->toU2022State.cs[0] = static_cast<int8_t>(tempState);
                    }
                    break;
                }
            }
            break;
#if !UCONFIG_ONLY_HTML_CONVERSION
        case ISO_2022_CN:
            {
                StateEnum tempState = static_cast<StateEnum>(nextStateToUnicodeCN[offset]);
                switch(tempState) {
                case INVALID_STATE:
                    *err = U_UNSUPPORTED_ESCAPE_SEQUENCE;
                    break;
                case SS2_STATE:
                    if(myData2022->toU2022State.cs[2]!=0) {
                        if(myData2022->toU2022State.g<2) {
                            myData2022->toU2022State.prevG=myData2022->toU2022State.g;
                        }
                        myData2022->toU2022State.g=2;
                    } else {
                        /* illegal to have SS2 before a matching designator */
                        *err = U_ILLEGAL_ESCAPE_SEQUENCE;
                    }
                    break;
                case SS3_STATE:
                    if(myData2022->toU2022State.cs[3]!=0) {
                        if(myData2022->toU2022State.g<2) {
                            myData2022->toU2022State.prevG=myData2022->toU2022State.g;
                        }
                        myData2022->toU2022State.g=3;
                    } else {
                        /* illegal to have SS3 before a matching designator */
                        *err = U_ILLEGAL_ESCAPE_SEQUENCE;
                    }
                    break;
                case ISO_IR_165:
                    if(myData2022->version==0) {
                        *err = U_UNSUPPORTED_ESCAPE_SEQUENCE;
                        break;
                    }
                    U_FALLTHROUGH;
                case GB2312_1:
                    U_FALLTHROUGH;
                case CNS_11643_1:
                    myData2022->toU2022State.cs[1] = static_cast<int8_t>(tempState);
                    break;
                case CNS_11643_2:
                    myData2022->toU2022State.cs[2] = static_cast<int8_t>(tempState);
                    break;
                default:
                    /* other CNS 11643 planes */
                    if(myData2022->version==0) {
                        *err = U_UNSUPPORTED_ESCAPE_SEQUENCE;
                    } else {
                        myData2022->toU2022State.cs[3] = static_cast<int8_t>(tempState);
                    }
                    break;
                }
            }
            break;
        case ISO_2022_KR:
            if(offset==0x30){
                /* nothing to be done, just accept this one escape sequence */
            } else {
                *err = U_UNSUPPORTED_ESCAPE_SEQUENCE;
            }
            break;
#endif  // !UCONFIG_ONLY_HTML_CONVERSION

        default:
            *err = U_ILLEGAL_ESCAPE_SEQUENCE;
            break;
        }
    }
    if(U_SUCCESS(*err)) {
        _this->toULength = 0;
    } else if(*err==U_ILLEGAL_ESCAPE_SEQUENCE) {
        if(_this->toULength>1) {
            /*
             * Ticket 5691: consistent illegal sequences:
             * - We include at least the first byte (ESC) in the illegal sequence.
             * - If any of the non-initial bytes could be the start of a character,
             *   we stop the illegal sequence before the first one of those.
             *   In escape sequences, all following bytes are "printable", that is,
             *   unless they are completely illegal (>7f in SBCS, outside 21..7e in DBCS),
             *   they are valid single/lead bytes.
             *   For simplicity, we always only report the initial ESC byte as the
             *   illegal sequence and back out all other bytes we looked at.
             */
            /* Back out some bytes. */
            int8_t backOutDistance=_this->toULength-1;
            int8_t bytesFromThisBuffer=_this->toULength-initialToULength;
            if(backOutDistance<=bytesFromThisBuffer) {
                /* same as initialToULength<=1 */
                *source-=backOutDistance;
            } else {
                /* Back out bytes from the previous buffer: Need to replay them. */
                _this->preToULength = static_cast<int8_t>(bytesFromThisBuffer - backOutDistance);
                /* same as -(initialToULength-1) */
                /* preToULength is negative! */
                uprv_memcpy(_this->preToU, _this->toUBytes+1, -_this->preToULength);
                *source-=bytesFromThisBuffer;
            }
            _this->toULength=1;
        }
    } else if(*err==U_UNSUPPORTED_ESCAPE_SEQUENCE) {
        _this->toUCallbackReason = UCNV_UNASSIGNED;
    }
}

#if !UCONFIG_ONLY_HTML_CONVERSION
/*Checks the characters of the buffer against valid 2022 escape sequences
*if the match we return a pointer to the initial start of the sequence otherwise
*we return sourceLimit
*/
/*for 2022 looks ahead in the stream
 *to determine the longest possible convertible
 *data stream
 */
static inline const char*
getEndOfBuffer_2022(const char** source,
                   const char* sourceLimit,
                   UBool /*flush*/){

    const char* mySource = *source;

#ifdef U_ENABLE_GENERIC_ISO_2022
    if (*source >= sourceLimit)
        return sourceLimit;

    do{

        if (*mySource == ESC_2022){
            int8_t i;
            int32_t key = 0;
            int32_t offset;
            UCNV_TableStates_2022 value = VALID_NON_TERMINAL_2022;

            /* Kludge: I could not
            * figure out the reason for validating an escape sequence
            * twice - once here and once in changeState_2022().
            * is it possible to have an ESC character in a ISO2022
            * byte stream which is valid in a code page? Is it legal?
            */
            for (i=0;
            (mySource+i < sourceLimit)&&(value == VALID_NON_TERMINAL_2022);
            i++) {
                value =  getKey_2022(*(mySource+i), &key, &offset);
            }
            if (value > 0 || *mySource==ESC_2022)
                return mySource;

            if ((value == VALID_NON_TERMINAL_2022)&&(!flush) )
                return sourceLimit;
        }
    }while (++mySource < sourceLimit);

    return sourceLimit;
#else
    while(mySource < sourceLimit && *mySource != ESC_2022) {
        ++mySource;
    }
    return mySource;
#endif
}
#endif

/* This inline function replicates code in _MBCSFromUChar32() function in ucnvmbcs.c
 * any future change in _MBCSFromUChar32() function should be reflected here.
 * @return number of bytes in *value; negative number if fallback; 0 if no mapping
 */
static inline int32_t
MBCS_FROM_UCHAR32_ISO2022(UConverterSharedData* sharedData,
                                         UChar32 c,
                                         uint32_t* value,
                                         UBool useFallback,
                                         int outputType)
{
    const int32_t *cx;
    const uint16_t *table;
    uint32_t stage2Entry;
    uint32_t myValue;
    int32_t length;
    const uint8_t *p;
    /*
     * TODO(markus): Use and require new, faster MBCS conversion table structures.
     * Use internal version of ucnv_open() that verifies that the new structures are available,
     * else U_INTERNAL_PROGRAM_ERROR.
     */
    /* BMP-only codepages are stored without stage 1 entries for supplementary code points */
    if(c<0x10000 || (sharedData->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY)) {
        table=sharedData->mbcs.fromUnicodeTable;
        stage2Entry=MBCS_STAGE_2_FROM_U(table, c);
        /* get the bytes and the length for the output */
        if(outputType==MBCS_OUTPUT_2){
            myValue=MBCS_VALUE_2_FROM_STAGE_2(sharedData->mbcs.fromUnicodeBytes, stage2Entry, c);
            if(myValue<=0xff) {
                length=1;
            } else {
                length=2;
            }
        } else /* outputType==MBCS_OUTPUT_3 */ {
            p=MBCS_POINTER_3_FROM_STAGE_2(sharedData->mbcs.fromUnicodeBytes, stage2Entry, c);
            myValue = (static_cast<uint32_t>(*p) << 16) | (static_cast<uint32_t>(p[1]) << 8) | p[2];
            if(myValue<=0xff) {
                length=1;
            } else if(myValue<=0xffff) {
                length=2;
            } else {
                length=3;
            }
        }
        /* is this code point assigned, or do we use fallbacks? */
        if((stage2Entry&(1<<(16+(c&0xf))))!=0) {
            /* assigned */
            *value=myValue;
            return length;
        } else if(FROM_U_USE_FALLBACK(useFallback, c) && myValue!=0) {
            /*
             * We allow a 0 byte output if the "assigned" bit is set for this entry.
             * There is no way with this data structure for fallback output
             * to be a zero byte.
             */
            *value=myValue;
            return -length;
        }
    }

    cx=sharedData->mbcs.extIndexes;
    if(cx!=nullptr) {
        return ucnv_extSimpleMatchFromU(cx, c, value, useFallback);
    }

    /* unassigned */
    return 0;
}

/* This inline function replicates code in _MBCSSingleFromUChar32() function in ucnvmbcs.c
 * any future change in _MBCSSingleFromUChar32() function should be reflected here.
 * @param retval pointer to output byte
 * @return 1 roundtrip byte  0 no mapping  -1 fallback byte
 */
static inline int32_t
MBCS_SINGLE_FROM_UCHAR32(UConverterSharedData* sharedData,
                                       UChar32 c,
                                       uint32_t* retval,
                                       UBool useFallback)
{
    const uint16_t *table;
    int32_t value;
    /* BMP-only codepages are stored without stage 1 entries for supplementary code points */
    if(c>=0x10000 && !(sharedData->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY)) {
        return 0;
    }
    /* convert the Unicode code point in c into codepage bytes (same as in _MBCSFromUnicodeWithOffsets) */
    table=sharedData->mbcs.fromUnicodeTable;
    /* get the byte for the output */
    value=MBCS_SINGLE_RESULT_FROM_U(table, (uint16_t *)sharedData->mbcs.fromUnicodeBytes, c);
    /* is this code point assigned, or do we use fallbacks? */
    *retval = static_cast<uint32_t>(value & 0xff);
    if(value>=0xf00) {
        return 1;  /* roundtrip */
    } else if(useFallback ? value>=0x800 : value>=0xc00) {
        return -1;  /* fallback taken */
    } else {
        return 0;  /* no mapping */
    }
}

/*
 * Check that the result is a 2-byte value with each byte in the range A1..FE
 * (strict EUC DBCS) before accepting it and subtracting 0x80 from each byte
 * to move it to the ISO 2022 range 21..7E.
 * Return 0 if out of range.
 */
static inline uint32_t
_2022FromGR94DBCS(uint32_t value) {
    if (static_cast<uint16_t>(value - 0xa1a1) <= (0xfefe - 0xa1a1) &&
        static_cast<uint8_t>(value - 0xa1) <= (0xfe - 0xa1)
    ) {
        return value - 0x8080;  /* shift down to 21..7e byte range */
    } else {
        return 0;  /* not valid for ISO 2022 */
    }
}

#if 0 /* 5691: Call sites now check for validity. They can just += 0x8080 after that. */
/*
 * This method does the reverse of _2022FromGR94DBCS(). Given the 2022 code point, it returns the
 * 2 byte value that is in the range A1..FE for each byte. Otherwise it returns the 2022 code point
 * unchanged. 
 */
static inline uint32_t
_2022ToGR94DBCS(uint32_t value) {
    uint32_t returnValue = value + 0x8080;
    if( (uint16_t)(returnValue - 0xa1a1) <= (0xfefe - 0xa1a1) &&
        (uint8_t)(returnValue - 0xa1) <= (0xfe - 0xa1)) {
        return returnValue;
    } else {
        return value;
    }
}
#endif

#ifdef U_ENABLE_GENERIC_ISO_2022

/**********************************************************************************
*  ISO-2022 Converter
*
*
*/

static void U_CALLCONV
T_UConverter_toUnicode_ISO_2022_OFFSETS_LOGIC(UConverterToUnicodeArgs* args,
                                                           UErrorCode* err){
    const char* mySourceLimit, *realSourceLimit;
    const char* sourceStart;
    const char16_t* myTargetStart;
    UConverter* saveThis;
    UConverterDataISO2022* myData;
    int8_t length;

    saveThis = args->converter;
    myData=((UConverterDataISO2022*)(saveThis->extraInfo));

    realSourceLimit = args->sourceLimit;
    while (args->source < realSourceLimit) {
        if(myData->key == 0) { /* are we in the middle of an escape sequence? */
            /*Find the end of the buffer e.g : Next Escape Seq | end of Buffer*/
            mySourceLimit = getEndOfBuffer_2022(&(args->source), realSourceLimit, args->flush);

            if(args->source < mySourceLimit) {
                if(myData->currentConverter==nullptr) {
                    myData->currentConverter = ucnv_open("ASCII",err);
                    if(U_FAILURE(*err)){
                        return;
                    }

                    myData->currentConverter->fromCharErrorBehaviour = UCNV_TO_U_CALLBACK_STOP;
                    saveThis->mode = UCNV_SO;
                }

                /* convert to before the ESC or until the end of the buffer */
                myData->isFirstBuffer=false;
                sourceStart = args->source;
                myTargetStart = args->target;
                args->converter = myData->currentConverter;
                ucnv_toUnicode(args->converter,
                    &args->target,
                    args->targetLimit,
                    &args->source,
                    mySourceLimit,
                    args->offsets,
                    (UBool)(args->flush && mySourceLimit == realSourceLimit),
                    err);
                args->converter = saveThis;

                if (*err == U_BUFFER_OVERFLOW_ERROR) {
                    /* move the overflow buffer */
                    length = saveThis->UCharErrorBufferLength = myData->currentConverter->UCharErrorBufferLength;
                    myData->currentConverter->UCharErrorBufferLength = 0;
                    if(length > 0) {
                        uprv_memcpy(saveThis->UCharErrorBuffer,
                                    myData->currentConverter->UCharErrorBuffer,
                                    length*U_SIZEOF_UCHAR);
                    }
                    return;
                }

                /*
                 * At least one of:
                 * -Error while converting
                 * -Done with entire buffer
                 * -Need to write offsets or update the current offset
                 *  (leave that up to the code in ucnv.c)
                 *
                 * or else we just stopped at an ESC byte and continue with changeState_2022()
                 */
                if (U_FAILURE(*err) ||
                    (args->source == realSourceLimit) ||
                    (args->offsets != nullptr && (args->target != myTargetStart || args->source != sourceStart) ||
                    (mySourceLimit < realSourceLimit && myData->currentConverter->toULength > 0))
                ) {
                    /* copy partial or error input for truncated detection and error handling */
                    if(U_FAILURE(*err)) {
                        length = saveThis->invalidCharLength = myData->currentConverter->invalidCharLength;
                        if(length > 0) {
                            uprv_memcpy(saveThis->invalidCharBuffer, myData->currentConverter->invalidCharBuffer, length);
                        }
                    } else {
                        length = saveThis->toULength = myData->currentConverter->toULength;
                        if(length > 0) {
                            uprv_memcpy(saveThis->toUBytes, myData->currentConverter->toUBytes, length);
                            if(args->source < mySourceLimit) {
                                *err = U_TRUNCATED_CHAR_FOUND; /* truncated input before ESC */
                            }
                        }
                    }
                    return;
                }
            }
        }

        sourceStart = args->source;
        changeState_2022(args->converter,
               &(args->source),
               realSourceLimit,
               ISO_2022,
               err);
        if (U_FAILURE(*err) || (args->source != sourceStart && args->offsets != nullptr)) {
            /* let the ucnv.c code update its current offset */
            return;
        }
    }
}

#endif

/*
 * To Unicode Callback helper function
 */
static void
toUnicodeCallback(UConverter *cnv,
                  const uint32_t sourceChar, const uint32_t targetUniChar,
                  UErrorCode* err){
    if(sourceChar>0xff){
        cnv->toUBytes[0] = static_cast<uint8_t>(sourceChar >> 8);
        cnv->toUBytes[1] = static_cast<uint8_t>(sourceChar);
        cnv->toULength = 2;
    }
    else{
        cnv->toUBytes[0] = static_cast<char>(sourceChar);
        cnv->toULength = 1;
    }

    if(targetUniChar == (missingCharMarker-1/*0xfffe*/)){
        *err = U_INVALID_CHAR_FOUND;
    }
    else{
        *err = U_ILLEGAL_CHAR_FOUND;
    }
}

/**************************************ISO-2022-JP*************************************************/

/************************************** IMPORTANT **************************************************
* The UConverter_fromUnicode_ISO2022_JP converter does not use ucnv_fromUnicode() functions for SBCS,DBCS and
* MBCS; instead, the values are obtained directly by calling _MBCSFromUChar32().
* The converter iterates over each Unicode codepoint
* to obtain the equivalent codepoints from the codepages supported. Since the source buffer is
* processed one char at a time it would make sense to reduce the extra processing a canned converter
* would do as far as possible.
*
* If the implementation of these macros or structure of sharedData struct change in the future, make
* sure that ISO-2022 is also changed.
***************************************************************************************************
*/

/***************************************************************************************************
* Rules for ISO-2022-jp encoding
* (i)   Escape sequences must be fully contained within a line they should not
*       span new lines or CRs
* (ii)  If the last character on a line is represented by two bytes then an ASCII or
*       JIS-Roman character escape sequence should follow before the line terminates
* (iii) If the first character on the line is represented by two bytes then a two
*       byte character escape sequence should precede it
* (iv)  If no escape sequence is encountered then the characters are ASCII
* (v)   Latin(ISO-8859-1) and Greek(ISO-8859-7) characters must be designated to G2,
*       and invoked with SS2 (ESC N).
* (vi)  If there is any G0 designation in text, there must be a switch to
*       ASCII or to JIS X 0201-Roman before a space character (but not
*       necessarily before "ESC 4/14 2/0" or "ESC N ' '") or control
*       characters such as tab or CRLF.
* (vi)  Supported encodings:
*          ASCII, JISX201, JISX208, JISX212, GB2312, KSC5601, ISO-8859-1,ISO-8859-7
*
*  source : RFC-1554
*
*          JISX201, JISX208,JISX212 : new .cnv data files created
*          KSC5601 : alias to ibm-949 mapping table
*          GB2312 : alias to ibm-1386 mapping table
*          ISO-8859-1 : Algorithmic implemented as LATIN1 case
*          ISO-8859-7 : alias to ibm-9409 mapping table
*/

/* preference order of JP charsets */
static const StateEnum jpCharsetPref[]={
    ASCII,
    JISX201,
    ISO8859_1,
    JISX208,
    ISO8859_7,
    JISX212,
    GB2312,
    KSC5601,
    HWKANA_7BIT
};

/*
 * The escape sequences must be in order of the enum constants like JISX201  = 3,
 * not in order of jpCharsetPref[]!
 */
static const char escSeqChars[][6] ={
    "\x1B\x28\x42",         /* <ESC>(B  ASCII       */
    "\x1B\x2E\x41",         /* <ESC>.A  ISO-8859-1  */
    "\x1B\x2E\x46",         /* <ESC>.F  ISO-8859-7  */
    "\x1B\x28\x4A",         /* <ESC>(J  JISX-201    */
    "\x1B\x24\x42",         /* <ESC>$B  JISX-208    */
    "\x1B\x24\x28\x44",     /* <ESC>$(D JISX-212    */
    "\x1B\x24\x41",         /* <ESC>$A  GB2312      */
    "\x1B\x24\x28\x43",     /* <ESC>$(C KSC5601     */
    "\x1B\x28\x49"          /* <ESC>(I  HWKANA_7BIT */

};
static  const int8_t escSeqCharsLen[] ={
    3, /* length of <ESC>(B  ASCII       */
    3, /* length of <ESC>.A  ISO-8859-1  */
    3, /* length of <ESC>.F  ISO-8859-7  */
    3, /* length of <ESC>(J  JISX-201    */
    3, /* length of <ESC>$B  JISX-208    */
    4, /* length of <ESC>$(D JISX-212    */
    3, /* length of <ESC>$A  GB2312      */
    4, /* length of <ESC>$(C KSC5601     */
    3  /* length of <ESC>(I  HWKANA_7BIT */
};

/*
* The iteration over various code pages works this way:
* i)   Get the currentState from myConverterData->currentState
* ii)  Check if the character is mapped to a valid character in the currentState
*      Yes ->  a) set the initIterState to currentState
*       b) remain in this state until an invalid character is found
*      No  ->  a) go to the next code page and find the character
* iii) Before changing the state increment the current state check if the current state
*      is equal to the intitIteration state
*      Yes ->  A character that cannot be represented in any of the supported encodings
*       break and return a U_INVALID_CHARACTER error
*      No  ->  Continue and find the character in next code page
*
*
* TODO: Implement a priority technique where the users are allowed to set the priority of code pages
*/

/* Map 00..7F to Unicode according to JIS X 0201. */
static inline uint32_t
jisx201ToU(uint32_t value) {
    if(value < 0x5c) {
        return value;
    } else if(value == 0x5c) {
        return 0xa5;
    } else if(value == 0x7e) {
        return 0x203e;
    } else /* value <= 0x7f */ {
        return value;
    }
}

/* Map Unicode to 00..7F according to JIS X 0201. Return U+FFFE if unmappable. */
static inline uint32_t
jisx201FromU(uint32_t value) {
    if(value<=0x7f) {
        if(value!=0x5c && value!=0x7e) {
            return value;
        }
    } else if(value==0xa5) {
        return 0x5c;
    } else if(value==0x203e) {
        return 0x7e;
    }
    return 0xfffe;
}

/*
 * Take a valid Shift-JIS byte pair, check that it is in the range corresponding
 * to JIS X 0208, and convert it to a pair of 21..7E bytes.
 * Return 0 if the byte pair is out of range.
 */
static inline uint32_t
_2022FromSJIS(uint32_t value) {
    uint8_t trail;

    if(value > 0xEFFC) {
        return 0;  /* beyond JIS X 0208 */
    }

    trail = static_cast<uint8_t>(value);

    value &= 0xff00;  /* lead byte */
    if(value <= 0x9f00) {
        value -= 0x7000;
    } else /* 0xe000 <= value <= 0xef00 */ {
        value -= 0xb000;
    }
    value <<= 1;

    if(trail <= 0x9e) {
        value -= 0x100;
        if(trail <= 0x7e) {
            value |= trail - 0x1f;
        } else {
            value |= trail - 0x20;
        }
    } else /* trail <= 0xfc */ {
        value |= trail - 0x7e;
    }
    return value;
}

/*
 * Convert a pair of JIS X 0208 21..7E bytes to Shift-JIS.
 * If either byte is outside 21..7E make sure that the result is not valid
 * for Shift-JIS so that the converter catches it.
 * Some invalid byte values already turn into equally invalid Shift-JIS
 * byte values and need not be tested explicitly.
 */
static inline void
_2022ToSJIS(uint8_t c1, uint8_t c2, char bytes[2]) {
    if(c1&1) {
        ++c1;
        if(c2 <= 0x5f) {
            c2 += 0x1f;
        } else if(c2 <= 0x7e) {
            c2 += 0x20;
        } else {
            c2 = 0;  /* invalid */
        }
    } else {
        if (static_cast<uint8_t>(c2 - 0x21) <= ((0x7e) - 0x21)) {
            c2 += 0x7e;
        } else {
            c2 = 0;  /* invalid */
        }
    }
    c1 >>= 1;
    if(c1 <= 0x2f) {
        c1 += 0x70;
    } else if(c1 <= 0x3f) {
        c1 += 0xb0;
    } else {
        c1 = 0;  /* invalid */
    }
    bytes[0] = static_cast<char>(c1);
    bytes[1] = static_cast<char>(c2);
}

/*
 * JIS X 0208 has fallbacks from Unicode half-width Katakana to full-width (DBCS)
 * Katakana.
 * Now that we use a Shift-JIS table for JIS X 0208 we need to hardcode these fallbacks
 * because Shift-JIS roundtrips half-width Katakana to single bytes.
 * These were the only fallbacks in ICU's jisx-208.ucm file.
 */
static const uint16_t hwkana_fb[HWKANA_END - HWKANA_START + 1] = {
    0x2123,  /* U+FF61 */
    0x2156,
    0x2157,
    0x2122,
    0x2126,
    0x2572,
    0x2521,
    0x2523,
    0x2525,
    0x2527,
    0x2529,
    0x2563,
    0x2565,
    0x2567,
    0x2543,
    0x213C,  /* U+FF70 */
    0x2522,
    0x2524,
    0x2526,
    0x2528,
    0x252A,
    0x252B,
    0x252D,
    0x252F,
    0x2531,
    0x2533,
    0x2535,
    0x2537,
    0x2539,
    0x253B,
    0x253D,
    0x253F,  /* U+FF80 */
    0x2541,
    0x2544,
    0x2546,
    0x2548,
    0x254A,
    0x254B,
    0x254C,
    0x254D,
    0x254E,
    0x254F,
    0x2552,
    0x2555,
    0x2558,
    0x255B,
    0x255E,
    0x255F,  /* U+FF90 */
    0x2560,
    0x2561,
    0x2562,
    0x2564,
    0x2566,
    0x2568,
    0x2569,
    0x256A,
    0x256B,
    0x256C,
    0x256D,
    0x256F,
    0x2573,
    0x212B,
    0x212C   /* U+FF9F */
};

static void U_CALLCONV
UConverter_fromUnicode_ISO_2022_JP_OFFSETS_LOGIC(UConverterFromUnicodeArgs* args, UErrorCode* err) {
    UConverter *cnv = args->converter;
    UConverterDataISO2022 *converterData;
    ISO2022State *pFromU2022State;
    uint8_t* target = reinterpret_cast<uint8_t*>(args->target);
    const uint8_t* targetLimit = reinterpret_cast<const uint8_t*>(args->targetLimit);
    const char16_t* source = args->source;
    const char16_t* sourceLimit = args->sourceLimit;
    int32_t* offsets = args->offsets;
    UChar32 sourceChar;
    char buffer[8];
    int32_t len, outLen;
    int8_t choices[10];
    int32_t choiceCount;
    uint32_t targetValue = 0;
    UBool useFallback;

    int32_t i;
    int8_t cs, g;

    /* set up the state */
    converterData = static_cast<UConverterDataISO2022*>(cnv->extraInfo);
    pFromU2022State   = &converterData->fromU2022State;

    choiceCount = 0;

    /* check if the last codepoint of previous buffer was a lead surrogate*/
    if((sourceChar = cnv->fromUChar32)!=0 && target< targetLimit) {
        goto getTrail;
    }

    while(source < sourceLimit) {
        if(target < targetLimit) {

            sourceChar  = *(source++);
            /*check if the char is a First surrogate*/
            if(U16_IS_SURROGATE(sourceChar)) {
                if(U16_IS_SURROGATE_LEAD(sourceChar)) {
getTrail:
                    /*look ahead to find the trail surrogate*/
                    if(source < sourceLimit) {
                        /* test the following code unit */
                        char16_t trail = *source;
                        if(U16_IS_TRAIL(trail)) {
                            source++;
                            sourceChar=U16_GET_SUPPLEMENTARY(sourceChar, trail);
                            cnv->fromUChar32=0x00;
                            /* convert this supplementary code point */
                            /* exit this condition tree */
                        } else {
                            /* this is an unmatched lead code unit (1st surrogate) */
                            /* callback(illegal) */
                            *err=U_ILLEGAL_CHAR_FOUND;
                            cnv->fromUChar32=sourceChar;
                            break;
                        }
                    } else {
                        /* no more input */
                        cnv->fromUChar32=sourceChar;
                        break;
                    }
                } else {
                    /* this is an unmatched trail code unit (2nd surrogate) */
                    /* callback(illegal) */
                    *err=U_ILLEGAL_CHAR_FOUND;
                    cnv->fromUChar32=sourceChar;
                    break;
                }
            }

            /* do not convert SO/SI/ESC */
            if(IS_2022_CONTROL(sourceChar)) {
                /* callback(illegal) */
                *err=U_ILLEGAL_CHAR_FOUND;
                cnv->fromUChar32=sourceChar;
                break;
            }

            /* do the conversion */

            if(choiceCount == 0) {
                uint16_t csm;

                /*
                 * The csm variable keeps track of which charsets are allowed
                 * and not used yet while building the choices[].
                 */
                csm = jpCharsetMasks[converterData->version];
                choiceCount = 0;

                /* JIS7/8: try single-byte half-width Katakana before JISX208 */
                if(converterData->version == 3 || converterData->version == 4) {
                    choices[choiceCount++] = static_cast<int8_t>(HWKANA_7BIT);
                }
                /* Do not try single-byte half-width Katakana for other versions. */
                csm &= ~CSM(HWKANA_7BIT);

                /* try the current G0 charset */
                choices[choiceCount++] = cs = pFromU2022State->cs[0];
                csm &= ~CSM(cs);

                /* try the current G2 charset */
                if((cs = pFromU2022State->cs[2]) != 0) {
                    choices[choiceCount++] = cs;
                    csm &= ~CSM(cs);
                }

                /* try all the other possible charsets */
                for(i = 0; i < UPRV_LENGTHOF(jpCharsetPref); ++i) {
                    cs = static_cast<int8_t>(jpCharsetPref[i]);
                    if(CSM(cs) & csm) {
                        choices[choiceCount++] = cs;
                        csm &= ~CSM(cs);
                    }
                }
            }

            cs = g = 0;
            /*
             * len==0: no mapping found yet
             * len<0: found a fallback result: continue looking for a roundtrip but no further fallbacks
             * len>0: found a roundtrip result, done
             */
            len = 0;
            /*
             * We will turn off useFallback after finding a fallback,
             * but we still get fallbacks from PUA code points as usual.
             * Therefore, we will also need to check that we don't overwrite
             * an early fallback with a later one.
             */
            useFallback = cnv->useFallback;

            for(i = 0; i < choiceCount && len <= 0; ++i) {
                uint32_t value;
                int32_t len2;
                int8_t cs0 = choices[i];
                switch(cs0) {
                case ASCII:
                    if(sourceChar <= 0x7f) {
                        targetValue = static_cast<uint32_t>(sourceChar);
                        len = 1;
                        cs = cs0;
                        g = 0;
                    }
                    break;
                case ISO8859_1:
                    if(GR96_START <= sourceChar && sourceChar <= GR96_END) {
                        targetValue = static_cast<uint32_t>(sourceChar) - 0x80;
                        len = 1;
                        cs = cs0;
                        g = 2;
                    }
                    break;
                case HWKANA_7BIT:
                    if (static_cast<uint32_t>(sourceChar - HWKANA_START) <= (HWKANA_END - HWKANA_START)) {
                        if(converterData->version==3) {
                            /* JIS7: use G1 (SO) */
                            /* Shift U+FF61..U+FF9F to bytes 21..5F. */
                            targetValue = static_cast<uint32_t>(sourceChar - (HWKANA_START - 0x21));
                            len = 1;
                            pFromU2022State->cs[1] = cs = cs0; /* do not output an escape sequence */
                            g = 1;
                        } else if(converterData->version==4) {
                            /* JIS8: use 8-bit bytes with any single-byte charset, see escape sequence output below */
                            /* Shift U+FF61..U+FF9F to bytes A1..DF. */
                            targetValue = static_cast<uint32_t>(sourceChar - (HWKANA_START - 0xa1));
                            len = 1;

                            cs = pFromU2022State->cs[0];
                            if(IS_JP_DBCS(cs)) {
                                /* switch from a DBCS charset to JISX201 */
                                cs = static_cast<int8_t>(JISX201);
                            }
                            /* else stay in the current G0 charset */
                            g = 0;
                        }
                        /* else do not use HWKANA_7BIT with other versions */
                    }
                    break;
                case JISX201:
                    /* G0 SBCS */
                    value = jisx201FromU(sourceChar);
                    if(value <= 0x7f) {
                        targetValue = value;
                        len = 1;
                        cs = cs0;
                        g = 0;
                        useFallback = false;
                    }
                    break;
                case JISX208:
                    /* G0 DBCS from Shift-JIS table */
                    len2 = MBCS_FROM_UCHAR32_ISO2022(
                                converterData->myConverterArray[cs0],
                                sourceChar, &value,
                                useFallback, MBCS_OUTPUT_2);
                    if(len2 == 2 || (len2 == -2 && len == 0)) {  /* only accept DBCS: abs(len)==2 */
                        value = _2022FromSJIS(value);
                        if(value != 0) {
                            targetValue = value;
                            len = len2;
                            cs = cs0;
                            g = 0;
                            useFallback = false;
                        }
                    } else if(len == 0 && useFallback &&
                              static_cast<uint32_t>(sourceChar - HWKANA_START) <= (HWKANA_END - HWKANA_START)) {
                        targetValue = hwkana_fb[sourceChar - HWKANA_START];
                        len = -2;
                        cs = cs0;
                        g = 0;
                        useFallback = false;
                    }
                    break;
                case ISO8859_7:
                    /* G0 SBCS forced to 7-bit output */
                    len2 = MBCS_SINGLE_FROM_UCHAR32(
                                converterData->myConverterArray[cs0],
                                sourceChar, &value,
                                useFallback);
                    if(len2 != 0 && !(len2 < 0 && len != 0) && GR96_START <= value && value <= GR96_END) {
                        targetValue = value - 0x80;
                        len = len2;
                        cs = cs0;
                        g = 2;
                        useFallback = false;
                    }
                    break;
                default:
                    /* G0 DBCS */
                    len2 = MBCS_FROM_UCHAR32_ISO2022(
                                converterData->myConverterArray[cs0],
                                sourceChar, &value,
                                useFallback, MBCS_OUTPUT_2);
                    if(len2 == 2 || (len2 == -2 && len == 0)) {  /* only accept DBCS: abs(len)==2 */
                        if(cs0 == KSC5601) {
                            /*
                             * Check for valid bytes for the encoding scheme.
                             * This is necessary because the sub-converter (windows-949)
                             * has a broader encoding scheme than is valid for 2022.
                             */
                            value = _2022FromGR94DBCS(value);
                            if(value == 0) {
                                break;
                            }
                        }
                        targetValue = value;
                        len = len2;
                        cs = cs0;
                        g = 0;
                        useFallback = false;
                    }
                    break;
                }
            }

            if(len != 0) {
                if(len < 0) {
                    len = -len;  /* fallback */
                }
                outLen = 0; /* count output bytes */

                /* write SI if necessary (only for JIS7) */
                if(pFromU2022State->g == 1 && g == 0) {
                    buffer[outLen++] = UCNV_SI;
                    pFromU2022State->g = 0;
                }

                /* write the designation sequence if necessary */
                if(cs != pFromU2022State->cs[g]) {
                    int32_t escLen = escSeqCharsLen[cs];
                    uprv_memcpy(buffer + outLen, escSeqChars[cs], escLen);
                    outLen += escLen;
                    pFromU2022State->cs[g] = cs;

                    /* invalidate the choices[] */
                    choiceCount = 0;
                }

                /* write the shift sequence if necessary */
                if(g != pFromU2022State->g) {
                    switch(g) {
                    /* case 0 handled before writing escapes */
                    case 1:
                        buffer[outLen++] = UCNV_SO;
                        pFromU2022State->g = 1;
                        break;
                    default: /* case 2 */
                        buffer[outLen++] = 0x1b;
                        buffer[outLen++] = 0x4e;
                        break;
                    /* no case 3: no SS3 in ISO-2022-JP-x */
                    }
                }

                /* write the output bytes */
                if(len == 1) {
                    buffer[outLen++] = static_cast<char>(targetValue);
                } else /* len == 2 */ {
                    buffer[outLen++] = static_cast<char>(targetValue >> 8);
                    buffer[outLen++] = static_cast<char>(targetValue);
                }
            } else {
                /*
                 * if we cannot find the character after checking all codepages
                 * then this is an error
                 */
                *err = U_INVALID_CHAR_FOUND;
                cnv->fromUChar32=sourceChar;
                break;
            }

            if(sourceChar == CR || sourceChar == LF) {
                /* reset the G2 state at the end of a line (conversion got us into ASCII or JISX201 already) */
                pFromU2022State->cs[2] = 0;
                choiceCount = 0;
            }

            /* output outLen>0 bytes in buffer[] */
            if(outLen == 1) {
                *target++ = buffer[0];
                if(offsets) {
                    *offsets++ = static_cast<int32_t>(source - args->source - 1); /* -1: known to be ASCII */
                }
            } else if(outLen == 2 && (target + 2) <= targetLimit) {
                *target++ = buffer[0];
                *target++ = buffer[1];
                if(offsets) {
                    int32_t sourceIndex = static_cast<int32_t>(source - args->source - U16_LENGTH(sourceChar));
                    *offsets++ = sourceIndex;
                    *offsets++ = sourceIndex;
                }
            } else {
                fromUWriteUInt8(
                    cnv,
                    buffer, outLen,
                    &target, reinterpret_cast<const char*>(targetLimit),
                    &offsets, static_cast<int32_t>(source - args->source - U16_LENGTH(sourceChar)),
                    err);
                if(U_FAILURE(*err)) {
                    break;
                }
            }
        } /* end if(myTargetIndex<myTargetLength) */
        else{
            *err =U_BUFFER_OVERFLOW_ERROR;
            break;
        }

    }/* end while(mySourceIndex<mySourceLength) */

    /*
     * the end of the input stream and detection of truncated input
     * are handled by the framework, but for ISO-2022-JP conversion
     * we need to be in ASCII mode at the very end
     *
     * conditions:
     *   successful
     *   in SO mode or not in ASCII mode
     *   end of input and no truncated input
     */
    if( U_SUCCESS(*err) &&
        (pFromU2022State->g!=0 || pFromU2022State->cs[0]!=ASCII) &&
        args->flush && source>=sourceLimit && cnv->fromUChar32==0
    ) {
        int32_t sourceIndex;

        outLen = 0;

        if(pFromU2022State->g != 0) {
            buffer[outLen++] = UCNV_SI;
            pFromU2022State->g = 0;
        }

        if(pFromU2022State->cs[0] != ASCII) {
            int32_t escLen = escSeqCharsLen[ASCII];
            uprv_memcpy(buffer + outLen, escSeqChars[ASCII], escLen);
            outLen += escLen;
            pFromU2022State->cs[0] = static_cast<int8_t>(ASCII);
        }

        /* get the source index of the last input character */
        /*
         * TODO this would be simpler and more reliable if we used a pair
         * of sourceIndex/prevSourceIndex like in ucnvmbcs.c
         * so that we could simply use the prevSourceIndex here;
         * this code gives an incorrect result for the rare case of an unmatched
         * trail surrogate that is alone in the last buffer of the text stream
         */
        sourceIndex = static_cast<int32_t>(source - args->source);
        if(sourceIndex>0) {
            --sourceIndex;
            if( U16_IS_TRAIL(args->source[sourceIndex]) &&
                (sourceIndex==0 || U16_IS_LEAD(args->source[sourceIndex-1]))
            ) {
                --sourceIndex;
            }
        } else {
            sourceIndex=-1;
        }

        fromUWriteUInt8(
            cnv,
            buffer, outLen,
            &target, reinterpret_cast<const char*>(targetLimit),
            &offsets, sourceIndex,
            err);
    }

    /*save the state and return */
    args->source = source;
    args->target = reinterpret_cast<char*>(target);
}

/*************** to unicode *******************/

static void U_CALLCONV
UConverter_toUnicode_ISO_2022_JP_OFFSETS_LOGIC(UConverterToUnicodeArgs *args,
                                               UErrorCode* err){
    char tempBuf[2];
    const char* mySource = const_cast<char*>(args->source);
    char16_t *myTarget = args->target;
    const char *mySourceLimit = args->sourceLimit;
    uint32_t targetUniChar = 0x0000;
    uint32_t mySourceChar = 0x0000;
    uint32_t tmpSourceChar = 0x0000;
    UConverterDataISO2022* myData;
    ISO2022State *pToU2022State;
    StateEnum cs;

    myData = static_cast<UConverterDataISO2022*>(args->converter->extraInfo);
    pToU2022State = &myData->toU2022State;

    if(myData->key != 0) {
        /* continue with a partial escape sequence */
        goto escape;
    } else if(args->converter->toULength == 1 && mySource < mySourceLimit && myTarget < args->targetLimit) {
        /* continue with a partial double-byte character */
        mySourceChar = args->converter->toUBytes[0];
        args->converter->toULength = 0;
        cs = static_cast<StateEnum>(pToU2022State->cs[pToU2022State->g]);
        targetUniChar = missingCharMarker;
        goto getTrailByte;
    }

    while(mySource < mySourceLimit){

        targetUniChar =missingCharMarker;

        if(myTarget < args->targetLimit){

            mySourceChar = static_cast<unsigned char>(*mySource++);

            switch(mySourceChar) {
            case UCNV_SI:
                if(myData->version==3) {
                    pToU2022State->g=0;
                    continue;
                } else {
                    /* only JIS7 uses SI/SO, not ISO-2022-JP-x */
                    myData->isEmptySegment = false; /* reset this, we have a different error */
                    break;
                }

            case UCNV_SO:
                if(myData->version==3) {
                    /* JIS7: switch to G1 half-width Katakana */
                    pToU2022State->cs[1] = static_cast<int8_t>(HWKANA_7BIT);
                    pToU2022State->g=1;
                    continue;
                } else {
                    /* only JIS7 uses SI/SO, not ISO-2022-JP-x */
                    myData->isEmptySegment = false; /* reset this, we have a different error */
                    break;
                }

            case ESC_2022:
                mySource--;
escape:
                {
                    const char * mySourceBefore = mySource;
                    int8_t toULengthBefore = args->converter->toULength;

                    changeState_2022(args->converter,&(mySource),
                        mySourceLimit, ISO_2022_JP,err);

                    /* If in ISO-2022-JP only and we successfully completed an escape sequence, but previous segment was empty, create an error */
                    if(myData->version==0 && myData->key==0 && U_SUCCESS(*err) && myData->isEmptySegment) {
                        *err = U_ILLEGAL_ESCAPE_SEQUENCE;
                        args->converter->toUCallbackReason = UCNV_IRREGULAR;
                        args->converter->toULength = static_cast<int8_t>(toULengthBefore + (mySource - mySourceBefore));
                    }
                }

                /* invalid or illegal escape sequence */
                if(U_FAILURE(*err)){
                    args->target = myTarget;
                    args->source = mySource;
                    myData->isEmptySegment = false; /* Reset to avoid future spurious errors */
                    return;
                }
                /* If we successfully completed an escape sequence, we begin a new segment, empty so far */
                if(myData->key==0) {
                    myData->isEmptySegment = true;
                }
                continue;

            /* ISO-2022-JP does not use single-byte (C1) SS2 and SS3 */

            case CR:
            case LF:
                /* automatically reset to single-byte mode */
                if (static_cast<StateEnum>(pToU2022State->cs[0]) != ASCII &&
                    static_cast<StateEnum>(pToU2022State->cs[0]) != JISX201) {
                    pToU2022State->cs[0] = static_cast<int8_t>(ASCII);
                }
                pToU2022State->cs[2] = 0;
                pToU2022State->g = 0;
                U_FALLTHROUGH;
            default:
                /* convert one or two bytes */
                myData->isEmptySegment = false;
                cs = static_cast<StateEnum>(pToU2022State->cs[pToU2022State->g]);
                if (static_cast<uint8_t>(mySourceChar - 0xa1) <= (0xdf - 0xa1) && myData->version == 4 &&
                    !IS_JP_DBCS(cs)
                ) {
                    /* 8-bit halfwidth katakana in any single-byte mode for JIS8 */
                    targetUniChar = mySourceChar + (HWKANA_START - 0xa1);

                    /* return from a single-shift state to the previous one */
                    if(pToU2022State->g >= 2) {
                        pToU2022State->g=pToU2022State->prevG;
                    }
                } else switch(cs) {
                case ASCII:
                    if(mySourceChar <= 0x7f) {
                        targetUniChar = mySourceChar;
                    }
                    break;
                case ISO8859_1:
                    if(mySourceChar <= 0x7f) {
                        targetUniChar = mySourceChar + 0x80;
                    }
                    /* return from a single-shift state to the previous one */
                    pToU2022State->g=pToU2022State->prevG;
                    break;
                case ISO8859_7:
                    if(mySourceChar <= 0x7f) {
                        /* convert mySourceChar+0x80 to use a normal 8-bit table */
                        targetUniChar =
                            _MBCS_SINGLE_SIMPLE_GET_NEXT_BMP(
                                myData->myConverterArray[cs],
                                mySourceChar + 0x80);
                    }
                    /* return from a single-shift state to the previous one */
                    pToU2022State->g=pToU2022State->prevG;
                    break;
                case JISX201:
                    if(mySourceChar <= 0x7f) {
                        targetUniChar = jisx201ToU(mySourceChar);
                    }
                    break;
                case HWKANA_7BIT:
                    if (static_cast<uint8_t>(mySourceChar - 0x21) <= (0x5f - 0x21)) {
                        /* 7-bit halfwidth Katakana */
                        targetUniChar = mySourceChar + (HWKANA_START - 0x21);
                    }
                    break;
                default:
                    /* G0 DBCS */
                    if(mySource < mySourceLimit) {
                        int leadIsOk, trailIsOk;
                        uint8_t trailByte;
getTrailByte:
                        trailByte = static_cast<uint8_t>(*mySource);
                        /*
                         * Ticket 5691: consistent illegal sequences:
                         * - We include at least the first byte in the illegal sequence.
                         * - If any of the non-initial bytes could be the start of a character,
                         *   we stop the illegal sequence before the first one of those.
                         *
                         * In ISO-2022 DBCS, if the second byte is in the 21..7e range or is
                         * an ESC/SO/SI, we report only the first byte as the illegal sequence.
                         * Otherwise we convert or report the pair of bytes.
                         */
                        leadIsOk = static_cast<uint8_t>(mySourceChar - 0x21) <= (0x7e - 0x21);
                        trailIsOk = static_cast<uint8_t>(trailByte - 0x21) <= (0x7e - 0x21);
                        if (leadIsOk && trailIsOk) {
                            ++mySource;
                            tmpSourceChar = (mySourceChar << 8) | trailByte;
                            if(cs == JISX208) {
                                _2022ToSJIS(static_cast<uint8_t>(mySourceChar), trailByte, tempBuf);
                                mySourceChar = tmpSourceChar;
                            } else {
                                /* Copy before we modify tmpSourceChar so toUnicodeCallback() sees the correct bytes. */
                                mySourceChar = tmpSourceChar;
                                if (cs == KSC5601) {
                                    tmpSourceChar += 0x8080;  /* = _2022ToGR94DBCS(tmpSourceChar) */
                                }
                                tempBuf[0] = static_cast<char>(tmpSourceChar >> 8);
                                tempBuf[1] = static_cast<char>(tmpSourceChar);
                            }
                            targetUniChar = ucnv_MBCSSimpleGetNextUChar(myData->myConverterArray[cs], tempBuf, 2, false);
                        } else if (!(trailIsOk || IS_2022_CONTROL(trailByte))) {
                            /* report a pair of illegal bytes if the second byte is not a DBCS starter */
                            ++mySource;
                            /* add another bit so that the code below writes 2 bytes in case of error */
                            mySourceChar = 0x10000 | (mySourceChar << 8) | trailByte;
                        }
                    } else {
                        args->converter->toUBytes[0] = static_cast<uint8_t>(mySourceChar);
                        args->converter->toULength = 1;
                        goto endloop;
                    }
                }  /* End of inner switch */
                break;
            }  /* End of outer switch */
            if(targetUniChar < (missingCharMarker-1/*0xfffe*/)){
                if(args->offsets){
                    args->offsets[myTarget - args->target] = static_cast<int32_t>(mySource - args->source - (mySourceChar <= 0xff ? 1 : 2));
                }
                *(myTarget++) = static_cast<char16_t>(targetUniChar);
            }
            else if(targetUniChar > missingCharMarker){
                /* disassemble the surrogate pair and write to output*/
                targetUniChar-=0x0010000;
                *myTarget = static_cast<char16_t>(0xd800 + static_cast<char16_t>(targetUniChar >> 10));
                if(args->offsets){
                    args->offsets[myTarget - args->target] = static_cast<int32_t>(mySource - args->source - (mySourceChar <= 0xff ? 1 : 2));
                }
                ++myTarget;
                if(myTarget< args->targetLimit){
                    *myTarget = static_cast<char16_t>(0xdc00 + static_cast<char16_t>(targetUniChar & 0x3ff));
                    if(args->offsets){
                        args->offsets[myTarget - args->target] = static_cast<int32_t>(mySource - args->source - (mySourceChar <= 0xff ? 1 : 2));
                    }
                    ++myTarget;
                }else{
                    args->converter->UCharErrorBuffer[args->converter->UCharErrorBufferLength++]=
                                    static_cast<char16_t>(0xdc00 + static_cast<char16_t>(targetUniChar & 0x3ff));
                }

            }
            else{
                /* Call the callback function*/
                toUnicodeCallback(args->converter,mySourceChar,targetUniChar,err);
                break;
            }
        }
        else{    /* goes with "if(myTarget < args->targetLimit)"  way up near top of function */
            *err =U_BUFFER_OVERFLOW_ERROR;
            break;
        }
    }
endloop:
    args->target = myTarget;
    args->source = mySource;
}


#if !UCONFIG_ONLY_HTML_CONVERSION
/***************************************************************
*   Rules for ISO-2022-KR encoding
*   i) The KSC5601 designator sequence should appear only once in a file,
*      at the beginning of a line before any KSC5601 characters. This usually
*      means that it appears by itself on the first line of the file
*  ii) There are only 2 shifting sequences SO to shift into double byte mode
*      and SI to shift into single byte mode
*/
static void U_CALLCONV
UConverter_fromUnicode_ISO_2022_KR_OFFSETS_LOGIC_IBM(UConverterFromUnicodeArgs* args, UErrorCode* err){

    UConverter* saveConv = args->converter;
    UConverterDataISO2022* myConverterData = static_cast<UConverterDataISO2022*>(saveConv->extraInfo);
    args->converter=myConverterData->currentConverter;

    myConverterData->currentConverter->fromUChar32 = saveConv->fromUChar32;
    ucnv_MBCSFromUnicodeWithOffsets(args,err);
    saveConv->fromUChar32 = myConverterData->currentConverter->fromUChar32;

    if(*err == U_BUFFER_OVERFLOW_ERROR) {
        if(myConverterData->currentConverter->charErrorBufferLength > 0) {
            uprv_memcpy(
                saveConv->charErrorBuffer,
                myConverterData->currentConverter->charErrorBuffer,
                myConverterData->currentConverter->charErrorBufferLength);
        }
        saveConv->charErrorBufferLength = myConverterData->currentConverter->charErrorBufferLength;
        myConverterData->currentConverter->charErrorBufferLength = 0;
    }
    args->converter=saveConv;
}

static void U_CALLCONV
UConverter_fromUnicode_ISO_2022_KR_OFFSETS_LOGIC(UConverterFromUnicodeArgs* args, UErrorCode* err){

    const char16_t *source = args->source;
    const char16_t *sourceLimit = args->sourceLimit;
    unsigned char *target = reinterpret_cast<unsigned char*>(args->target);
    unsigned char *targetLimit = reinterpret_cast<unsigned char*>(const_cast<char*>(args->targetLimit));
    int32_t* offsets = args->offsets;
    uint32_t targetByteUnit = 0x0000;
    UChar32 sourceChar = 0x0000;
    UBool isTargetByteDBCS;
    UBool oldIsTargetByteDBCS;
    UConverterDataISO2022 *converterData;
    UConverterSharedData* sharedData;
    UBool useFallback;
    int32_t length =0;

    converterData = static_cast<UConverterDataISO2022*>(args->converter->extraInfo);
    /* if the version is 1 then the user is requesting
     * conversion with ibm-25546 pass the arguments to
     * MBCS converter and return
     */
    if(converterData->version==1){
        UConverter_fromUnicode_ISO_2022_KR_OFFSETS_LOGIC_IBM(args,err);
        return;
    }

    /* initialize data */
    sharedData = converterData->currentConverter->sharedData;
    useFallback = args->converter->useFallback;
    isTargetByteDBCS = static_cast<UBool>(args->converter->fromUnicodeStatus);
    oldIsTargetByteDBCS = isTargetByteDBCS;

    isTargetByteDBCS = static_cast<UBool>(args->converter->fromUnicodeStatus);
    if((sourceChar = args->converter->fromUChar32)!=0 && target <targetLimit) {
        goto getTrail;
    }
    while(source < sourceLimit){

        targetByteUnit = missingCharMarker;

        if(target < (unsigned char*) args->targetLimit){
            sourceChar = *source++;

            /* do not convert SO/SI/ESC */
            if(IS_2022_CONTROL(sourceChar)) {
                /* callback(illegal) */
                *err=U_ILLEGAL_CHAR_FOUND;
                args->converter->fromUChar32=sourceChar;
                break;
            }

            length = MBCS_FROM_UCHAR32_ISO2022(sharedData,sourceChar,&targetByteUnit,useFallback,MBCS_OUTPUT_2);
            if(length < 0) {
                length = -length;  /* fallback */
            }
            /* only DBCS or SBCS characters are expected*/
            /* DB characters with high bit set to 1 are expected */
            if( length > 2 || length==0 ||
                (length == 1 && targetByteUnit > 0x7f) ||
                (length == 2 &&
                    (static_cast<uint16_t>(targetByteUnit - 0xa1a1) > (0xfefe - 0xa1a1) ||
                    static_cast<uint8_t>(targetByteUnit - 0xa1) > (0xfe - 0xa1)))
            ) {
                targetByteUnit=missingCharMarker;
            }
            if (targetByteUnit != missingCharMarker){

                oldIsTargetByteDBCS = isTargetByteDBCS;
                isTargetByteDBCS = static_cast<UBool>(targetByteUnit > 0x00FF);
                  /* append the shift sequence */
                if (oldIsTargetByteDBCS != isTargetByteDBCS ){

                    if (isTargetByteDBCS)
                        *target++ = UCNV_SO;
                    else
                        *target++ = UCNV_SI;
                    if(offsets)
                        *(offsets++) = static_cast<int32_t>(source - args->source - 1);
                }
                /* write the targetUniChar  to target */
                if(targetByteUnit <= 0x00FF){
                    if( target < targetLimit){
                        *(target++) = static_cast<unsigned char>(targetByteUnit);
                        if(offsets){
                            *(offsets++) = static_cast<int32_t>(source - args->source - 1);
                        }

                    }else{
                        args->converter->charErrorBuffer[args->converter->charErrorBufferLength++] = static_cast<unsigned char>(targetByteUnit);
                        *err = U_BUFFER_OVERFLOW_ERROR;
                    }
                }else{
                    if(target < targetLimit){
                        *(target++) = static_cast<unsigned char>((targetByteUnit >> 8) - 0x80);
                        if(offsets){
                            *(offsets++) = static_cast<int32_t>(source - args->source - 1);
                        }
                        if(target < targetLimit){
                            *(target++) = static_cast<unsigned char>(targetByteUnit - 0x80);
                            if(offsets){
                                *(offsets++) = static_cast<int32_t>(source - args->source - 1);
                            }
                        }else{
                            args->converter->charErrorBuffer[args->converter->charErrorBufferLength++] = static_cast<unsigned char>(targetByteUnit - 0x80);
                            *err = U_BUFFER_OVERFLOW_ERROR;
                        }
                    }else{
                        args->converter->charErrorBuffer[args->converter->charErrorBufferLength++] = static_cast<unsigned char>((targetByteUnit >> 8) - 0x80);
                        args->converter->charErrorBuffer[args->converter->charErrorBufferLength++] = static_cast<unsigned char>(targetByteUnit - 0x80);
                        *err = U_BUFFER_OVERFLOW_ERROR;
                    }
                }

            }
            else{
                /* oops.. the code point is unassingned
                 * set the error and reason
                 */

                /*check if the char is a First surrogate*/
                if(U16_IS_SURROGATE(sourceChar)) {
                    if(U16_IS_SURROGATE_LEAD(sourceChar)) {
getTrail:
                        /*look ahead to find the trail surrogate*/
                        if(source <  sourceLimit) {
                            /* test the following code unit */
                            char16_t trail = *source;
                            if(U16_IS_TRAIL(trail)) {
                                source++;
                                sourceChar=U16_GET_SUPPLEMENTARY(sourceChar, trail);
                                *err = U_INVALID_CHAR_FOUND;
                                /* convert this surrogate code point */
                                /* exit this condition tree */
                            } else {
                                /* this is an unmatched lead code unit (1st surrogate) */
                                /* callback(illegal) */
                                *err=U_ILLEGAL_CHAR_FOUND;
                            }
                        } else {
                            /* no more input */
                            *err = U_ZERO_ERROR;
                        }
                    } else {
                        /* this is an unmatched trail code unit (2nd surrogate) */
                        /* callback(illegal) */
                        *err=U_ILLEGAL_CHAR_FOUND;
                    }
                } else {
                    /* callback(unassigned) for a BMP code point */
                    *err = U_INVALID_CHAR_FOUND;
                }

                args->converter->fromUChar32=sourceChar;
                break;
            }
        } /* end if(myTargetIndex<myTargetLength) */
        else{
            *err =U_BUFFER_OVERFLOW_ERROR;
            break;
        }

    }/* end while(mySourceIndex<mySourceLength) */

    /*
     * the end of the input stream and detection of truncated input
     * are handled by the framework, but for ISO-2022-KR conversion
     * we need to be in ASCII mode at the very end
     *
     * conditions:
     *   successful
     *   not in ASCII mode
     *   end of input and no truncated input
     */
    if( U_SUCCESS(*err) &&
        isTargetByteDBCS &&
        args->flush && source>=sourceLimit && args->converter->fromUChar32==0
    ) {
        int32_t sourceIndex;

        /* we are switching to ASCII */
        isTargetByteDBCS=false;

        /* get the source index of the last input character */
        /*
         * TODO this would be simpler and more reliable if we used a pair
         * of sourceIndex/prevSourceIndex like in ucnvmbcs.c
         * so that we could simply use the prevSourceIndex here;
         * this code gives an incorrect result for the rare case of an unmatched
         * trail surrogate that is alone in the last buffer of the text stream
         */
        sourceIndex = static_cast<int32_t>(source - args->source);
        if(sourceIndex>0) {
            --sourceIndex;
            if( U16_IS_TRAIL(args->source[sourceIndex]) &&
                (sourceIndex==0 || U16_IS_LEAD(args->source[sourceIndex-1]))
            ) {
                --sourceIndex;
            }
        } else {
            sourceIndex=-1;
        }

        fromUWriteUInt8(
            args->converter,
            SHIFT_IN_STR, 1,
            &target, reinterpret_cast<const char*>(targetLimit),
            &offsets, sourceIndex,
            err);
    }

    /*save the state and return */
    args->source = source;
    args->target = reinterpret_cast<char*>(target);
    args->converter->fromUnicodeStatus = static_cast<uint32_t>(isTargetByteDBCS);
}

/************************ To Unicode ***************************************/

static void U_CALLCONV
UConverter_toUnicode_ISO_2022_KR_OFFSETS_LOGIC_IBM(UConverterToUnicodeArgs *args,
                                                            UErrorCode* err){
    char const* sourceStart;
    UConverterDataISO2022* myData = static_cast<UConverterDataISO2022*>(args->converter->extraInfo);

    UConverterToUnicodeArgs subArgs;
    int32_t minArgsSize;

    /* set up the subconverter arguments */
    if(args->size<sizeof(UConverterToUnicodeArgs)) {
        minArgsSize = args->size;
    } else {
        minArgsSize = static_cast<int32_t>(sizeof(UConverterToUnicodeArgs));
    }

    uprv_memcpy(&subArgs, args, minArgsSize);
    subArgs.size = static_cast<uint16_t>(minArgsSize);
    subArgs.converter = myData->currentConverter;

    /* remember the original start of the input for offsets */
    sourceStart = args->source;

    if(myData->key != 0) {
        /* continue with a partial escape sequence */
        goto escape;
    }

    while(U_SUCCESS(*err) && args->source < args->sourceLimit) {
        /*Find the end of the buffer e.g : Next Escape Seq | end of Buffer*/
        subArgs.source = args->source;
        subArgs.sourceLimit = getEndOfBuffer_2022(&(args->source), args->sourceLimit, args->flush);
        if(subArgs.source != subArgs.sourceLimit) {
            /*
             * get the current partial byte sequence
             *
             * it needs to be moved between the public and the subconverter
             * so that the conversion framework, which only sees the public
             * converter, can handle truncated and illegal input etc.
             */
            if(args->converter->toULength > 0) {
                uprv_memcpy(subArgs.converter->toUBytes, args->converter->toUBytes, args->converter->toULength);
            }
            subArgs.converter->toULength = args->converter->toULength;

            /*
             * Convert up to the end of the input, or to before the next escape character.
             * Does not handle conversion extensions because the preToU[] state etc.
             * is not copied.
             */
            ucnv_MBCSToUnicodeWithOffsets(&subArgs, err);

            if(args->offsets != nullptr && sourceStart != args->source) {
                /* update offsets to base them on the actual start of the input */
                int32_t *offsets = args->offsets;
                char16_t *target = args->target;
                int32_t delta = static_cast<int32_t>(args->source - sourceStart);
                while(target < subArgs.target) {
                    if(*offsets >= 0) {
                        *offsets += delta;
                    }
                    ++offsets;
                    ++target;
                }
            }
            args->source = subArgs.source;
            args->target = subArgs.target;
            args->offsets = subArgs.offsets;

            /* copy input/error/overflow buffers */
            if(subArgs.converter->toULength > 0) {
                uprv_memcpy(args->converter->toUBytes, subArgs.converter->toUBytes, subArgs.converter->toULength);
            }
            args->converter->toULength = subArgs.converter->toULength;

            if(*err == U_BUFFER_OVERFLOW_ERROR) {
                if(subArgs.converter->UCharErrorBufferLength > 0) {
                    uprv_memcpy(args->converter->UCharErrorBuffer, subArgs.converter->UCharErrorBuffer,
                                subArgs.converter->UCharErrorBufferLength);
                }
                args->converter->UCharErrorBufferLength=subArgs.converter->UCharErrorBufferLength;
                subArgs.converter->UCharErrorBufferLength = 0;
            }
        }

        if (U_FAILURE(*err) || (args->source == args->sourceLimit)) {
            return;
        }

escape:
        changeState_2022(args->converter,
               &(args->source),
               args->sourceLimit,
               ISO_2022_KR,
               err);
    }
}

static void U_CALLCONV
UConverter_toUnicode_ISO_2022_KR_OFFSETS_LOGIC(UConverterToUnicodeArgs *args,
                                                            UErrorCode* err){
    char tempBuf[2];
    const char* mySource = const_cast<char*>(args->source);
    char16_t *myTarget = args->target;
    const char *mySourceLimit = args->sourceLimit;
    UChar32 targetUniChar = 0x0000;
    char16_t mySourceChar = 0x0000;
    UConverterDataISO2022* myData;
    UConverterSharedData* sharedData ;
    UBool useFallback;

    myData = static_cast<UConverterDataISO2022*>(args->converter->extraInfo);
    if(myData->version==1){
        UConverter_toUnicode_ISO_2022_KR_OFFSETS_LOGIC_IBM(args,err);
        return;
    }

    /* initialize state */
    sharedData = myData->currentConverter->sharedData;
    useFallback = args->converter->useFallback;

    if(myData->key != 0) {
        /* continue with a partial escape sequence */
        goto escape;
    } else if(args->converter->toULength == 1 && mySource < mySourceLimit && myTarget < args->targetLimit) {
        /* continue with a partial double-byte character */
        mySourceChar = args->converter->toUBytes[0];
        args->converter->toULength = 0;
        goto getTrailByte;
    }

    while(mySource< mySourceLimit){

        if(myTarget < args->targetLimit){

            mySourceChar = static_cast<unsigned char>(*mySource++);

            if(mySourceChar==UCNV_SI){
                myData->toU2022State.g = 0;
                if (myData->isEmptySegment) {
                    myData->isEmptySegment = false; /* we are handling it, reset to avoid future spurious errors */
                    *err = U_ILLEGAL_ESCAPE_SEQUENCE;
                    args->converter->toUCallbackReason = UCNV_IRREGULAR;
                    args->converter->toUBytes[0] = static_cast<uint8_t>(mySourceChar);
                    args->converter->toULength = 1;
                    args->target = myTarget;
                    args->source = mySource;
                    return;
                }
                /*consume the source */
                continue;
            }else if(mySourceChar==UCNV_SO){
                myData->toU2022State.g = 1;
                myData->isEmptySegment = true; /* Begin a new segment, empty so far */
                /*consume the source */
                continue;
            }else if(mySourceChar==ESC_2022){
                mySource--;
escape:
                myData->isEmptySegment = false; /* Any invalid ESC sequences will be detected separately, so just reset this */
                changeState_2022(args->converter,&(mySource),
                                mySourceLimit, ISO_2022_KR, err);
                if(U_FAILURE(*err)){
                    args->target = myTarget;
                    args->source = mySource;
                    return;
                }
                continue;
            }

            myData->isEmptySegment = false; /* Any invalid char errors will be detected separately, so just reset this */
            if(myData->toU2022State.g == 1) {
                if(mySource < mySourceLimit) {
                    int leadIsOk, trailIsOk;
                    uint8_t trailByte;
getTrailByte:
                    targetUniChar = missingCharMarker;
                    trailByte = static_cast<uint8_t>(*mySource);
                    /*
                     * Ticket 5691: consistent illegal sequences:
                     * - We include at least the first byte in the illegal sequence.
                     * - If any of the non-initial bytes could be the start of a character,
                     *   we stop the illegal sequence before the first one of those.
                     *
                     * In ISO-2022 DBCS, if the second byte is in the 21..7e range or is
                     * an ESC/SO/SI, we report only the first byte as the illegal sequence.
                     * Otherwise we convert or report the pair of bytes.
                     */
                    leadIsOk = static_cast<uint8_t>(mySourceChar - 0x21) <= (0x7e - 0x21);
                    trailIsOk = static_cast<uint8_t>(trailByte - 0x21) <= (0x7e - 0x21);
                    if (leadIsOk && trailIsOk) {
                        ++mySource;
                        tempBuf[0] = static_cast<char>(mySourceChar + 0x80);
                        tempBuf[1] = static_cast<char>(trailByte + 0x80);
                        targetUniChar = ucnv_MBCSSimpleGetNextUChar(sharedData, tempBuf, 2, useFallback);
                        mySourceChar = (mySourceChar << 8) | trailByte;
                    } else if (!(trailIsOk || IS_2022_CONTROL(trailByte))) {
                        /* report a pair of illegal bytes if the second byte is not a DBCS starter */
                        ++mySource;
                        /* add another bit so that the code below writes 2 bytes in case of error */
                        mySourceChar = static_cast<char16_t>(0x10000 | (mySourceChar << 8) | trailByte);
                    }
                } else {
                    args->converter->toUBytes[0] = static_cast<uint8_t>(mySourceChar);
                    args->converter->toULength = 1;
                    break;
                }
            }
            else if(mySourceChar <= 0x7f) {
                targetUniChar = ucnv_MBCSSimpleGetNextUChar(sharedData, mySource - 1, 1, useFallback);
            } else {
                targetUniChar = 0xffff;
            }
            if(targetUniChar < 0xfffe){
                if(args->offsets) {
                    args->offsets[myTarget - args->target] = static_cast<int32_t>(mySource - args->source - (mySourceChar <= 0xff ? 1 : 2));
                }
                *(myTarget++) = static_cast<char16_t>(targetUniChar);
            }
            else {
                /* Call the callback function*/
                toUnicodeCallback(args->converter,mySourceChar,targetUniChar,err);
                break;
            }
        }
        else{
            *err =U_BUFFER_OVERFLOW_ERROR;
            break;
        }
    }
    args->target = myTarget;
    args->source = mySource;
}

/*************************** END ISO2022-KR *********************************/

/*************************** ISO-2022-CN *********************************
*
* Rules for ISO-2022-CN Encoding:
* i)   The designator sequence must appear once on a line before any instance
*      of character set it designates.
* ii)  If two lines contain characters from the same character set, both lines
*      must include the designator sequence.
* iii) Once the designator sequence is known, a shifting sequence has to be found
*      to invoke the  shifting
* iv)  All lines start in ASCII and end in ASCII.
* v)   Four shifting sequences are employed for this purpose:
*
*      Sequcence   ASCII Eq    Charsets
*      ----------  -------    ---------
*      SI           <SI>        US-ASCII
*      SO           <SO>        CNS-11643-1992 Plane 1, GB2312, ISO-IR-165
*      SS2          <ESC>N      CNS-11643-1992 Plane 2
*      SS3          <ESC>O      CNS-11643-1992 Planes 3-7
*
* vi)
*      SOdesignator  : ESC "$" ")" finalchar_for_SO
*      SS2designator : ESC "$" "*" finalchar_for_SS2
*      SS3designator : ESC "$" "+" finalchar_for_SS3
*
*      ESC $ ) A       Indicates the bytes following SO are Chinese
*       characters as defined in GB 2312-80, until
*       another SOdesignation appears
*
*
*      ESC $ ) E       Indicates the bytes following SO are as defined
*       in ISO-IR-165 (for details, see section 2.1),
*       until another SOdesignation appears
*
*      ESC $ ) G       Indicates the bytes following SO are as defined
*       in CNS 11643-plane-1, until another
*       SOdesignation appears
*
*      ESC $ * H       Indicates the two bytes immediately following
*       SS2 is a Chinese character as defined in CNS
*       11643-plane-2, until another SS2designation
*       appears
*       (Meaning <ESC>N must precede every 2 byte
*        sequence.)
*
*      ESC $ + I       Indicates the immediate two bytes following SS3
*       is a Chinese character as defined in CNS
*       11643-plane-3, until another SS3designation
*       appears
*       (Meaning <ESC>O must precede every 2 byte
*        sequence.)
*
*      ESC $ + J       Indicates the immediate two bytes following SS3
*       is a Chinese character as defined in CNS
*       11643-plane-4, until another SS3designation
*       appears
*       (In English: <ESC>O must precede every 2 byte
*        sequence.)
*
*      ESC $ + K       Indicates the immediate two bytes following SS3
*       is a Chinese character as defined in CNS
*       11643-plane-5, until another SS3designation
*       appears
*
*      ESC $ + L       Indicates the immediate two bytes following SS3
*       is a Chinese character as defined in CNS
*       11643-plane-6, until another SS3designation
*       appears
*
*      ESC $ + M       Indicates the immediate two bytes following SS3
*       is a Chinese character as defined in CNS
*       11643-plane-7, until another SS3designation
*       appears
*
*       As in ISO-2022-CN, each line starts in ASCII, and ends in ASCII, and
*       has its own designation information before any Chinese characters
*       appear
*
*/

/* The following are defined this way to make the strings truly readonly */
static const char GB_2312_80_STR[] = "\x1B\x24\x29\x41";
static const char ISO_IR_165_STR[] = "\x1B\x24\x29\x45";
static const char CNS_11643_1992_Plane_1_STR[] = "\x1B\x24\x29\x47";
static const char CNS_11643_1992_Plane_2_STR[] = "\x1B\x24\x2A\x48";
static const char CNS_11643_1992_Plane_3_STR[] = "\x1B\x24\x2B\x49";
static const char CNS_11643_1992_Plane_4_STR[] = "\x1B\x24\x2B\x4A";
static const char CNS_11643_1992_Plane_5_STR[] = "\x1B\x24\x2B\x4B";
static const char CNS_11643_1992_Plane_6_STR[] = "\x1B\x24\x2B\x4C";
static const char CNS_11643_1992_Plane_7_STR[] = "\x1B\x24\x2B\x4D";

/********************** ISO2022-CN Data **************************/
static const char* const escSeqCharsCN[10] ={
        SHIFT_IN_STR,                   /* 0 ASCII */
        GB_2312_80_STR,                 /* 1 GB2312_1 */
        ISO_IR_165_STR,                 /* 2 ISO_IR_165 */
        CNS_11643_1992_Plane_1_STR,
        CNS_11643_1992_Plane_2_STR,
        CNS_11643_1992_Plane_3_STR,
        CNS_11643_1992_Plane_4_STR,
        CNS_11643_1992_Plane_5_STR,
        CNS_11643_1992_Plane_6_STR,
        CNS_11643_1992_Plane_7_STR
};

static void U_CALLCONV
UConverter_fromUnicode_ISO_2022_CN_OFFSETS_LOGIC(UConverterFromUnicodeArgs* args, UErrorCode* err){
    UConverter *cnv = args->converter;
    UConverterDataISO2022 *converterData;
    ISO2022State *pFromU2022State;
    uint8_t* target = reinterpret_cast<uint8_t*>(args->target);
    const uint8_t* targetLimit = reinterpret_cast<const uint8_t*>(args->targetLimit);
    const char16_t* source = args->source;
    const char16_t* sourceLimit = args->sourceLimit;
    int32_t* offsets = args->offsets;
    UChar32 sourceChar;
    char buffer[8];
    int32_t len;
    int8_t choices[3];
    int32_t choiceCount;
    uint32_t targetValue = 0;
    UBool useFallback;

    /* set up the state */
    converterData = static_cast<UConverterDataISO2022*>(cnv->extraInfo);
    pFromU2022State   = &converterData->fromU2022State;

    choiceCount = 0;

    /* check if the last codepoint of previous buffer was a lead surrogate*/
    if((sourceChar = cnv->fromUChar32)!=0 && target< targetLimit) {
        goto getTrail;
    }

    while( source < sourceLimit){
        if(target < targetLimit){

            sourceChar  = *(source++);
            /*check if the char is a First surrogate*/
             if(U16_IS_SURROGATE(sourceChar)) {
                if(U16_IS_SURROGATE_LEAD(sourceChar)) {
getTrail:
                    /*look ahead to find the trail surrogate*/
                    if(source < sourceLimit) {
                        /* test the following code unit */
                        char16_t trail = *source;
                        if(U16_IS_TRAIL(trail)) {
                            source++;
                            sourceChar=U16_GET_SUPPLEMENTARY(sourceChar, trail);
                            cnv->fromUChar32=0x00;
                            /* convert this supplementary code point */
                            /* exit this condition tree */
                        } else {
                            /* this is an unmatched lead code unit (1st surrogate) */
                            /* callback(illegal) */
                            *err=U_ILLEGAL_CHAR_FOUND;
                            cnv->fromUChar32=sourceChar;
                            break;
                        }
                    } else {
                        /* no more input */
                        cnv->fromUChar32=sourceChar;
                        break;
                    }
                } else {
                    /* this is an unmatched trail code unit (2nd surrogate) */
                    /* callback(illegal) */
                    *err=U_ILLEGAL_CHAR_FOUND;
                    cnv->fromUChar32=sourceChar;
                    break;
                }
            }

            /* do the conversion */
            if(sourceChar <= 0x007f ){
                /* do not convert SO/SI/ESC */
                if(IS_2022_CONTROL(sourceChar)) {
                    /* callback(illegal) */
                    *err=U_ILLEGAL_CHAR_FOUND;
                    cnv->fromUChar32=sourceChar;
                    break;
                }

                /* US-ASCII */
                if(pFromU2022State->g == 0) {
                    buffer[0] = static_cast<char>(sourceChar);
                    len = 1;
                } else {
                    buffer[0] = UCNV_SI;
                    buffer[1] = static_cast<char>(sourceChar);
                    len = 2;
                    pFromU2022State->g = 0;
                    choiceCount = 0;
                }
                if(sourceChar == CR || sourceChar == LF) {
                    /* reset the state at the end of a line */
                    uprv_memset(pFromU2022State, 0, sizeof(ISO2022State));
                    choiceCount = 0;
                }
            }
            else{
                /* convert U+0080..U+10ffff */
                int32_t i;
                int8_t cs, g;

                if(choiceCount == 0) {
                    /* try the current SO/G1 converter first */
                    choices[0] = pFromU2022State->cs[1];

                    /* default to GB2312_1 if none is designated yet */
                    if(choices[0] == 0) {
                        choices[0] = GB2312_1;
                    }

                    if(converterData->version == 0) {
                        /* ISO-2022-CN */

                        /* try the other SO/G1 converter; a CNS_11643_1 lookup may result in any plane */
                        if(choices[0] == GB2312_1) {
                            choices[1] = static_cast<int8_t>(CNS_11643_1);
                        } else {
                            choices[1] = static_cast<int8_t>(GB2312_1);
                        }

                        choiceCount = 2;
                    } else if (converterData->version == 1) {
                        /* ISO-2022-CN-EXT */

                        /* try one of the other converters */
                        switch(choices[0]) {
                        case GB2312_1:
                            choices[1] = static_cast<int8_t>(CNS_11643_1);
                            choices[2] = static_cast<int8_t>(ISO_IR_165);
                            break;
                        case ISO_IR_165:
                            choices[1] = static_cast<int8_t>(GB2312_1);
                            choices[2] = static_cast<int8_t>(CNS_11643_1);
                            break;
                        default: /* CNS_11643_x */
                            choices[1] = static_cast<int8_t>(GB2312_1);
                            choices[2] = static_cast<int8_t>(ISO_IR_165);
                            break;
                        }

                        choiceCount = 3;
                    } else {
                        choices[0] = static_cast<int8_t>(CNS_11643_1);
                        choices[1] = static_cast<int8_t>(GB2312_1);
                    }
                }

                cs = g = 0;
                /*
                 * len==0: no mapping found yet
                 * len<0: found a fallback result: continue looking for a roundtrip but no further fallbacks
                 * len>0: found a roundtrip result, done
                 */
                len = 0;
                /*
                 * We will turn off useFallback after finding a fallback,
                 * but we still get fallbacks from PUA code points as usual.
                 * Therefore, we will also need to check that we don't overwrite
                 * an early fallback with a later one.
                 */
                useFallback = cnv->useFallback;

                for(i = 0; i < choiceCount && len <= 0; ++i) {
                    int8_t cs0 = choices[i];
                    if(cs0 > 0) {
                        uint32_t value;
                        int32_t len2;
                        if(cs0 >= CNS_11643_0) {
                            len2 = MBCS_FROM_UCHAR32_ISO2022(
                                        converterData->myConverterArray[CNS_11643],
                                        sourceChar,
                                        &value,
                                        useFallback,
                                        MBCS_OUTPUT_3);
                            if(len2 == 3 || (len2 == -3 && len == 0)) {
                                targetValue = value;
                                cs = static_cast<int8_t>(CNS_11643_0 + (value >> 16) - 0x80);
                                if(len2 >= 0) {
                                    len = 2;
                                } else {
                                    len = -2;
                                    useFallback = false;
                                }
                                if(cs == CNS_11643_1) {
                                    g = 1;
                                } else if(cs == CNS_11643_2) {
                                    g = 2;
                                } else /* plane 3..7 */ if(converterData->version == 1) {
                                    g = 3;
                                } else {
                                    /* ISO-2022-CN (without -EXT) does not support plane 3..7 */
                                    len = 0;
                                }
                            }
                        } else {
                            /* GB2312_1 or ISO-IR-165 */
                            U_ASSERT(cs0<UCNV_2022_MAX_CONVERTERS);
                            len2 = MBCS_FROM_UCHAR32_ISO2022(
                                        converterData->myConverterArray[cs0],
                                        sourceChar,
                                        &value,
                                        useFallback,
                                        MBCS_OUTPUT_2);
                            if(len2 == 2 || (len2 == -2 && len == 0)) {
                                targetValue = value;
                                len = len2;
                                cs = cs0;
                                g = 1;
                                useFallback = false;
                            }
                        }
                    }
                }

                if(len != 0) {
                    len = 0; /* count output bytes; it must have been abs(len) == 2 */

                    /* write the designation sequence if necessary */
                    if(cs != pFromU2022State->cs[g]) {
                        if(cs < CNS_11643) {
                            uprv_memcpy(buffer, escSeqCharsCN[cs], 4);
                        } else {
                            U_ASSERT(cs >= CNS_11643_1);
                            uprv_memcpy(buffer, escSeqCharsCN[CNS_11643 + (cs - CNS_11643_1)], 4);
                        }
                        len = 4;
                        pFromU2022State->cs[g] = cs;
                        if(g == 1) {
                            /* changing the SO/G1 charset invalidates the choices[] */
                            choiceCount = 0;
                        }
                    }

                    /* write the shift sequence if necessary */
                    if(g != pFromU2022State->g) {
                        switch(g) {
                        case 1:
                            buffer[len++] = UCNV_SO;

                            /* set the new state only if it is the locking shift SO/G1, not for SS2 or SS3 */
                            pFromU2022State->g = 1;
                            break;
                        case 2:
                            buffer[len++] = 0x1b;
                            buffer[len++] = 0x4e;
                            break;
                        default: /* case 3 */
                            buffer[len++] = 0x1b;
                            buffer[len++] = 0x4f;
                            break;
                        }
                    }

                    /* write the two output bytes */
                    buffer[len++] = static_cast<char>(targetValue >> 8);
                    buffer[len++] = static_cast<char>(targetValue);
                } else {
                    /* if we cannot find the character after checking all codepages
                     * then this is an error
                     */
                    *err = U_INVALID_CHAR_FOUND;
                    cnv->fromUChar32=sourceChar;
                    break;
                }
            }

            /* output len>0 bytes in buffer[] */
            if(len == 1) {
                *target++ = buffer[0];
                if(offsets) {
                    *offsets++ = static_cast<int32_t>(source - args->source - 1); /* -1: known to be ASCII */
                }
            } else if(len == 2 && (target + 2) <= targetLimit) {
                *target++ = buffer[0];
                *target++ = buffer[1];
                if(offsets) {
                    int32_t sourceIndex = static_cast<int32_t>(source - args->source - U16_LENGTH(sourceChar));
                    *offsets++ = sourceIndex;
                    *offsets++ = sourceIndex;
                }
            } else {
                fromUWriteUInt8(
                    cnv,
                    buffer, len,
                    &target, reinterpret_cast<const char*>(targetLimit),
                    &offsets, static_cast<int32_t>(source - args->source - U16_LENGTH(sourceChar)),
                    err);
                if(U_FAILURE(*err)) {
                    break;
                }
            }
        } /* end if(myTargetIndex<myTargetLength) */
        else{
            *err =U_BUFFER_OVERFLOW_ERROR;
            break;
        }

    }/* end while(mySourceIndex<mySourceLength) */

    /*
     * the end of the input stream and detection of truncated input
     * are handled by the framework, but for ISO-2022-CN conversion
     * we need to be in ASCII mode at the very end
     *
     * conditions:
     *   successful
     *   not in ASCII mode
     *   end of input and no truncated input
     */
    if( U_SUCCESS(*err) &&
        pFromU2022State->g!=0 &&
        args->flush && source>=sourceLimit && cnv->fromUChar32==0
    ) {
        int32_t sourceIndex;

        /* we are switching to ASCII */
        pFromU2022State->g=0;

        /* get the source index of the last input character */
        /*
         * TODO this would be simpler and more reliable if we used a pair
         * of sourceIndex/prevSourceIndex like in ucnvmbcs.c
         * so that we could simply use the prevSourceIndex here;
         * this code gives an incorrect result for the rare case of an unmatched
         * trail surrogate that is alone in the last buffer of the text stream
         */
        sourceIndex = static_cast<int32_t>(source - args->source);
        if(sourceIndex>0) {
            --sourceIndex;
            if( U16_IS_TRAIL(args->source[sourceIndex]) &&
                (sourceIndex==0 || U16_IS_LEAD(args->source[sourceIndex-1]))
            ) {
                --sourceIndex;
            }
        } else {
            sourceIndex=-1;
        }

        fromUWriteUInt8(
            cnv,
            SHIFT_IN_STR, 1,
            &target, reinterpret_cast<const char*>(targetLimit),
            &offsets, sourceIndex,
            err);
    }

    /*save the state and return */
    args->source = source;
    args->target = reinterpret_cast<char*>(target);
}


static void U_CALLCONV
UConverter_toUnicode_ISO_2022_CN_OFFSETS_LOGIC(UConverterToUnicodeArgs *args,
                                               UErrorCode* err){
    char tempBuf[3];
    const char* mySource = const_cast<char*>(args->source);
    char16_t *myTarget = args->target;
    const char *mySourceLimit = args->sourceLimit;
    uint32_t targetUniChar = 0x0000;
    uint32_t mySourceChar = 0x0000;
    UConverterDataISO2022* myData;
    ISO2022State *pToU2022State;

    myData = static_cast<UConverterDataISO2022*>(args->converter->extraInfo);
    pToU2022State = &myData->toU2022State;

    if(myData->key != 0) {
        /* continue with a partial escape sequence */
        goto escape;
    } else if(args->converter->toULength == 1 && mySource < mySourceLimit && myTarget < args->targetLimit) {
        /* continue with a partial double-byte character */
        mySourceChar = args->converter->toUBytes[0];
        args->converter->toULength = 0;
        targetUniChar = missingCharMarker;
        goto getTrailByte;
    }

    while(mySource < mySourceLimit){

        targetUniChar =missingCharMarker;

        if(myTarget < args->targetLimit){

            mySourceChar = static_cast<unsigned char>(*mySource++);

            switch(mySourceChar){
            case UCNV_SI:
                pToU2022State->g=0;
                if (myData->isEmptySegment) {
                    myData->isEmptySegment = false; /* we are handling it, reset to avoid future spurious errors */
                    *err = U_ILLEGAL_ESCAPE_SEQUENCE;
                    args->converter->toUCallbackReason = UCNV_IRREGULAR;
                    args->converter->toUBytes[0] = static_cast<uint8_t>(mySourceChar);
                    args->converter->toULength = 1;
                    args->target = myTarget;
                    args->source = mySource;
                    return;
                }
                continue;

            case UCNV_SO:
                if(pToU2022State->cs[1] != 0) {
                    pToU2022State->g=1;
                    myData->isEmptySegment = true; /* Begin a new segment, empty so far */
                    continue;
                } else {
                    /* illegal to have SO before a matching designator */
                    myData->isEmptySegment = false; /* Handling a different error, reset this to avoid future spurious errs */
                    break;
                }

            case ESC_2022:
                mySource--;
escape:
                {
                    const char * mySourceBefore = mySource;
                    int8_t toULengthBefore = args->converter->toULength;

                    changeState_2022(args->converter,&(mySource),
                        mySourceLimit, ISO_2022_CN,err);

                    /* After SO there must be at least one character before a designator (designator error handled separately) */
                    if(myData->key==0 && U_SUCCESS(*err) && myData->isEmptySegment) {
                        *err = U_ILLEGAL_ESCAPE_SEQUENCE;
                        args->converter->toUCallbackReason = UCNV_IRREGULAR;
                        args->converter->toULength = static_cast<int8_t>(toULengthBefore + (mySource - mySourceBefore));
                    }
                }

                /* invalid or illegal escape sequence */
                if(U_FAILURE(*err)){
                    args->target = myTarget;
                    args->source = mySource;
                    myData->isEmptySegment = false; /* Reset to avoid future spurious errors */
                    return;
                }
                continue;

            /* ISO-2022-CN does not use single-byte (C1) SS2 and SS3 */

            case CR:
            case LF:
                uprv_memset(pToU2022State, 0, sizeof(ISO2022State));
                U_FALLTHROUGH;
            default:
                /* convert one or two bytes */
                myData->isEmptySegment = false;
                if(pToU2022State->g != 0) {
                    if(mySource < mySourceLimit) {
                        UConverterSharedData *cnv;
                        StateEnum tempState;
                        int32_t tempBufLen;
                        int leadIsOk, trailIsOk;
                        uint8_t trailByte;
getTrailByte:
                        trailByte = static_cast<uint8_t>(*mySource);
                        /*
                         * Ticket 5691: consistent illegal sequences:
                         * - We include at least the first byte in the illegal sequence.
                         * - If any of the non-initial bytes could be the start of a character,
                         *   we stop the illegal sequence before the first one of those.
                         *
                         * In ISO-2022 DBCS, if the second byte is in the 21..7e range or is
                         * an ESC/SO/SI, we report only the first byte as the illegal sequence.
                         * Otherwise we convert or report the pair of bytes.
                         */
                        leadIsOk = static_cast<uint8_t>(mySourceChar - 0x21) <= (0x7e - 0x21);
                        trailIsOk = static_cast<uint8_t>(trailByte - 0x21) <= (0x7e - 0x21);
                        if (leadIsOk && trailIsOk) {
                            ++mySource;
                            tempState = static_cast<StateEnum>(pToU2022State->cs[pToU2022State->g]);
                            if(tempState >= CNS_11643_0) {
                                cnv = myData->myConverterArray[CNS_11643];
                                tempBuf[0] = static_cast<char>(0x80 + (tempState - CNS_11643_0));
                                tempBuf[1] = static_cast<char>(mySourceChar);
                                tempBuf[2] = static_cast<char>(trailByte);
                                tempBufLen = 3;

                            }else{
                                U_ASSERT(tempState<UCNV_2022_MAX_CONVERTERS);
                                cnv = myData->myConverterArray[tempState];
                                tempBuf[0] = static_cast<char>(mySourceChar);
                                tempBuf[1] = static_cast<char>(trailByte);
                                tempBufLen = 2;
                            }
                            targetUniChar = ucnv_MBCSSimpleGetNextUChar(cnv, tempBuf, tempBufLen, false);
                            mySourceChar = (mySourceChar << 8) | trailByte;
                        } else if (!(trailIsOk || IS_2022_CONTROL(trailByte))) {
                            /* report a pair of illegal bytes if the second byte is not a DBCS starter */
                            ++mySource;
                            /* add another bit so that the code below writes 2 bytes in case of error */
                            mySourceChar = 0x10000 | (mySourceChar << 8) | trailByte;
                        }
                        if(pToU2022State->g>=2) {
                            /* return from a single-shift state to the previous one */
                            pToU2022State->g=pToU2022State->prevG;
                        }
                    } else {
                        args->converter->toUBytes[0] = static_cast<uint8_t>(mySourceChar);
                        args->converter->toULength = 1;
                        goto endloop;
                    }
                }
                else{
                    if(mySourceChar <= 0x7f) {
                        targetUniChar = static_cast<char16_t>(mySourceChar);
                    }
                }
                break;
            }
            if(targetUniChar < (missingCharMarker-1/*0xfffe*/)){
                if(args->offsets){
                    args->offsets[myTarget - args->target] = static_cast<int32_t>(mySource - args->source - (mySourceChar <= 0xff ? 1 : 2));
                }
                *(myTarget++) = static_cast<char16_t>(targetUniChar);
            }
            else if(targetUniChar > missingCharMarker){
                /* disassemble the surrogate pair and write to output*/
                targetUniChar-=0x0010000;
                *myTarget = static_cast<char16_t>(0xd800 + static_cast<char16_t>(targetUniChar >> 10));
                if(args->offsets){
                    args->offsets[myTarget - args->target] = static_cast<int32_t>(mySource - args->source - (mySourceChar <= 0xff ? 1 : 2));
                }
                ++myTarget;
                if(myTarget< args->targetLimit){
                    *myTarget = static_cast<char16_t>(0xdc00 + static_cast<char16_t>(targetUniChar & 0x3ff));
                    if(args->offsets){
                        args->offsets[myTarget - args->target] = static_cast<int32_t>(mySource - args->source - (mySourceChar <= 0xff ? 1 : 2));
                    }
                    ++myTarget;
                }else{
                    args->converter->UCharErrorBuffer[args->converter->UCharErrorBufferLength++]=
                                    static_cast<char16_t>(0xdc00 + static_cast<char16_t>(targetUniChar & 0x3ff));
                }

            }
            else{
                /* Call the callback function*/
                toUnicodeCallback(args->converter,mySourceChar,targetUniChar,err);
                break;
            }
        }
        else{
            *err =U_BUFFER_OVERFLOW_ERROR;
            break;
        }
    }
endloop:
    args->target = myTarget;
    args->source = mySource;
}
#endif /* #if !UCONFIG_ONLY_HTML_CONVERSION */

static void U_CALLCONV
_ISO_2022_WriteSub(UConverterFromUnicodeArgs *args, int32_t offsetIndex, UErrorCode *err) {
    UConverter *cnv = args->converter;
    UConverterDataISO2022* myConverterData = static_cast<UConverterDataISO2022*>(cnv->extraInfo);
    ISO2022State *pFromU2022State=&myConverterData->fromU2022State;
    char *p, *subchar;
    char buffer[8];
    int32_t length;

    subchar = reinterpret_cast<char*>(cnv->subChars);
    length=cnv->subCharLen; /* assume length==1 for most variants */

    p = buffer;
    switch(myConverterData->locale[0]){
    case 'j':
        {
            int8_t cs;

            if(pFromU2022State->g == 1) {
                /* JIS7: switch from G1 to G0 */
                pFromU2022State->g = 0;
                *p++ = UCNV_SI;
            }

            cs = pFromU2022State->cs[0];
            if(cs != ASCII && cs != JISX201) {
                /* not in ASCII or JIS X 0201: switch to ASCII */
                pFromU2022State->cs[0] = static_cast<int8_t>(ASCII);
                *p++ = '\x1b';
                *p++ = '\x28';
                *p++ = '\x42';
            }

            *p++ = subchar[0];
            break;
        }
    case 'c':
        if(pFromU2022State->g != 0) {
            /* not in ASCII mode: switch to ASCII */
            pFromU2022State->g = 0;
            *p++ = UCNV_SI;
        }
        *p++ = subchar[0];
        break;
    case 'k':
        if(myConverterData->version == 0) {
            if(length == 1) {
                if(args->converter->fromUnicodeStatus) {
                    /* in DBCS mode: switch to SBCS */
                    args->converter->fromUnicodeStatus = 0;
                    *p++ = UCNV_SI;
                }
                *p++ = subchar[0];
            } else /* length == 2*/ {
                if(!args->converter->fromUnicodeStatus) {
                    /* in SBCS mode: switch to DBCS */
                    args->converter->fromUnicodeStatus = 1;
                    *p++ = UCNV_SO;
                }
                *p++ = subchar[0];
                *p++ = subchar[1];
            }
            break;
        } else {
            /* save the subconverter's substitution string */
            uint8_t *currentSubChars = myConverterData->currentConverter->subChars;
            int8_t currentSubCharLen = myConverterData->currentConverter->subCharLen;

            /* set our substitution string into the subconverter */
            myConverterData->currentConverter->subChars = reinterpret_cast<uint8_t*>(subchar);
            myConverterData->currentConverter->subCharLen = static_cast<int8_t>(length);

            /* let the subconverter write the subchar, set/retrieve fromUChar32 state */
            args->converter = myConverterData->currentConverter;
            myConverterData->currentConverter->fromUChar32 = cnv->fromUChar32;
            ucnv_cbFromUWriteSub(args, 0, err);
            cnv->fromUChar32 = myConverterData->currentConverter->fromUChar32;
            args->converter = cnv;

            /* restore the subconverter's substitution string */
            myConverterData->currentConverter->subChars = currentSubChars;
            myConverterData->currentConverter->subCharLen = currentSubCharLen;

            if(*err == U_BUFFER_OVERFLOW_ERROR) {
                if(myConverterData->currentConverter->charErrorBufferLength > 0) {
                    uprv_memcpy(
                        cnv->charErrorBuffer,
                        myConverterData->currentConverter->charErrorBuffer,
                        myConverterData->currentConverter->charErrorBufferLength);
                }
                cnv->charErrorBufferLength = myConverterData->currentConverter->charErrorBufferLength;
                myConverterData->currentConverter->charErrorBufferLength = 0;
            }
            return;
        }
    default:
        /* not expected */
        break;
    }
    ucnv_cbFromUWriteBytes(args,
                           buffer, static_cast<int32_t>(p - buffer),
                           offsetIndex, err);
}

/*
 * Structure for cloning an ISO 2022 converter into a single memory block.
 */
struct cloneStruct
{
    UConverter cnv;
    UConverter currentConverter;
    UConverterDataISO2022 mydata;
};


U_CDECL_BEGIN

static UConverter * U_CALLCONV
_ISO_2022_SafeClone(
            const UConverter *cnv,
            void *stackBuffer,
            int32_t *pBufferSize,
            UErrorCode *status)
{
    struct cloneStruct * localClone;
    UConverterDataISO2022 *cnvData;
    int32_t i, size;

    if (U_FAILURE(*status)){
        return nullptr;
    }

    if (*pBufferSize == 0) { /* 'preflighting' request - set needed size into *pBufferSize */
        *pBufferSize = (int32_t)sizeof(struct cloneStruct);
        return nullptr;
    }

    cnvData = (UConverterDataISO2022 *)cnv->extraInfo;
    localClone = (struct cloneStruct *)stackBuffer;

    /* ucnv.c/ucnv_safeClone() copied the main UConverter already */

    uprv_memcpy(&localClone->mydata, cnvData, sizeof(UConverterDataISO2022));
    localClone->cnv.extraInfo = &localClone->mydata; /* set pointer to extra data */
    localClone->cnv.isExtraLocal = true;

    /* share the subconverters */

    if(cnvData->currentConverter != nullptr) {
        size = (int32_t)sizeof(UConverter);
        localClone->mydata.currentConverter =
            ucnv_safeClone(cnvData->currentConverter,
                            &localClone->currentConverter,
                            &size, status);
        if(U_FAILURE(*status)) {
            return nullptr;
        }
    }

    for(i=0; i<UCNV_2022_MAX_CONVERTERS; ++i) {
        if(cnvData->myConverterArray[i] != nullptr) {
            ucnv_incrementRefCount(cnvData->myConverterArray[i]);
        }
    }

    return &localClone->cnv;
}

U_CDECL_END

static void U_CALLCONV
_ISO_2022_GetUnicodeSet(const UConverter *cnv,
                    const USetAdder *sa,
                    UConverterUnicodeSet which,
                    UErrorCode *pErrorCode)
{
    int32_t i;
    UConverterDataISO2022* cnvData;

    if (U_FAILURE(*pErrorCode)) {
        return;
    }
#ifdef U_ENABLE_GENERIC_ISO_2022
    if (cnv->sharedData == &_ISO2022Data) {
        /* We use UTF-8 in this case */
        sa->addRange(sa->set, 0, 0xd7FF);
        sa->addRange(sa->set, 0xE000, 0x10FFFF);
        return;
    }
#endif

    cnvData = static_cast<UConverterDataISO2022*>(cnv->extraInfo);

    /* open a set and initialize it with code points that are algorithmically round-tripped */
    switch(cnvData->locale[0]){
    case 'j':
        /* include JIS X 0201 which is hardcoded */
        sa->add(sa->set, 0xa5);
        sa->add(sa->set, 0x203e);
        if(jpCharsetMasks[cnvData->version]&CSM(ISO8859_1)) {
            /* include Latin-1 for some variants of JP */
            sa->addRange(sa->set, 0, 0xff);
        } else {
            /* include ASCII for JP */
            sa->addRange(sa->set, 0, 0x7f);
        }
        if(cnvData->version==3 || cnvData->version==4 || which==UCNV_ROUNDTRIP_AND_FALLBACK_SET) {
            /*
             * Do not test (jpCharsetMasks[cnvData->version]&CSM(HWKANA_7BIT))!=0
             * because the bit is on for all JP versions although only versions 3 & 4 (JIS7 & JIS8)
             * use half-width Katakana.
             * This is because all ISO-2022-JP variants are lenient in that they accept (in toUnicode)
             * half-width Katakana via the ESC ( I sequence.
             * However, we only emit (fromUnicode) half-width Katakana according to the
             * definition of each variant.
             *
             * When including fallbacks,
             * we need to include half-width Katakana Unicode code points for all JP variants because
             * JIS X 0208 has hardcoded fallbacks for them (which map to full-width Katakana).
             */
            /* include half-width Katakana for JP */
            sa->addRange(sa->set, HWKANA_START, HWKANA_END);
        }
        break;
#if !UCONFIG_ONLY_HTML_CONVERSION
    case 'c':
    case 'z':
        /* include ASCII for CN */
        sa->addRange(sa->set, 0, 0x7f);
        break;
    case 'k':
        /* there is only one converter for KR, and it is not in the myConverterArray[] */
        cnvData->currentConverter->sharedData->impl->getUnicodeSet(
                cnvData->currentConverter, sa, which, pErrorCode);
        /* the loop over myConverterArray[] will simply not find another converter */
        break;
#endif
    default:
        break;
    }

#if 0  /* Replaced by ucnv_MBCSGetFilteredUnicodeSetForUnicode() until we implement ucnv_getUnicodeSet() with reverse fallbacks. */
            if( (cnvData->locale[0]=='c' || cnvData->locale[0]=='z') &&
                cnvData->version==0 && i==CNS_11643
            ) {
                /* special handling for non-EXT ISO-2022-CN: add only code points for CNS planes 1 and 2 */
                ucnv_MBCSGetUnicodeSetForBytes(
                        cnvData->myConverterArray[i],
                        sa, UCNV_ROUNDTRIP_SET,
                        0, 0x81, 0x82,
                        pErrorCode);
            }
#endif

    for (i=0; i<UCNV_2022_MAX_CONVERTERS; i++) {
        UConverterSetFilter filter;
        if(cnvData->myConverterArray[i]!=nullptr) {
            if(cnvData->locale[0]=='j' && i==JISX208) {
                /*
                 * Only add code points that map to Shift-JIS codes
                 * corresponding to JIS X 0208.
                 */
                filter=UCNV_SET_FILTER_SJIS;
#if !UCONFIG_ONLY_HTML_CONVERSION
            } else if( (cnvData->locale[0]=='c' || cnvData->locale[0]=='z') &&
                       cnvData->version==0 && i==CNS_11643) {
                /*
                 * Version-specific for CN:
                 * CN version 0 does not map CNS planes 3..7 although
                 * they are all available in the CNS conversion table;
                 * CN version 1 (-EXT) does map them all.
                 * The two versions create different Unicode sets.
                 */
                filter=UCNV_SET_FILTER_2022_CN;
            } else if(i==KSC5601) {
                /*
                 * Some of the KSC 5601 tables (convrtrs.txt has this aliases on multiple tables)
                 * are broader than GR94.
                 */
                filter=UCNV_SET_FILTER_GR94DBCS;
#endif
            } else {
                filter=UCNV_SET_FILTER_NONE;
            }
            ucnv_MBCSGetFilteredUnicodeSetForUnicode(cnvData->myConverterArray[i], sa, which, filter, pErrorCode);
        }
    }

    /*
     * ISO 2022 converters must not convert SO/SI/ESC despite what
     * sub-converters do by themselves.
     * Remove these characters from the set.
     */
    sa->remove(sa->set, 0x0e);
    sa->remove(sa->set, 0x0f);
    sa->remove(sa->set, 0x1b);

    /* ISO 2022 converters do not convert C1 controls either */
    sa->removeRange(sa->set, 0x80, 0x9f);
}

static const UConverterImpl _ISO2022Impl={
    UCNV_ISO_2022,

    nullptr,
    nullptr,

    _ISO2022Open,
    _ISO2022Close,
    _ISO2022Reset,

#ifdef U_ENABLE_GENERIC_ISO_2022
    T_UConverter_toUnicode_ISO_2022_OFFSETS_LOGIC,
    T_UConverter_toUnicode_ISO_2022_OFFSETS_LOGIC,
    ucnv_fromUnicode_UTF8,
    ucnv_fromUnicode_UTF8_OFFSETS_LOGIC,
#else
    nullptr,
    nullptr,
    nullptr,
    nullptr,
#endif
    nullptr,

    nullptr,
    _ISO2022getName,
    _ISO_2022_WriteSub,
    _ISO_2022_SafeClone,
    _ISO_2022_GetUnicodeSet,

    nullptr,
    nullptr
};
static const UConverterStaticData _ISO2022StaticData={
    sizeof(UConverterStaticData),
    "ISO_2022",
    2022,
    UCNV_IBM,
    UCNV_ISO_2022,
    1,
    3, /* max 3 bytes per char16_t from UTF-8 (4 bytes from surrogate _pair_) */
    { 0x1a, 0, 0, 0 },
    1,
    false,
    false,
    0,
    0,
    { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } /* reserved */
};
const UConverterSharedData _ISO2022Data=
        UCNV_IMMUTABLE_SHARED_DATA_INITIALIZER(&_ISO2022StaticData, &_ISO2022Impl);

/*************JP****************/
static const UConverterImpl _ISO2022JPImpl={
    UCNV_ISO_2022,

    nullptr,
    nullptr,

    _ISO2022Open,
    _ISO2022Close,
    _ISO2022Reset,

    UConverter_toUnicode_ISO_2022_JP_OFFSETS_LOGIC,
    UConverter_toUnicode_ISO_2022_JP_OFFSETS_LOGIC,
    UConverter_fromUnicode_ISO_2022_JP_OFFSETS_LOGIC,
    UConverter_fromUnicode_ISO_2022_JP_OFFSETS_LOGIC,
    nullptr,

    nullptr,
    _ISO2022getName,
    _ISO_2022_WriteSub,
    _ISO_2022_SafeClone,
    _ISO_2022_GetUnicodeSet,

    nullptr,
    nullptr
};
static const UConverterStaticData _ISO2022JPStaticData={
    sizeof(UConverterStaticData),
    "ISO_2022_JP",
    0,
    UCNV_IBM,
    UCNV_ISO_2022,
    1,
    6, /* max 6 bytes per char16_t: 4-byte escape sequence + DBCS */
    { 0x1a, 0, 0, 0 },
    1,
    false,
    false,
    0,
    0,
    { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } /* reserved */
};

namespace {

const UConverterSharedData _ISO2022JPData=
        UCNV_IMMUTABLE_SHARED_DATA_INITIALIZER(&_ISO2022JPStaticData, &_ISO2022JPImpl);

}  // namespace

#if !UCONFIG_ONLY_HTML_CONVERSION
/************* KR ***************/
static const UConverterImpl _ISO2022KRImpl={
    UCNV_ISO_2022,

    nullptr,
    nullptr,

    _ISO2022Open,
    _ISO2022Close,
    _ISO2022Reset,

    UConverter_toUnicode_ISO_2022_KR_OFFSETS_LOGIC,
    UConverter_toUnicode_ISO_2022_KR_OFFSETS_LOGIC,
    UConverter_fromUnicode_ISO_2022_KR_OFFSETS_LOGIC,
    UConverter_fromUnicode_ISO_2022_KR_OFFSETS_LOGIC,
    nullptr,

    nullptr,
    _ISO2022getName,
    _ISO_2022_WriteSub,
    _ISO_2022_SafeClone,
    _ISO_2022_GetUnicodeSet,

    nullptr,
    nullptr
};
static const UConverterStaticData _ISO2022KRStaticData={
    sizeof(UConverterStaticData),
    "ISO_2022_KR",
    0,
    UCNV_IBM,
    UCNV_ISO_2022,
    1,
    8, /* max 8 bytes per char16_t */
    { 0x1a, 0, 0, 0 },
    1,
    false,
    false,
    0,
    0,
    { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } /* reserved */
};

namespace {

const UConverterSharedData _ISO2022KRData=
        UCNV_IMMUTABLE_SHARED_DATA_INITIALIZER(&_ISO2022KRStaticData, &_ISO2022KRImpl);

}  // namespace

/*************** CN ***************/
static const UConverterImpl _ISO2022CNImpl={

    UCNV_ISO_2022,

    nullptr,
    nullptr,

    _ISO2022Open,
    _ISO2022Close,
    _ISO2022Reset,

    UConverter_toUnicode_ISO_2022_CN_OFFSETS_LOGIC,
    UConverter_toUnicode_ISO_2022_CN_OFFSETS_LOGIC,
    UConverter_fromUnicode_ISO_2022_CN_OFFSETS_LOGIC,
    UConverter_fromUnicode_ISO_2022_CN_OFFSETS_LOGIC,
    nullptr,

    nullptr,
    _ISO2022getName,
    _ISO_2022_WriteSub,
    _ISO_2022_SafeClone,
    _ISO_2022_GetUnicodeSet,

    nullptr,
    nullptr
};
static const UConverterStaticData _ISO2022CNStaticData={
    sizeof(UConverterStaticData),
    "ISO_2022_CN",
    0,
    UCNV_IBM,
    UCNV_ISO_2022,
    1,
    8, /* max 8 bytes per char16_t: 4-byte CNS designator + 2 bytes for SS2/SS3 + DBCS */
    { 0x1a, 0, 0, 0 },
    1,
    false,
    false,
    0,
    0,
    { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } /* reserved */
};

namespace {

const UConverterSharedData _ISO2022CNData=
        UCNV_IMMUTABLE_SHARED_DATA_INITIALIZER(&_ISO2022CNStaticData, &_ISO2022CNImpl);

}  // namespace
#endif /* #if !UCONFIG_ONLY_HTML_CONVERSION */

#endif /* #if !UCONFIG_NO_LEGACY_CONVERSION */