diff options
author | William Joye <wjoye@cfa.harvard.edu> | 2016-11-17 21:10:17 (GMT) |
---|---|---|
committer | William Joye <wjoye@cfa.harvard.edu> | 2016-11-17 21:10:17 (GMT) |
commit | 8096d34300076a1aa9cb517de49fb920a051939f (patch) | |
tree | 5f2b1b7c41f89fedc31af973d6a747ca674cfd24 /libxml2/encoding.c | |
parent | fc7f7edd0b8011cb71573b15462ef83068d9e54b (diff) | |
parent | 574585fa78070b0cc6b5ad22543e21a3502a122b (diff) | |
download | blt-8096d34300076a1aa9cb517de49fb920a051939f.zip blt-8096d34300076a1aa9cb517de49fb920a051939f.tar.gz blt-8096d34300076a1aa9cb517de49fb920a051939f.tar.bz2 |
Merge commit '574585fa78070b0cc6b5ad22543e21a3502a122b' as 'libxml2'
Diffstat (limited to 'libxml2/encoding.c')
-rw-r--r-- | libxml2/encoding.c | 4085 |
1 files changed, 4085 insertions, 0 deletions
diff --git a/libxml2/encoding.c b/libxml2/encoding.c new file mode 100644 index 0000000..e49c7f8 --- /dev/null +++ b/libxml2/encoding.c @@ -0,0 +1,4085 @@ +/* + * encoding.c : implements the encoding conversion functions needed for XML + * + * Related specs: + * rfc2044 (UTF-8 and UTF-16) F. Yergeau Alis Technologies + * rfc2781 UTF-16, an encoding of ISO 10646, P. Hoffman, F. Yergeau + * [ISO-10646] UTF-8 and UTF-16 in Annexes + * [ISO-8859-1] ISO Latin-1 characters codes. + * [UNICODE] The Unicode Consortium, "The Unicode Standard -- + * Worldwide Character Encoding -- Version 1.0", Addison- + * Wesley, Volume 1, 1991, Volume 2, 1992. UTF-8 is + * described in Unicode Technical Report #4. + * [US-ASCII] Coded Character Set--7-bit American Standard Code for + * Information Interchange, ANSI X3.4-1986. + * + * See Copyright for the status of this software. + * + * daniel@veillard.com + * + * Original code for IsoLatin1 and UTF-16 by "Martin J. Duerst" <duerst@w3.org> + */ + +#define IN_LIBXML +#include "libxml.h" + +#include <string.h> +#include <limits.h> + +#ifdef HAVE_CTYPE_H +#include <ctype.h> +#endif +#ifdef HAVE_STDLIB_H +#include <stdlib.h> +#endif +#ifdef LIBXML_ICONV_ENABLED +#ifdef HAVE_ERRNO_H +#include <errno.h> +#endif +#endif +#include <libxml/encoding.h> +#include <libxml/xmlmemory.h> +#ifdef LIBXML_HTML_ENABLED +#include <libxml/HTMLparser.h> +#endif +#include <libxml/globals.h> +#include <libxml/xmlerror.h> + +#include "buf.h" +#include "enc.h" + +static xmlCharEncodingHandlerPtr xmlUTF16LEHandler = NULL; +static xmlCharEncodingHandlerPtr xmlUTF16BEHandler = NULL; + +typedef struct _xmlCharEncodingAlias xmlCharEncodingAlias; +typedef xmlCharEncodingAlias *xmlCharEncodingAliasPtr; +struct _xmlCharEncodingAlias { + const char *name; + const char *alias; +}; + +static xmlCharEncodingAliasPtr xmlCharEncodingAliases = NULL; +static int xmlCharEncodingAliasesNb = 0; +static int xmlCharEncodingAliasesMax = 0; + +#if defined(LIBXML_ICONV_ENABLED) || defined(LIBXML_ICU_ENABLED) +#if 0 +#define DEBUG_ENCODING /* Define this to get encoding traces */ +#endif +#else +#ifdef LIBXML_ISO8859X_ENABLED +static void xmlRegisterCharEncodingHandlersISO8859x (void); +#endif +#endif + +static int xmlLittleEndian = 1; + +/** + * xmlEncodingErrMemory: + * @extra: extra informations + * + * Handle an out of memory condition + */ +static void +xmlEncodingErrMemory(const char *extra) +{ + __xmlSimpleError(XML_FROM_I18N, XML_ERR_NO_MEMORY, NULL, NULL, extra); +} + +/** + * xmlErrEncoding: + * @error: the error number + * @msg: the error message + * + * n encoding error + */ +static void LIBXML_ATTR_FORMAT(2,0) +xmlEncodingErr(xmlParserErrors error, const char *msg, const char *val) +{ + __xmlRaiseError(NULL, NULL, NULL, NULL, NULL, + XML_FROM_I18N, error, XML_ERR_FATAL, + NULL, 0, val, NULL, NULL, 0, 0, msg, val); +} + +#ifdef LIBXML_ICU_ENABLED +static uconv_t* +openIcuConverter(const char* name, int toUnicode) +{ + UErrorCode status = U_ZERO_ERROR; + uconv_t *conv = (uconv_t *) xmlMalloc(sizeof(uconv_t)); + if (conv == NULL) + return NULL; + + conv->uconv = ucnv_open(name, &status); + if (U_FAILURE(status)) + goto error; + + status = U_ZERO_ERROR; + if (toUnicode) { + ucnv_setToUCallBack(conv->uconv, UCNV_TO_U_CALLBACK_STOP, + NULL, NULL, NULL, &status); + } + else { + ucnv_setFromUCallBack(conv->uconv, UCNV_FROM_U_CALLBACK_STOP, + NULL, NULL, NULL, &status); + } + if (U_FAILURE(status)) + goto error; + + status = U_ZERO_ERROR; + conv->utf8 = ucnv_open("UTF-8", &status); + if (U_SUCCESS(status)) + return conv; + +error: + if (conv->uconv) + ucnv_close(conv->uconv); + xmlFree(conv); + return NULL; +} + +static void +closeIcuConverter(uconv_t *conv) +{ + if (conv != NULL) { + ucnv_close(conv->uconv); + ucnv_close(conv->utf8); + xmlFree(conv); + } +} +#endif /* LIBXML_ICU_ENABLED */ + +/************************************************************************ + * * + * Conversions To/From UTF8 encoding * + * * + ************************************************************************/ + +/** + * asciiToUTF8: + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @in: a pointer to an array of ASCII chars + * @inlen: the length of @in + * + * Take a block of ASCII chars in and try to convert it to an UTF-8 + * block of chars out. + * Returns 0 if success, or -1 otherwise + * The value of @inlen after return is the number of octets consumed + * if the return value is positive, else unpredictable. + * The value of @outlen after return is the number of octets consumed. + */ +static int +asciiToUTF8(unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + unsigned char* outstart = out; + const unsigned char* base = in; + const unsigned char* processed = in; + unsigned char* outend = out + *outlen; + const unsigned char* inend; + unsigned int c; + + inend = in + (*inlen); + while ((in < inend) && (out - outstart + 5 < *outlen)) { + c= *in++; + + if (out >= outend) + break; + if (c < 0x80) { + *out++ = c; + } else { + *outlen = out - outstart; + *inlen = processed - base; + return(-1); + } + + processed = (const unsigned char*) in; + } + *outlen = out - outstart; + *inlen = processed - base; + return(*outlen); +} + +#ifdef LIBXML_OUTPUT_ENABLED +/** + * UTF8Toascii: + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @in: a pointer to an array of UTF-8 chars + * @inlen: the length of @in + * + * Take a block of UTF-8 chars in and try to convert it to an ASCII + * block of chars out. + * + * Returns 0 if success, -2 if the transcoding fails, or -1 otherwise + * The value of @inlen after return is the number of octets consumed + * if the return value is positive, else unpredictable. + * The value of @outlen after return is the number of octets consumed. + */ +static int +UTF8Toascii(unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + const unsigned char* processed = in; + const unsigned char* outend; + const unsigned char* outstart = out; + const unsigned char* instart = in; + const unsigned char* inend; + unsigned int c, d; + int trailing; + + if ((out == NULL) || (outlen == NULL) || (inlen == NULL)) return(-1); + if (in == NULL) { + /* + * initialization nothing to do + */ + *outlen = 0; + *inlen = 0; + return(0); + } + inend = in + (*inlen); + outend = out + (*outlen); + while (in < inend) { + d = *in++; + if (d < 0x80) { c= d; trailing= 0; } + else if (d < 0xC0) { + /* trailing byte in leading position */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } else if (d < 0xE0) { c= d & 0x1F; trailing= 1; } + else if (d < 0xF0) { c= d & 0x0F; trailing= 2; } + else if (d < 0xF8) { c= d & 0x07; trailing= 3; } + else { + /* no chance for this in Ascii */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + + if (inend - in < trailing) { + break; + } + + for ( ; trailing; trailing--) { + if ((in >= inend) || (((d= *in++) & 0xC0) != 0x80)) + break; + c <<= 6; + c |= d & 0x3F; + } + + /* assertion: c is a single UTF-4 value */ + if (c < 0x80) { + if (out >= outend) + break; + *out++ = c; + } else { + /* no chance for this in Ascii */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + processed = in; + } + *outlen = out - outstart; + *inlen = processed - instart; + return(*outlen); +} +#endif /* LIBXML_OUTPUT_ENABLED */ + +/** + * isolat1ToUTF8: + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @in: a pointer to an array of ISO Latin 1 chars + * @inlen: the length of @in + * + * Take a block of ISO Latin 1 chars in and try to convert it to an UTF-8 + * block of chars out. + * Returns the number of bytes written if success, or -1 otherwise + * The value of @inlen after return is the number of octets consumed + * if the return value is positive, else unpredictable. + * The value of @outlen after return is the number of octets consumed. + */ +int +isolat1ToUTF8(unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + unsigned char* outstart = out; + const unsigned char* base = in; + unsigned char* outend; + const unsigned char* inend; + const unsigned char* instop; + + if ((out == NULL) || (in == NULL) || (outlen == NULL) || (inlen == NULL)) + return(-1); + + outend = out + *outlen; + inend = in + (*inlen); + instop = inend; + + while ((in < inend) && (out < outend - 1)) { + if (*in >= 0x80) { + *out++ = (((*in) >> 6) & 0x1F) | 0xC0; + *out++ = ((*in) & 0x3F) | 0x80; + ++in; + } + if ((instop - in) > (outend - out)) instop = in + (outend - out); + while ((in < instop) && (*in < 0x80)) { + *out++ = *in++; + } + } + if ((in < inend) && (out < outend) && (*in < 0x80)) { + *out++ = *in++; + } + *outlen = out - outstart; + *inlen = in - base; + return(*outlen); +} + +/** + * UTF8ToUTF8: + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @inb: a pointer to an array of UTF-8 chars + * @inlenb: the length of @in in UTF-8 chars + * + * No op copy operation for UTF8 handling. + * + * Returns the number of bytes written, or -1 if lack of space. + * The value of *inlen after return is the number of octets consumed + * if the return value is positive, else unpredictable. + */ +static int +UTF8ToUTF8(unsigned char* out, int *outlen, + const unsigned char* inb, int *inlenb) +{ + int len; + + if ((out == NULL) || (inb == NULL) || (outlen == NULL) || (inlenb == NULL)) + return(-1); + if (*outlen > *inlenb) { + len = *inlenb; + } else { + len = *outlen; + } + if (len < 0) + return(-1); + + memcpy(out, inb, len); + + *outlen = len; + *inlenb = len; + return(*outlen); +} + + +#ifdef LIBXML_OUTPUT_ENABLED +/** + * UTF8Toisolat1: + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @in: a pointer to an array of UTF-8 chars + * @inlen: the length of @in + * + * Take a block of UTF-8 chars in and try to convert it to an ISO Latin 1 + * block of chars out. + * + * Returns the number of bytes written if success, -2 if the transcoding fails, + or -1 otherwise + * The value of @inlen after return is the number of octets consumed + * if the return value is positive, else unpredictable. + * The value of @outlen after return is the number of octets consumed. + */ +int +UTF8Toisolat1(unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + const unsigned char* processed = in; + const unsigned char* outend; + const unsigned char* outstart = out; + const unsigned char* instart = in; + const unsigned char* inend; + unsigned int c, d; + int trailing; + + if ((out == NULL) || (outlen == NULL) || (inlen == NULL)) return(-1); + if (in == NULL) { + /* + * initialization nothing to do + */ + *outlen = 0; + *inlen = 0; + return(0); + } + inend = in + (*inlen); + outend = out + (*outlen); + while (in < inend) { + d = *in++; + if (d < 0x80) { c= d; trailing= 0; } + else if (d < 0xC0) { + /* trailing byte in leading position */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } else if (d < 0xE0) { c= d & 0x1F; trailing= 1; } + else if (d < 0xF0) { c= d & 0x0F; trailing= 2; } + else if (d < 0xF8) { c= d & 0x07; trailing= 3; } + else { + /* no chance for this in IsoLat1 */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + + if (inend - in < trailing) { + break; + } + + for ( ; trailing; trailing--) { + if (in >= inend) + break; + if (((d= *in++) & 0xC0) != 0x80) { + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + c <<= 6; + c |= d & 0x3F; + } + + /* assertion: c is a single UTF-4 value */ + if (c <= 0xFF) { + if (out >= outend) + break; + *out++ = c; + } else { + /* no chance for this in IsoLat1 */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + processed = in; + } + *outlen = out - outstart; + *inlen = processed - instart; + return(*outlen); +} +#endif /* LIBXML_OUTPUT_ENABLED */ + +/** + * UTF16LEToUTF8: + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @inb: a pointer to an array of UTF-16LE passwd as a byte array + * @inlenb: the length of @in in UTF-16LE chars + * + * Take a block of UTF-16LE ushorts in and try to convert it to an UTF-8 + * block of chars out. This function assumes the endian property + * is the same between the native type of this machine and the + * inputed one. + * + * Returns the number of bytes written, or -1 if lack of space, or -2 + * if the transcoding fails (if *in is not a valid utf16 string) + * The value of *inlen after return is the number of octets consumed + * if the return value is positive, else unpredictable. + */ +static int +UTF16LEToUTF8(unsigned char* out, int *outlen, + const unsigned char* inb, int *inlenb) +{ + unsigned char* outstart = out; + const unsigned char* processed = inb; + unsigned char* outend = out + *outlen; + unsigned short* in = (unsigned short*) inb; + unsigned short* inend; + unsigned int c, d, inlen; + unsigned char *tmp; + int bits; + + if ((*inlenb % 2) == 1) + (*inlenb)--; + inlen = *inlenb / 2; + inend = in + inlen; + while ((in < inend) && (out - outstart + 5 < *outlen)) { + if (xmlLittleEndian) { + c= *in++; + } else { + tmp = (unsigned char *) in; + c = *tmp++; + c = c | (((unsigned int)*tmp) << 8); + in++; + } + if ((c & 0xFC00) == 0xD800) { /* surrogates */ + if (in >= inend) { /* (in > inend) shouldn't happens */ + break; + } + if (xmlLittleEndian) { + d = *in++; + } else { + tmp = (unsigned char *) in; + d = *tmp++; + d = d | (((unsigned int)*tmp) << 8); + in++; + } + if ((d & 0xFC00) == 0xDC00) { + c &= 0x03FF; + c <<= 10; + c |= d & 0x03FF; + c += 0x10000; + } + else { + *outlen = out - outstart; + *inlenb = processed - inb; + return(-2); + } + } + + /* assertion: c is a single UTF-4 value */ + if (out >= outend) + break; + if (c < 0x80) { *out++= c; bits= -6; } + else if (c < 0x800) { *out++= ((c >> 6) & 0x1F) | 0xC0; bits= 0; } + else if (c < 0x10000) { *out++= ((c >> 12) & 0x0F) | 0xE0; bits= 6; } + else { *out++= ((c >> 18) & 0x07) | 0xF0; bits= 12; } + + for ( ; bits >= 0; bits-= 6) { + if (out >= outend) + break; + *out++= ((c >> bits) & 0x3F) | 0x80; + } + processed = (const unsigned char*) in; + } + *outlen = out - outstart; + *inlenb = processed - inb; + return(*outlen); +} + +#ifdef LIBXML_OUTPUT_ENABLED +/** + * UTF8ToUTF16LE: + * @outb: a pointer to an array of bytes to store the result + * @outlen: the length of @outb + * @in: a pointer to an array of UTF-8 chars + * @inlen: the length of @in + * + * Take a block of UTF-8 chars in and try to convert it to an UTF-16LE + * block of chars out. + * + * Returns the number of bytes written, or -1 if lack of space, or -2 + * if the transcoding failed. + */ +static int +UTF8ToUTF16LE(unsigned char* outb, int *outlen, + const unsigned char* in, int *inlen) +{ + unsigned short* out = (unsigned short*) outb; + const unsigned char* processed = in; + const unsigned char *const instart = in; + unsigned short* outstart= out; + unsigned short* outend; + const unsigned char* inend; + unsigned int c, d; + int trailing; + unsigned char *tmp; + unsigned short tmp1, tmp2; + + /* UTF16LE encoding has no BOM */ + if ((out == NULL) || (outlen == NULL) || (inlen == NULL)) return(-1); + if (in == NULL) { + *outlen = 0; + *inlen = 0; + return(0); + } + inend= in + *inlen; + outend = out + (*outlen / 2); + while (in < inend) { + d= *in++; + if (d < 0x80) { c= d; trailing= 0; } + else if (d < 0xC0) { + /* trailing byte in leading position */ + *outlen = (out - outstart) * 2; + *inlen = processed - instart; + return(-2); + } else if (d < 0xE0) { c= d & 0x1F; trailing= 1; } + else if (d < 0xF0) { c= d & 0x0F; trailing= 2; } + else if (d < 0xF8) { c= d & 0x07; trailing= 3; } + else { + /* no chance for this in UTF-16 */ + *outlen = (out - outstart) * 2; + *inlen = processed - instart; + return(-2); + } + + if (inend - in < trailing) { + break; + } + + for ( ; trailing; trailing--) { + if ((in >= inend) || (((d= *in++) & 0xC0) != 0x80)) + break; + c <<= 6; + c |= d & 0x3F; + } + + /* assertion: c is a single UTF-4 value */ + if (c < 0x10000) { + if (out >= outend) + break; + if (xmlLittleEndian) { + *out++ = c; + } else { + tmp = (unsigned char *) out; + *tmp = c ; + *(tmp + 1) = c >> 8 ; + out++; + } + } + else if (c < 0x110000) { + if (out+1 >= outend) + break; + c -= 0x10000; + if (xmlLittleEndian) { + *out++ = 0xD800 | (c >> 10); + *out++ = 0xDC00 | (c & 0x03FF); + } else { + tmp1 = 0xD800 | (c >> 10); + tmp = (unsigned char *) out; + *tmp = (unsigned char) tmp1; + *(tmp + 1) = tmp1 >> 8; + out++; + + tmp2 = 0xDC00 | (c & 0x03FF); + tmp = (unsigned char *) out; + *tmp = (unsigned char) tmp2; + *(tmp + 1) = tmp2 >> 8; + out++; + } + } + else + break; + processed = in; + } + *outlen = (out - outstart) * 2; + *inlen = processed - instart; + return(*outlen); +} + +/** + * UTF8ToUTF16: + * @outb: a pointer to an array of bytes to store the result + * @outlen: the length of @outb + * @in: a pointer to an array of UTF-8 chars + * @inlen: the length of @in + * + * Take a block of UTF-8 chars in and try to convert it to an UTF-16 + * block of chars out. + * + * Returns the number of bytes written, or -1 if lack of space, or -2 + * if the transcoding failed. + */ +static int +UTF8ToUTF16(unsigned char* outb, int *outlen, + const unsigned char* in, int *inlen) +{ + if (in == NULL) { + /* + * initialization, add the Byte Order Mark for UTF-16LE + */ + if (*outlen >= 2) { + outb[0] = 0xFF; + outb[1] = 0xFE; + *outlen = 2; + *inlen = 0; +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "Added FFFE Byte Order Mark\n"); +#endif + return(2); + } + *outlen = 0; + *inlen = 0; + return(0); + } + return (UTF8ToUTF16LE(outb, outlen, in, inlen)); +} +#endif /* LIBXML_OUTPUT_ENABLED */ + +/** + * UTF16BEToUTF8: + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @inb: a pointer to an array of UTF-16 passed as a byte array + * @inlenb: the length of @in in UTF-16 chars + * + * Take a block of UTF-16 ushorts in and try to convert it to an UTF-8 + * block of chars out. This function assumes the endian property + * is the same between the native type of this machine and the + * inputed one. + * + * Returns the number of bytes written, or -1 if lack of space, or -2 + * if the transcoding fails (if *in is not a valid utf16 string) + * The value of *inlen after return is the number of octets consumed + * if the return value is positive, else unpredictable. + */ +static int +UTF16BEToUTF8(unsigned char* out, int *outlen, + const unsigned char* inb, int *inlenb) +{ + unsigned char* outstart = out; + const unsigned char* processed = inb; + unsigned char* outend = out + *outlen; + unsigned short* in = (unsigned short*) inb; + unsigned short* inend; + unsigned int c, d, inlen; + unsigned char *tmp; + int bits; + + if ((*inlenb % 2) == 1) + (*inlenb)--; + inlen = *inlenb / 2; + inend= in + inlen; + while (in < inend) { + if (xmlLittleEndian) { + tmp = (unsigned char *) in; + c = *tmp++; + c = c << 8; + c = c | (unsigned int) *tmp; + in++; + } else { + c= *in++; + } + if ((c & 0xFC00) == 0xD800) { /* surrogates */ + if (in >= inend) { /* (in > inend) shouldn't happens */ + *outlen = out - outstart; + *inlenb = processed - inb; + return(-2); + } + if (xmlLittleEndian) { + tmp = (unsigned char *) in; + d = *tmp++; + d = d << 8; + d = d | (unsigned int) *tmp; + in++; + } else { + d= *in++; + } + if ((d & 0xFC00) == 0xDC00) { + c &= 0x03FF; + c <<= 10; + c |= d & 0x03FF; + c += 0x10000; + } + else { + *outlen = out - outstart; + *inlenb = processed - inb; + return(-2); + } + } + + /* assertion: c is a single UTF-4 value */ + if (out >= outend) + break; + if (c < 0x80) { *out++= c; bits= -6; } + else if (c < 0x800) { *out++= ((c >> 6) & 0x1F) | 0xC0; bits= 0; } + else if (c < 0x10000) { *out++= ((c >> 12) & 0x0F) | 0xE0; bits= 6; } + else { *out++= ((c >> 18) & 0x07) | 0xF0; bits= 12; } + + for ( ; bits >= 0; bits-= 6) { + if (out >= outend) + break; + *out++= ((c >> bits) & 0x3F) | 0x80; + } + processed = (const unsigned char*) in; + } + *outlen = out - outstart; + *inlenb = processed - inb; + return(*outlen); +} + +#ifdef LIBXML_OUTPUT_ENABLED +/** + * UTF8ToUTF16BE: + * @outb: a pointer to an array of bytes to store the result + * @outlen: the length of @outb + * @in: a pointer to an array of UTF-8 chars + * @inlen: the length of @in + * + * Take a block of UTF-8 chars in and try to convert it to an UTF-16BE + * block of chars out. + * + * Returns the number of byte written, or -1 by lack of space, or -2 + * if the transcoding failed. + */ +static int +UTF8ToUTF16BE(unsigned char* outb, int *outlen, + const unsigned char* in, int *inlen) +{ + unsigned short* out = (unsigned short*) outb; + const unsigned char* processed = in; + const unsigned char *const instart = in; + unsigned short* outstart= out; + unsigned short* outend; + const unsigned char* inend; + unsigned int c, d; + int trailing; + unsigned char *tmp; + unsigned short tmp1, tmp2; + + /* UTF-16BE has no BOM */ + if ((outb == NULL) || (outlen == NULL) || (inlen == NULL)) return(-1); + if (in == NULL) { + *outlen = 0; + *inlen = 0; + return(0); + } + inend= in + *inlen; + outend = out + (*outlen / 2); + while (in < inend) { + d= *in++; + if (d < 0x80) { c= d; trailing= 0; } + else if (d < 0xC0) { + /* trailing byte in leading position */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } else if (d < 0xE0) { c= d & 0x1F; trailing= 1; } + else if (d < 0xF0) { c= d & 0x0F; trailing= 2; } + else if (d < 0xF8) { c= d & 0x07; trailing= 3; } + else { + /* no chance for this in UTF-16 */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + + if (inend - in < trailing) { + break; + } + + for ( ; trailing; trailing--) { + if ((in >= inend) || (((d= *in++) & 0xC0) != 0x80)) break; + c <<= 6; + c |= d & 0x3F; + } + + /* assertion: c is a single UTF-4 value */ + if (c < 0x10000) { + if (out >= outend) break; + if (xmlLittleEndian) { + tmp = (unsigned char *) out; + *tmp = c >> 8; + *(tmp + 1) = c; + out++; + } else { + *out++ = c; + } + } + else if (c < 0x110000) { + if (out+1 >= outend) break; + c -= 0x10000; + if (xmlLittleEndian) { + tmp1 = 0xD800 | (c >> 10); + tmp = (unsigned char *) out; + *tmp = tmp1 >> 8; + *(tmp + 1) = (unsigned char) tmp1; + out++; + + tmp2 = 0xDC00 | (c & 0x03FF); + tmp = (unsigned char *) out; + *tmp = tmp2 >> 8; + *(tmp + 1) = (unsigned char) tmp2; + out++; + } else { + *out++ = 0xD800 | (c >> 10); + *out++ = 0xDC00 | (c & 0x03FF); + } + } + else + break; + processed = in; + } + *outlen = (out - outstart) * 2; + *inlen = processed - instart; + return(*outlen); +} +#endif /* LIBXML_OUTPUT_ENABLED */ + +/************************************************************************ + * * + * Generic encoding handling routines * + * * + ************************************************************************/ + +/** + * xmlDetectCharEncoding: + * @in: a pointer to the first bytes of the XML entity, must be at least + * 2 bytes long (at least 4 if encoding is UTF4 variant). + * @len: pointer to the length of the buffer + * + * Guess the encoding of the entity using the first bytes of the entity content + * according to the non-normative appendix F of the XML-1.0 recommendation. + * + * Returns one of the XML_CHAR_ENCODING_... values. + */ +xmlCharEncoding +xmlDetectCharEncoding(const unsigned char* in, int len) +{ + if (in == NULL) + return(XML_CHAR_ENCODING_NONE); + if (len >= 4) { + if ((in[0] == 0x00) && (in[1] == 0x00) && + (in[2] == 0x00) && (in[3] == 0x3C)) + return(XML_CHAR_ENCODING_UCS4BE); + if ((in[0] == 0x3C) && (in[1] == 0x00) && + (in[2] == 0x00) && (in[3] == 0x00)) + return(XML_CHAR_ENCODING_UCS4LE); + if ((in[0] == 0x00) && (in[1] == 0x00) && + (in[2] == 0x3C) && (in[3] == 0x00)) + return(XML_CHAR_ENCODING_UCS4_2143); + if ((in[0] == 0x00) && (in[1] == 0x3C) && + (in[2] == 0x00) && (in[3] == 0x00)) + return(XML_CHAR_ENCODING_UCS4_3412); + if ((in[0] == 0x4C) && (in[1] == 0x6F) && + (in[2] == 0xA7) && (in[3] == 0x94)) + return(XML_CHAR_ENCODING_EBCDIC); + if ((in[0] == 0x3C) && (in[1] == 0x3F) && + (in[2] == 0x78) && (in[3] == 0x6D)) + return(XML_CHAR_ENCODING_UTF8); + /* + * Although not part of the recommendation, we also + * attempt an "auto-recognition" of UTF-16LE and + * UTF-16BE encodings. + */ + if ((in[0] == 0x3C) && (in[1] == 0x00) && + (in[2] == 0x3F) && (in[3] == 0x00)) + return(XML_CHAR_ENCODING_UTF16LE); + if ((in[0] == 0x00) && (in[1] == 0x3C) && + (in[2] == 0x00) && (in[3] == 0x3F)) + return(XML_CHAR_ENCODING_UTF16BE); + } + if (len >= 3) { + /* + * Errata on XML-1.0 June 20 2001 + * We now allow an UTF8 encoded BOM + */ + if ((in[0] == 0xEF) && (in[1] == 0xBB) && + (in[2] == 0xBF)) + return(XML_CHAR_ENCODING_UTF8); + } + /* For UTF-16 we can recognize by the BOM */ + if (len >= 2) { + if ((in[0] == 0xFE) && (in[1] == 0xFF)) + return(XML_CHAR_ENCODING_UTF16BE); + if ((in[0] == 0xFF) && (in[1] == 0xFE)) + return(XML_CHAR_ENCODING_UTF16LE); + } + return(XML_CHAR_ENCODING_NONE); +} + +/** + * xmlCleanupEncodingAliases: + * + * Unregisters all aliases + */ +void +xmlCleanupEncodingAliases(void) { + int i; + + if (xmlCharEncodingAliases == NULL) + return; + + for (i = 0;i < xmlCharEncodingAliasesNb;i++) { + if (xmlCharEncodingAliases[i].name != NULL) + xmlFree((char *) xmlCharEncodingAliases[i].name); + if (xmlCharEncodingAliases[i].alias != NULL) + xmlFree((char *) xmlCharEncodingAliases[i].alias); + } + xmlCharEncodingAliasesNb = 0; + xmlCharEncodingAliasesMax = 0; + xmlFree(xmlCharEncodingAliases); + xmlCharEncodingAliases = NULL; +} + +/** + * xmlGetEncodingAlias: + * @alias: the alias name as parsed, in UTF-8 format (ASCII actually) + * + * Lookup an encoding name for the given alias. + * + * Returns NULL if not found, otherwise the original name + */ +const char * +xmlGetEncodingAlias(const char *alias) { + int i; + char upper[100]; + + if (alias == NULL) + return(NULL); + + if (xmlCharEncodingAliases == NULL) + return(NULL); + + for (i = 0;i < 99;i++) { + upper[i] = toupper(alias[i]); + if (upper[i] == 0) break; + } + upper[i] = 0; + + /* + * Walk down the list looking for a definition of the alias + */ + for (i = 0;i < xmlCharEncodingAliasesNb;i++) { + if (!strcmp(xmlCharEncodingAliases[i].alias, upper)) { + return(xmlCharEncodingAliases[i].name); + } + } + return(NULL); +} + +/** + * xmlAddEncodingAlias: + * @name: the encoding name as parsed, in UTF-8 format (ASCII actually) + * @alias: the alias name as parsed, in UTF-8 format (ASCII actually) + * + * Registers an alias @alias for an encoding named @name. Existing alias + * will be overwritten. + * + * Returns 0 in case of success, -1 in case of error + */ +int +xmlAddEncodingAlias(const char *name, const char *alias) { + int i; + char upper[100]; + + if ((name == NULL) || (alias == NULL)) + return(-1); + + for (i = 0;i < 99;i++) { + upper[i] = toupper(alias[i]); + if (upper[i] == 0) break; + } + upper[i] = 0; + + if (xmlCharEncodingAliases == NULL) { + xmlCharEncodingAliasesNb = 0; + xmlCharEncodingAliasesMax = 20; + xmlCharEncodingAliases = (xmlCharEncodingAliasPtr) + xmlMalloc(xmlCharEncodingAliasesMax * sizeof(xmlCharEncodingAlias)); + if (xmlCharEncodingAliases == NULL) + return(-1); + } else if (xmlCharEncodingAliasesNb >= xmlCharEncodingAliasesMax) { + xmlCharEncodingAliasesMax *= 2; + xmlCharEncodingAliases = (xmlCharEncodingAliasPtr) + xmlRealloc(xmlCharEncodingAliases, + xmlCharEncodingAliasesMax * sizeof(xmlCharEncodingAlias)); + } + /* + * Walk down the list looking for a definition of the alias + */ + for (i = 0;i < xmlCharEncodingAliasesNb;i++) { + if (!strcmp(xmlCharEncodingAliases[i].alias, upper)) { + /* + * Replace the definition. + */ + xmlFree((char *) xmlCharEncodingAliases[i].name); + xmlCharEncodingAliases[i].name = xmlMemStrdup(name); + return(0); + } + } + /* + * Add the definition + */ + xmlCharEncodingAliases[xmlCharEncodingAliasesNb].name = xmlMemStrdup(name); + xmlCharEncodingAliases[xmlCharEncodingAliasesNb].alias = xmlMemStrdup(upper); + xmlCharEncodingAliasesNb++; + return(0); +} + +/** + * xmlDelEncodingAlias: + * @alias: the alias name as parsed, in UTF-8 format (ASCII actually) + * + * Unregisters an encoding alias @alias + * + * Returns 0 in case of success, -1 in case of error + */ +int +xmlDelEncodingAlias(const char *alias) { + int i; + + if (alias == NULL) + return(-1); + + if (xmlCharEncodingAliases == NULL) + return(-1); + /* + * Walk down the list looking for a definition of the alias + */ + for (i = 0;i < xmlCharEncodingAliasesNb;i++) { + if (!strcmp(xmlCharEncodingAliases[i].alias, alias)) { + xmlFree((char *) xmlCharEncodingAliases[i].name); + xmlFree((char *) xmlCharEncodingAliases[i].alias); + xmlCharEncodingAliasesNb--; + memmove(&xmlCharEncodingAliases[i], &xmlCharEncodingAliases[i + 1], + sizeof(xmlCharEncodingAlias) * (xmlCharEncodingAliasesNb - i)); + return(0); + } + } + return(-1); +} + +/** + * xmlParseCharEncoding: + * @name: the encoding name as parsed, in UTF-8 format (ASCII actually) + * + * Compare the string to the encoding schemes already known. Note + * that the comparison is case insensitive accordingly to the section + * [XML] 4.3.3 Character Encoding in Entities. + * + * Returns one of the XML_CHAR_ENCODING_... values or XML_CHAR_ENCODING_NONE + * if not recognized. + */ +xmlCharEncoding +xmlParseCharEncoding(const char* name) +{ + const char *alias; + char upper[500]; + int i; + + if (name == NULL) + return(XML_CHAR_ENCODING_NONE); + + /* + * Do the alias resolution + */ + alias = xmlGetEncodingAlias(name); + if (alias != NULL) + name = alias; + + for (i = 0;i < 499;i++) { + upper[i] = toupper(name[i]); + if (upper[i] == 0) break; + } + upper[i] = 0; + + if (!strcmp(upper, "")) return(XML_CHAR_ENCODING_NONE); + if (!strcmp(upper, "UTF-8")) return(XML_CHAR_ENCODING_UTF8); + if (!strcmp(upper, "UTF8")) return(XML_CHAR_ENCODING_UTF8); + + /* + * NOTE: if we were able to parse this, the endianness of UTF16 is + * already found and in use + */ + if (!strcmp(upper, "UTF-16")) return(XML_CHAR_ENCODING_UTF16LE); + if (!strcmp(upper, "UTF16")) return(XML_CHAR_ENCODING_UTF16LE); + + if (!strcmp(upper, "ISO-10646-UCS-2")) return(XML_CHAR_ENCODING_UCS2); + if (!strcmp(upper, "UCS-2")) return(XML_CHAR_ENCODING_UCS2); + if (!strcmp(upper, "UCS2")) return(XML_CHAR_ENCODING_UCS2); + + /* + * NOTE: if we were able to parse this, the endianness of UCS4 is + * already found and in use + */ + if (!strcmp(upper, "ISO-10646-UCS-4")) return(XML_CHAR_ENCODING_UCS4LE); + if (!strcmp(upper, "UCS-4")) return(XML_CHAR_ENCODING_UCS4LE); + if (!strcmp(upper, "UCS4")) return(XML_CHAR_ENCODING_UCS4LE); + + + if (!strcmp(upper, "ISO-8859-1")) return(XML_CHAR_ENCODING_8859_1); + if (!strcmp(upper, "ISO-LATIN-1")) return(XML_CHAR_ENCODING_8859_1); + if (!strcmp(upper, "ISO LATIN 1")) return(XML_CHAR_ENCODING_8859_1); + + if (!strcmp(upper, "ISO-8859-2")) return(XML_CHAR_ENCODING_8859_2); + if (!strcmp(upper, "ISO-LATIN-2")) return(XML_CHAR_ENCODING_8859_2); + if (!strcmp(upper, "ISO LATIN 2")) return(XML_CHAR_ENCODING_8859_2); + + if (!strcmp(upper, "ISO-8859-3")) return(XML_CHAR_ENCODING_8859_3); + if (!strcmp(upper, "ISO-8859-4")) return(XML_CHAR_ENCODING_8859_4); + if (!strcmp(upper, "ISO-8859-5")) return(XML_CHAR_ENCODING_8859_5); + if (!strcmp(upper, "ISO-8859-6")) return(XML_CHAR_ENCODING_8859_6); + if (!strcmp(upper, "ISO-8859-7")) return(XML_CHAR_ENCODING_8859_7); + if (!strcmp(upper, "ISO-8859-8")) return(XML_CHAR_ENCODING_8859_8); + if (!strcmp(upper, "ISO-8859-9")) return(XML_CHAR_ENCODING_8859_9); + + if (!strcmp(upper, "ISO-2022-JP")) return(XML_CHAR_ENCODING_2022_JP); + if (!strcmp(upper, "SHIFT_JIS")) return(XML_CHAR_ENCODING_SHIFT_JIS); + if (!strcmp(upper, "EUC-JP")) return(XML_CHAR_ENCODING_EUC_JP); + +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, "Unknown encoding %s\n", name); +#endif + return(XML_CHAR_ENCODING_ERROR); +} + +/** + * xmlGetCharEncodingName: + * @enc: the encoding + * + * The "canonical" name for XML encoding. + * C.f. http://www.w3.org/TR/REC-xml#charencoding + * Section 4.3.3 Character Encoding in Entities + * + * Returns the canonical name for the given encoding + */ + +const char* +xmlGetCharEncodingName(xmlCharEncoding enc) { + switch (enc) { + case XML_CHAR_ENCODING_ERROR: + return(NULL); + case XML_CHAR_ENCODING_NONE: + return(NULL); + case XML_CHAR_ENCODING_UTF8: + return("UTF-8"); + case XML_CHAR_ENCODING_UTF16LE: + return("UTF-16"); + case XML_CHAR_ENCODING_UTF16BE: + return("UTF-16"); + case XML_CHAR_ENCODING_EBCDIC: + return("EBCDIC"); + case XML_CHAR_ENCODING_UCS4LE: + return("ISO-10646-UCS-4"); + case XML_CHAR_ENCODING_UCS4BE: + return("ISO-10646-UCS-4"); + case XML_CHAR_ENCODING_UCS4_2143: + return("ISO-10646-UCS-4"); + case XML_CHAR_ENCODING_UCS4_3412: + return("ISO-10646-UCS-4"); + case XML_CHAR_ENCODING_UCS2: + return("ISO-10646-UCS-2"); + case XML_CHAR_ENCODING_8859_1: + return("ISO-8859-1"); + case XML_CHAR_ENCODING_8859_2: + return("ISO-8859-2"); + case XML_CHAR_ENCODING_8859_3: + return("ISO-8859-3"); + case XML_CHAR_ENCODING_8859_4: + return("ISO-8859-4"); + case XML_CHAR_ENCODING_8859_5: + return("ISO-8859-5"); + case XML_CHAR_ENCODING_8859_6: + return("ISO-8859-6"); + case XML_CHAR_ENCODING_8859_7: + return("ISO-8859-7"); + case XML_CHAR_ENCODING_8859_8: + return("ISO-8859-8"); + case XML_CHAR_ENCODING_8859_9: + return("ISO-8859-9"); + case XML_CHAR_ENCODING_2022_JP: + return("ISO-2022-JP"); + case XML_CHAR_ENCODING_SHIFT_JIS: + return("Shift-JIS"); + case XML_CHAR_ENCODING_EUC_JP: + return("EUC-JP"); + case XML_CHAR_ENCODING_ASCII: + return(NULL); + } + return(NULL); +} + +/************************************************************************ + * * + * Char encoding handlers * + * * + ************************************************************************/ + + +/* the size should be growable, but it's not a big deal ... */ +#define MAX_ENCODING_HANDLERS 50 +static xmlCharEncodingHandlerPtr *handlers = NULL; +static int nbCharEncodingHandler = 0; + +/* + * The default is UTF-8 for XML, that's also the default used for the + * parser internals, so the default encoding handler is NULL + */ + +static xmlCharEncodingHandlerPtr xmlDefaultCharEncodingHandler = NULL; + +/** + * xmlNewCharEncodingHandler: + * @name: the encoding name, in UTF-8 format (ASCII actually) + * @input: the xmlCharEncodingInputFunc to read that encoding + * @output: the xmlCharEncodingOutputFunc to write that encoding + * + * Create and registers an xmlCharEncodingHandler. + * + * Returns the xmlCharEncodingHandlerPtr created (or NULL in case of error). + */ +xmlCharEncodingHandlerPtr +xmlNewCharEncodingHandler(const char *name, + xmlCharEncodingInputFunc input, + xmlCharEncodingOutputFunc output) { + xmlCharEncodingHandlerPtr handler; + const char *alias; + char upper[500]; + int i; + char *up = NULL; + + /* + * Do the alias resolution + */ + alias = xmlGetEncodingAlias(name); + if (alias != NULL) + name = alias; + + /* + * Keep only the uppercase version of the encoding. + */ + if (name == NULL) { + xmlEncodingErr(XML_I18N_NO_NAME, + "xmlNewCharEncodingHandler : no name !\n", NULL); + return(NULL); + } + for (i = 0;i < 499;i++) { + upper[i] = toupper(name[i]); + if (upper[i] == 0) break; + } + upper[i] = 0; + up = xmlMemStrdup(upper); + if (up == NULL) { + xmlEncodingErrMemory("xmlNewCharEncodingHandler : out of memory !\n"); + return(NULL); + } + + /* + * allocate and fill-up an handler block. + */ + handler = (xmlCharEncodingHandlerPtr) + xmlMalloc(sizeof(xmlCharEncodingHandler)); + if (handler == NULL) { + xmlFree(up); + xmlEncodingErrMemory("xmlNewCharEncodingHandler : out of memory !\n"); + return(NULL); + } + memset(handler, 0, sizeof(xmlCharEncodingHandler)); + handler->input = input; + handler->output = output; + handler->name = up; + +#ifdef LIBXML_ICONV_ENABLED + handler->iconv_in = NULL; + handler->iconv_out = NULL; +#endif +#ifdef LIBXML_ICU_ENABLED + handler->uconv_in = NULL; + handler->uconv_out = NULL; +#endif + + /* + * registers and returns the handler. + */ + xmlRegisterCharEncodingHandler(handler); +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "Registered encoding handler for %s\n", name); +#endif + return(handler); +} + +/** + * xmlInitCharEncodingHandlers: + * + * Initialize the char encoding support, it registers the default + * encoding supported. + * NOTE: while public, this function usually doesn't need to be called + * in normal processing. + */ +void +xmlInitCharEncodingHandlers(void) { + unsigned short int tst = 0x1234; + unsigned char *ptr = (unsigned char *) &tst; + + if (handlers != NULL) return; + + handlers = (xmlCharEncodingHandlerPtr *) + xmlMalloc(MAX_ENCODING_HANDLERS * sizeof(xmlCharEncodingHandlerPtr)); + + if (*ptr == 0x12) xmlLittleEndian = 0; + else if (*ptr == 0x34) xmlLittleEndian = 1; + else { + xmlEncodingErr(XML_ERR_INTERNAL_ERROR, + "Odd problem at endianness detection\n", NULL); + } + + if (handlers == NULL) { + xmlEncodingErrMemory("xmlInitCharEncodingHandlers : out of memory !\n"); + return; + } + xmlNewCharEncodingHandler("UTF-8", UTF8ToUTF8, UTF8ToUTF8); +#ifdef LIBXML_OUTPUT_ENABLED + xmlUTF16LEHandler = + xmlNewCharEncodingHandler("UTF-16LE", UTF16LEToUTF8, UTF8ToUTF16LE); + xmlUTF16BEHandler = + xmlNewCharEncodingHandler("UTF-16BE", UTF16BEToUTF8, UTF8ToUTF16BE); + xmlNewCharEncodingHandler("UTF-16", UTF16LEToUTF8, UTF8ToUTF16); + xmlNewCharEncodingHandler("ISO-8859-1", isolat1ToUTF8, UTF8Toisolat1); + xmlNewCharEncodingHandler("ASCII", asciiToUTF8, UTF8Toascii); + xmlNewCharEncodingHandler("US-ASCII", asciiToUTF8, UTF8Toascii); +#ifdef LIBXML_HTML_ENABLED + xmlNewCharEncodingHandler("HTML", NULL, UTF8ToHtml); +#endif +#else + xmlUTF16LEHandler = + xmlNewCharEncodingHandler("UTF-16LE", UTF16LEToUTF8, NULL); + xmlUTF16BEHandler = + xmlNewCharEncodingHandler("UTF-16BE", UTF16BEToUTF8, NULL); + xmlNewCharEncodingHandler("UTF-16", UTF16LEToUTF8, NULL); + xmlNewCharEncodingHandler("ISO-8859-1", isolat1ToUTF8, NULL); + xmlNewCharEncodingHandler("ASCII", asciiToUTF8, NULL); + xmlNewCharEncodingHandler("US-ASCII", asciiToUTF8, NULL); +#endif /* LIBXML_OUTPUT_ENABLED */ +#if !defined(LIBXML_ICONV_ENABLED) && !defined(LIBXML_ICU_ENABLED) +#ifdef LIBXML_ISO8859X_ENABLED + xmlRegisterCharEncodingHandlersISO8859x (); +#endif +#endif + +} + +/** + * xmlCleanupCharEncodingHandlers: + * + * Cleanup the memory allocated for the char encoding support, it + * unregisters all the encoding handlers and the aliases. + */ +void +xmlCleanupCharEncodingHandlers(void) { + xmlCleanupEncodingAliases(); + + if (handlers == NULL) return; + + for (;nbCharEncodingHandler > 0;) { + nbCharEncodingHandler--; + if (handlers[nbCharEncodingHandler] != NULL) { + if (handlers[nbCharEncodingHandler]->name != NULL) + xmlFree(handlers[nbCharEncodingHandler]->name); + xmlFree(handlers[nbCharEncodingHandler]); + } + } + xmlFree(handlers); + handlers = NULL; + nbCharEncodingHandler = 0; + xmlDefaultCharEncodingHandler = NULL; +} + +/** + * xmlRegisterCharEncodingHandler: + * @handler: the xmlCharEncodingHandlerPtr handler block + * + * Register the char encoding handler, surprising, isn't it ? + */ +void +xmlRegisterCharEncodingHandler(xmlCharEncodingHandlerPtr handler) { + if (handlers == NULL) xmlInitCharEncodingHandlers(); + if ((handler == NULL) || (handlers == NULL)) { + xmlEncodingErr(XML_I18N_NO_HANDLER, + "xmlRegisterCharEncodingHandler: NULL handler !\n", NULL); + return; + } + + if (nbCharEncodingHandler >= MAX_ENCODING_HANDLERS) { + xmlEncodingErr(XML_I18N_EXCESS_HANDLER, + "xmlRegisterCharEncodingHandler: Too many handler registered, see %s\n", + "MAX_ENCODING_HANDLERS"); + return; + } + handlers[nbCharEncodingHandler++] = handler; +} + +/** + * xmlGetCharEncodingHandler: + * @enc: an xmlCharEncoding value. + * + * Search in the registered set the handler able to read/write that encoding. + * + * Returns the handler or NULL if not found + */ +xmlCharEncodingHandlerPtr +xmlGetCharEncodingHandler(xmlCharEncoding enc) { + xmlCharEncodingHandlerPtr handler; + + if (handlers == NULL) xmlInitCharEncodingHandlers(); + switch (enc) { + case XML_CHAR_ENCODING_ERROR: + return(NULL); + case XML_CHAR_ENCODING_NONE: + return(NULL); + case XML_CHAR_ENCODING_UTF8: + return(NULL); + case XML_CHAR_ENCODING_UTF16LE: + return(xmlUTF16LEHandler); + case XML_CHAR_ENCODING_UTF16BE: + return(xmlUTF16BEHandler); + case XML_CHAR_ENCODING_EBCDIC: + handler = xmlFindCharEncodingHandler("EBCDIC"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("ebcdic"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("EBCDIC-US"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("IBM-037"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_UCS4BE: + handler = xmlFindCharEncodingHandler("ISO-10646-UCS-4"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("UCS-4"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("UCS4"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_UCS4LE: + handler = xmlFindCharEncodingHandler("ISO-10646-UCS-4"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("UCS-4"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("UCS4"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_UCS4_2143: + break; + case XML_CHAR_ENCODING_UCS4_3412: + break; + case XML_CHAR_ENCODING_UCS2: + handler = xmlFindCharEncodingHandler("ISO-10646-UCS-2"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("UCS-2"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("UCS2"); + if (handler != NULL) return(handler); + break; + + /* + * We used to keep ISO Latin encodings native in the + * generated data. This led to so many problems that + * this has been removed. One can still change this + * back by registering no-ops encoders for those + */ + case XML_CHAR_ENCODING_8859_1: + handler = xmlFindCharEncodingHandler("ISO-8859-1"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_8859_2: + handler = xmlFindCharEncodingHandler("ISO-8859-2"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_8859_3: + handler = xmlFindCharEncodingHandler("ISO-8859-3"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_8859_4: + handler = xmlFindCharEncodingHandler("ISO-8859-4"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_8859_5: + handler = xmlFindCharEncodingHandler("ISO-8859-5"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_8859_6: + handler = xmlFindCharEncodingHandler("ISO-8859-6"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_8859_7: + handler = xmlFindCharEncodingHandler("ISO-8859-7"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_8859_8: + handler = xmlFindCharEncodingHandler("ISO-8859-8"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_8859_9: + handler = xmlFindCharEncodingHandler("ISO-8859-9"); + if (handler != NULL) return(handler); + break; + + + case XML_CHAR_ENCODING_2022_JP: + handler = xmlFindCharEncodingHandler("ISO-2022-JP"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_SHIFT_JIS: + handler = xmlFindCharEncodingHandler("SHIFT-JIS"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("SHIFT_JIS"); + if (handler != NULL) return(handler); + handler = xmlFindCharEncodingHandler("Shift_JIS"); + if (handler != NULL) return(handler); + break; + case XML_CHAR_ENCODING_EUC_JP: + handler = xmlFindCharEncodingHandler("EUC-JP"); + if (handler != NULL) return(handler); + break; + default: + break; + } + +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "No handler found for encoding %d\n", enc); +#endif + return(NULL); +} + +/** + * xmlFindCharEncodingHandler: + * @name: a string describing the char encoding. + * + * Search in the registered set the handler able to read/write that encoding. + * + * Returns the handler or NULL if not found + */ +xmlCharEncodingHandlerPtr +xmlFindCharEncodingHandler(const char *name) { + const char *nalias; + const char *norig; + xmlCharEncoding alias; +#ifdef LIBXML_ICONV_ENABLED + xmlCharEncodingHandlerPtr enc; + iconv_t icv_in, icv_out; +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + xmlCharEncodingHandlerPtr encu; + uconv_t *ucv_in, *ucv_out; +#endif /* LIBXML_ICU_ENABLED */ + char upper[100]; + int i; + + if (handlers == NULL) xmlInitCharEncodingHandlers(); + if (name == NULL) return(xmlDefaultCharEncodingHandler); + if (name[0] == 0) return(xmlDefaultCharEncodingHandler); + + /* + * Do the alias resolution + */ + norig = name; + nalias = xmlGetEncodingAlias(name); + if (nalias != NULL) + name = nalias; + + /* + * Check first for directly registered encoding names + */ + for (i = 0;i < 99;i++) { + upper[i] = toupper(name[i]); + if (upper[i] == 0) break; + } + upper[i] = 0; + + if (handlers != NULL) { + for (i = 0;i < nbCharEncodingHandler; i++) { + if (!strcmp(upper, handlers[i]->name)) { +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "Found registered handler for encoding %s\n", name); +#endif + return(handlers[i]); + } + } + } + +#ifdef LIBXML_ICONV_ENABLED + /* check whether iconv can handle this */ + icv_in = iconv_open("UTF-8", name); + icv_out = iconv_open(name, "UTF-8"); + if (icv_in == (iconv_t) -1) { + icv_in = iconv_open("UTF-8", upper); + } + if (icv_out == (iconv_t) -1) { + icv_out = iconv_open(upper, "UTF-8"); + } + if ((icv_in != (iconv_t) -1) && (icv_out != (iconv_t) -1)) { + enc = (xmlCharEncodingHandlerPtr) + xmlMalloc(sizeof(xmlCharEncodingHandler)); + if (enc == NULL) { + iconv_close(icv_in); + iconv_close(icv_out); + return(NULL); + } + memset(enc, 0, sizeof(xmlCharEncodingHandler)); + enc->name = xmlMemStrdup(name); + enc->input = NULL; + enc->output = NULL; + enc->iconv_in = icv_in; + enc->iconv_out = icv_out; +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "Found iconv handler for encoding %s\n", name); +#endif + return enc; + } else if ((icv_in != (iconv_t) -1) || icv_out != (iconv_t) -1) { + xmlEncodingErr(XML_ERR_INTERNAL_ERROR, + "iconv : problems with filters for '%s'\n", name); + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + /* check whether icu can handle this */ + ucv_in = openIcuConverter(name, 1); + ucv_out = openIcuConverter(name, 0); + if (ucv_in != NULL && ucv_out != NULL) { + encu = (xmlCharEncodingHandlerPtr) + xmlMalloc(sizeof(xmlCharEncodingHandler)); + if (encu == NULL) { + closeIcuConverter(ucv_in); + closeIcuConverter(ucv_out); + return(NULL); + } + memset(encu, 0, sizeof(xmlCharEncodingHandler)); + encu->name = xmlMemStrdup(name); + encu->input = NULL; + encu->output = NULL; + encu->uconv_in = ucv_in; + encu->uconv_out = ucv_out; +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "Found ICU converter handler for encoding %s\n", name); +#endif + return encu; + } else if (ucv_in != NULL || ucv_out != NULL) { + closeIcuConverter(ucv_in); + closeIcuConverter(ucv_out); + xmlEncodingErr(XML_ERR_INTERNAL_ERROR, + "ICU converter : problems with filters for '%s'\n", name); + } +#endif /* LIBXML_ICU_ENABLED */ + +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "No handler found for encoding %s\n", name); +#endif + + /* + * Fallback using the canonical names + */ + alias = xmlParseCharEncoding(norig); + if (alias != XML_CHAR_ENCODING_ERROR) { + const char* canon; + canon = xmlGetCharEncodingName(alias); + if ((canon != NULL) && (strcmp(name, canon))) { + return(xmlFindCharEncodingHandler(canon)); + } + } + + /* If "none of the above", give up */ + return(NULL); +} + +/************************************************************************ + * * + * ICONV based generic conversion functions * + * * + ************************************************************************/ + +#ifdef LIBXML_ICONV_ENABLED +/** + * xmlIconvWrapper: + * @cd: iconv converter data structure + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @in: a pointer to an array of ISO Latin 1 chars + * @inlen: the length of @in + * + * Returns 0 if success, or + * -1 by lack of space, or + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + * -3 if there the last byte can't form a single output char. + * + * The value of @inlen after return is the number of octets consumed + * as the return value is positive, else unpredictable. + * The value of @outlen after return is the number of ocetes consumed. + */ +static int +xmlIconvWrapper(iconv_t cd, unsigned char *out, int *outlen, + const unsigned char *in, int *inlen) { + size_t icv_inlen, icv_outlen; + const char *icv_in = (const char *) in; + char *icv_out = (char *) out; + int ret; + + if ((out == NULL) || (outlen == NULL) || (inlen == NULL) || (in == NULL)) { + if (outlen != NULL) *outlen = 0; + return(-1); + } + icv_inlen = *inlen; + icv_outlen = *outlen; + ret = iconv(cd, (ICONV_CONST char **) &icv_in, &icv_inlen, &icv_out, &icv_outlen); + *inlen -= icv_inlen; + *outlen -= icv_outlen; + if ((icv_inlen != 0) || (ret == -1)) { +#ifdef EILSEQ + if (errno == EILSEQ) { + return -2; + } else +#endif +#ifdef E2BIG + if (errno == E2BIG) { + return -1; + } else +#endif +#ifdef EINVAL + if (errno == EINVAL) { + return -3; + } else +#endif + { + return -3; + } + } + return 0; +} +#endif /* LIBXML_ICONV_ENABLED */ + +/************************************************************************ + * * + * ICU based generic conversion functions * + * * + ************************************************************************/ + +#ifdef LIBXML_ICU_ENABLED +/** + * xmlUconvWrapper: + * @cd: ICU uconverter data structure + * @toUnicode : non-zero if toUnicode. 0 otherwise. + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @in: a pointer to an array of ISO Latin 1 chars + * @inlen: the length of @in + * + * Returns 0 if success, or + * -1 by lack of space, or + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + * -3 if there the last byte can't form a single output char. + * + * The value of @inlen after return is the number of octets consumed + * as the return value is positive, else unpredictable. + * The value of @outlen after return is the number of ocetes consumed. + */ +static int +xmlUconvWrapper(uconv_t *cd, int toUnicode, unsigned char *out, int *outlen, + const unsigned char *in, int *inlen) { + const char *ucv_in = (const char *) in; + char *ucv_out = (char *) out; + UErrorCode err = U_ZERO_ERROR; + + if ((out == NULL) || (outlen == NULL) || (inlen == NULL) || (in == NULL)) { + if (outlen != NULL) *outlen = 0; + return(-1); + } + + /* + * TODO(jungshik) + * 1. is ucnv_convert(To|From)Algorithmic better? + * 2. had we better use an explicit pivot buffer? + * 3. error returned comes from 'fromUnicode' only even + * when toUnicode is true ! + */ + if (toUnicode) { + /* encoding => UTF-16 => UTF-8 */ + ucnv_convertEx(cd->utf8, cd->uconv, &ucv_out, ucv_out + *outlen, + &ucv_in, ucv_in + *inlen, NULL, NULL, NULL, NULL, + 0, TRUE, &err); + } else { + /* UTF-8 => UTF-16 => encoding */ + ucnv_convertEx(cd->uconv, cd->utf8, &ucv_out, ucv_out + *outlen, + &ucv_in, ucv_in + *inlen, NULL, NULL, NULL, NULL, + 0, TRUE, &err); + } + *inlen = ucv_in - (const char*) in; + *outlen = ucv_out - (char *) out; + if (U_SUCCESS(err)) + return 0; + if (err == U_BUFFER_OVERFLOW_ERROR) + return -1; + if (err == U_INVALID_CHAR_FOUND || err == U_ILLEGAL_CHAR_FOUND) + return -2; + /* if (err == U_TRUNCATED_CHAR_FOUND) */ + return -3; +} +#endif /* LIBXML_ICU_ENABLED */ + +/************************************************************************ + * * + * The real API used by libxml for on-the-fly conversion * + * * + ************************************************************************/ + +/** + * xmlCharEncFirstLineInt: + * @handler: char enconding transformation data structure + * @out: an xmlBuffer for the output. + * @in: an xmlBuffer for the input + * @len: number of bytes to convert for the first line, or -1 + * + * Front-end for the encoding handler input function, but handle only + * the very first line, i.e. limit itself to 45 chars. + * + * Returns the number of byte written if success, or + * -1 general error + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + */ +int +xmlCharEncFirstLineInt(xmlCharEncodingHandler *handler, xmlBufferPtr out, + xmlBufferPtr in, int len) { + int ret = -2; + int written; + int toconv; + + if (handler == NULL) return(-1); + if (out == NULL) return(-1); + if (in == NULL) return(-1); + + /* calculate space available */ + written = out->size - out->use - 1; /* count '\0' */ + toconv = in->use; + /* + * echo '<?xml version="1.0" encoding="UCS4"?>' | wc -c => 38 + * 45 chars should be sufficient to reach the end of the encoding + * declaration without going too far inside the document content. + * on UTF-16 this means 90bytes, on UCS4 this means 180 + * The actual value depending on guessed encoding is passed as @len + * if provided + */ + if (len >= 0) { + if (toconv > len) + toconv = len; + } else { + if (toconv > 180) + toconv = 180; + } + if (toconv * 2 >= written) { + xmlBufferGrow(out, toconv * 2); + written = out->size - out->use - 1; + } + + if (handler->input != NULL) { + ret = handler->input(&out->content[out->use], &written, + in->content, &toconv); + xmlBufferShrink(in, toconv); + out->use += written; + out->content[out->use] = 0; + } +#ifdef LIBXML_ICONV_ENABLED + else if (handler->iconv_in != NULL) { + ret = xmlIconvWrapper(handler->iconv_in, &out->content[out->use], + &written, in->content, &toconv); + xmlBufferShrink(in, toconv); + out->use += written; + out->content[out->use] = 0; + if (ret == -1) ret = -3; + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + else if (handler->uconv_in != NULL) { + ret = xmlUconvWrapper(handler->uconv_in, 1, &out->content[out->use], + &written, in->content, &toconv); + xmlBufferShrink(in, toconv); + out->use += written; + out->content[out->use] = 0; + if (ret == -1) ret = -3; + } +#endif /* LIBXML_ICU_ENABLED */ +#ifdef DEBUG_ENCODING + switch (ret) { + case 0: + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input\n", + toconv, written); + break; + case -1: + xmlGenericError(xmlGenericErrorContext,"converted %d bytes to %d bytes of input, %d left\n", + toconv, written, in->use); + break; + case -2: + xmlGenericError(xmlGenericErrorContext, + "input conversion failed due to input error\n"); + break; + case -3: + xmlGenericError(xmlGenericErrorContext,"converted %d bytes to %d bytes of input, %d left\n", + toconv, written, in->use); + break; + default: + xmlGenericError(xmlGenericErrorContext,"Unknown input conversion failed %d\n", ret); + } +#endif /* DEBUG_ENCODING */ + /* + * Ignore when input buffer is not on a boundary + */ + if (ret == -3) ret = 0; + if (ret == -1) ret = 0; + return(ret); +} + +/** + * xmlCharEncFirstLine: + * @handler: char enconding transformation data structure + * @out: an xmlBuffer for the output. + * @in: an xmlBuffer for the input + * + * Front-end for the encoding handler input function, but handle only + * the very first line, i.e. limit itself to 45 chars. + * + * Returns the number of byte written if success, or + * -1 general error + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + */ +int +xmlCharEncFirstLine(xmlCharEncodingHandler *handler, xmlBufferPtr out, + xmlBufferPtr in) { + return(xmlCharEncFirstLineInt(handler, out, in, -1)); +} + +/** + * xmlCharEncFirstLineInput: + * @input: a parser input buffer + * @len: number of bytes to convert for the first line, or -1 + * + * Front-end for the encoding handler input function, but handle only + * the very first line. Point is that this is based on autodetection + * of the encoding and once that first line is converted we may find + * out that a different decoder is needed to process the input. + * + * Returns the number of byte written if success, or + * -1 general error + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + */ +int +xmlCharEncFirstLineInput(xmlParserInputBufferPtr input, int len) +{ + int ret = -2; + size_t written; + size_t toconv; + int c_in; + int c_out; + xmlBufPtr in; + xmlBufPtr out; + + if ((input == NULL) || (input->encoder == NULL) || + (input->buffer == NULL) || (input->raw == NULL)) + return (-1); + out = input->buffer; + in = input->raw; + + toconv = xmlBufUse(in); + if (toconv == 0) + return (0); + written = xmlBufAvail(out) - 1; /* count '\0' */ + /* + * echo '<?xml version="1.0" encoding="UCS4"?>' | wc -c => 38 + * 45 chars should be sufficient to reach the end of the encoding + * declaration without going too far inside the document content. + * on UTF-16 this means 90bytes, on UCS4 this means 180 + * The actual value depending on guessed encoding is passed as @len + * if provided + */ + if (len >= 0) { + if (toconv > (unsigned int) len) + toconv = len; + } else { + if (toconv > 180) + toconv = 180; + } + if (toconv * 2 >= written) { + xmlBufGrow(out, toconv * 2); + written = xmlBufAvail(out) - 1; + } + if (written > 360) + written = 360; + + c_in = toconv; + c_out = written; + if (input->encoder->input != NULL) { + ret = input->encoder->input(xmlBufEnd(out), &c_out, + xmlBufContent(in), &c_in); + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + } +#ifdef LIBXML_ICONV_ENABLED + else if (input->encoder->iconv_in != NULL) { + ret = xmlIconvWrapper(input->encoder->iconv_in, xmlBufEnd(out), + &c_out, xmlBufContent(in), &c_in); + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + if (ret == -1) + ret = -3; + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + else if (input->encoder->uconv_in != NULL) { + ret = xmlUconvWrapper(input->encoder->uconv_in, 1, xmlBufEnd(out), + &c_out, xmlBufContent(in), &c_in); + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + if (ret == -1) + ret = -3; + } +#endif /* LIBXML_ICU_ENABLED */ + switch (ret) { + case 0: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input\n", + c_in, c_out); +#endif + break; + case -1: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input, %d left\n", + c_in, c_out, (int)xmlBufUse(in)); +#endif + break; + case -3: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input, %d left\n", + c_in, c_out, (int)xmlBufUse(in)); +#endif + break; + case -2: { + char buf[50]; + const xmlChar *content = xmlBufContent(in); + + snprintf(&buf[0], 49, "0x%02X 0x%02X 0x%02X 0x%02X", + content[0], content[1], + content[2], content[3]); + buf[49] = 0; + xmlEncodingErr(XML_I18N_CONV_FAILED, + "input conversion failed due to input error, bytes %s\n", + buf); + } + } + /* + * Ignore when input buffer is not on a boundary + */ + if (ret == -3) ret = 0; + if (ret == -1) ret = 0; + return(ret); +} + +/** + * xmlCharEncInput: + * @input: a parser input buffer + * @flush: try to flush all the raw buffer + * + * Generic front-end for the encoding handler on parser input + * + * Returns the number of byte written if success, or + * -1 general error + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + */ +int +xmlCharEncInput(xmlParserInputBufferPtr input, int flush) +{ + int ret = -2; + size_t written; + size_t toconv; + int c_in; + int c_out; + xmlBufPtr in; + xmlBufPtr out; + + if ((input == NULL) || (input->encoder == NULL) || + (input->buffer == NULL) || (input->raw == NULL)) + return (-1); + out = input->buffer; + in = input->raw; + + toconv = xmlBufUse(in); + if (toconv == 0) + return (0); + if ((toconv > 64 * 1024) && (flush == 0)) + toconv = 64 * 1024; + written = xmlBufAvail(out); + if (written > 0) + written--; /* count '\0' */ + if (toconv * 2 >= written) { + xmlBufGrow(out, toconv * 2); + written = xmlBufAvail(out); + if (written > 0) + written--; /* count '\0' */ + } + if ((written > 128 * 1024) && (flush == 0)) + written = 128 * 1024; + + c_in = toconv; + c_out = written; + if (input->encoder->input != NULL) { + ret = input->encoder->input(xmlBufEnd(out), &c_out, + xmlBufContent(in), &c_in); + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + } +#ifdef LIBXML_ICONV_ENABLED + else if (input->encoder->iconv_in != NULL) { + ret = xmlIconvWrapper(input->encoder->iconv_in, xmlBufEnd(out), + &c_out, xmlBufContent(in), &c_in); + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + if (ret == -1) + ret = -3; + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + else if (input->encoder->uconv_in != NULL) { + ret = xmlUconvWrapper(input->encoder->uconv_in, 1, xmlBufEnd(out), + &c_out, xmlBufContent(in), &c_in); + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + if (ret == -1) + ret = -3; + } +#endif /* LIBXML_ICU_ENABLED */ + switch (ret) { + case 0: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input\n", + c_in, c_out); +#endif + break; + case -1: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input, %d left\n", + c_in, c_out, (int)xmlBufUse(in)); +#endif + break; + case -3: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input, %d left\n", + c_in, c_out, (int)xmlBufUse(in)); +#endif + break; + case -2: { + char buf[50]; + const xmlChar *content = xmlBufContent(in); + + snprintf(&buf[0], 49, "0x%02X 0x%02X 0x%02X 0x%02X", + content[0], content[1], + content[2], content[3]); + buf[49] = 0; + xmlEncodingErr(XML_I18N_CONV_FAILED, + "input conversion failed due to input error, bytes %s\n", + buf); + } + } + /* + * Ignore when input buffer is not on a boundary + */ + if (ret == -3) + ret = 0; + return (c_out? c_out : ret); +} + +/** + * xmlCharEncInFunc: + * @handler: char encoding transformation data structure + * @out: an xmlBuffer for the output. + * @in: an xmlBuffer for the input + * + * Generic front-end for the encoding handler input function + * + * Returns the number of byte written if success, or + * -1 general error + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + */ +int +xmlCharEncInFunc(xmlCharEncodingHandler * handler, xmlBufferPtr out, + xmlBufferPtr in) +{ + int ret = -2; + int written; + int toconv; + + if (handler == NULL) + return (-1); + if (out == NULL) + return (-1); + if (in == NULL) + return (-1); + + toconv = in->use; + if (toconv == 0) + return (0); + written = out->size - out->use -1; /* count '\0' */ + if (toconv * 2 >= written) { + xmlBufferGrow(out, out->size + toconv * 2); + written = out->size - out->use - 1; + } + if (handler->input != NULL) { + ret = handler->input(&out->content[out->use], &written, + in->content, &toconv); + xmlBufferShrink(in, toconv); + out->use += written; + out->content[out->use] = 0; + } +#ifdef LIBXML_ICONV_ENABLED + else if (handler->iconv_in != NULL) { + ret = xmlIconvWrapper(handler->iconv_in, &out->content[out->use], + &written, in->content, &toconv); + xmlBufferShrink(in, toconv); + out->use += written; + out->content[out->use] = 0; + if (ret == -1) + ret = -3; + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + else if (handler->uconv_in != NULL) { + ret = xmlUconvWrapper(handler->uconv_in, 1, &out->content[out->use], + &written, in->content, &toconv); + xmlBufferShrink(in, toconv); + out->use += written; + out->content[out->use] = 0; + if (ret == -1) + ret = -3; + } +#endif /* LIBXML_ICU_ENABLED */ + switch (ret) { + case 0: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input\n", + toconv, written); +#endif + break; + case -1: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input, %d left\n", + toconv, written, in->use); +#endif + break; + case -3: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of input, %d left\n", + toconv, written, in->use); +#endif + break; + case -2: { + char buf[50]; + + snprintf(&buf[0], 49, "0x%02X 0x%02X 0x%02X 0x%02X", + in->content[0], in->content[1], + in->content[2], in->content[3]); + buf[49] = 0; + xmlEncodingErr(XML_I18N_CONV_FAILED, + "input conversion failed due to input error, bytes %s\n", + buf); + } + } + /* + * Ignore when input buffer is not on a boundary + */ + if (ret == -3) + ret = 0; + return (written? written : ret); +} + +#ifdef LIBXML_OUTPUT_ENABLED +/** + * xmlCharEncOutput: + * @output: a parser output buffer + * @init: is this an initialization call without data + * + * Generic front-end for the encoding handler on parser output + * a first call with @init == 1 has to be made first to initiate the + * output in case of non-stateless encoding needing to initiate their + * state or the output (like the BOM in UTF16). + * In case of UTF8 sequence conversion errors for the given encoder, + * the content will be automatically remapped to a CharRef sequence. + * + * Returns the number of byte written if success, or + * -1 general error + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + */ +int +xmlCharEncOutput(xmlOutputBufferPtr output, int init) +{ + int ret = -2; + size_t written; + size_t writtentot = 0; + size_t toconv; + int c_in; + int c_out; + xmlBufPtr in; + xmlBufPtr out; + int charref_len = 0; + + if ((output == NULL) || (output->encoder == NULL) || + (output->buffer == NULL) || (output->conv == NULL)) + return (-1); + out = output->conv; + in = output->buffer; + +retry: + + written = xmlBufAvail(out); + if (written > 0) + written--; /* count '\0' */ + + /* + * First specific handling of the initialization call + */ + if (init) { + c_in = 0; + c_out = written; + if (output->encoder->output != NULL) { + ret = output->encoder->output(xmlBufEnd(out), &c_out, + NULL, &c_in); + if (ret > 0) /* Gennady: check return value */ + xmlBufAddLen(out, c_out); + } +#ifdef LIBXML_ICONV_ENABLED + else if (output->encoder->iconv_out != NULL) { + ret = xmlIconvWrapper(output->encoder->iconv_out, xmlBufEnd(out), + &c_out, NULL, &c_in); + xmlBufAddLen(out, c_out); + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + else if (output->encoder->uconv_out != NULL) { + ret = xmlUconvWrapper(output->encoder->uconv_out, 0, xmlBufEnd(out), + &c_out, NULL, &c_in); + xmlBufAddLen(out, c_out); + } +#endif /* LIBXML_ICU_ENABLED */ +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "initialized encoder\n"); +#endif + return(0); + } + + /* + * Conversion itself. + */ + toconv = xmlBufUse(in); + if (toconv == 0) + return (0); + if (toconv > 64 * 1024) + toconv = 64 * 1024; + if (toconv * 4 >= written) { + xmlBufGrow(out, toconv * 4); + written = xmlBufAvail(out) - 1; + } + if (written > 256 * 1024) + written = 256 * 1024; + + c_in = toconv; + c_out = written; + if (output->encoder->output != NULL) { + ret = output->encoder->output(xmlBufEnd(out), &c_out, + xmlBufContent(in), &c_in); + if (c_out > 0) { + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + writtentot += c_out; + } + } +#ifdef LIBXML_ICONV_ENABLED + else if (output->encoder->iconv_out != NULL) { + ret = xmlIconvWrapper(output->encoder->iconv_out, xmlBufEnd(out), + &c_out, xmlBufContent(in), &c_in); + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + writtentot += c_out; + if (ret == -1) { + if (c_out > 0) { + /* + * Can be a limitation of iconv + */ + charref_len = 0; + goto retry; + } + ret = -3; + } + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + else if (output->encoder->uconv_out != NULL) { + ret = xmlUconvWrapper(output->encoder->uconv_out, 0, xmlBufEnd(out), + &c_out, xmlBufContent(in), &c_in); + xmlBufShrink(in, c_in); + xmlBufAddLen(out, c_out); + writtentot += c_out; + if (ret == -1) { + if (c_out > 0) { + /* + * Can be a limitation of uconv + */ + charref_len = 0; + goto retry; + } + ret = -3; + } + } +#endif /* LIBXML_ICU_ENABLED */ + else { + xmlEncodingErr(XML_I18N_NO_OUTPUT, + "xmlCharEncOutFunc: no output function !\n", NULL); + return(-1); + } + + if (ret >= 0) output += ret; + + /* + * Attempt to handle error cases + */ + switch (ret) { + case 0: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of output\n", + c_in, c_out); +#endif + break; + case -1: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "output conversion failed by lack of space\n"); +#endif + break; + case -3: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext,"converted %d bytes to %d bytes of output %d left\n", + c_in, c_out, (int) xmlBufUse(in)); +#endif + break; + case -2: { + int len = (int) xmlBufUse(in); + xmlChar *content = xmlBufContent(in); + int cur; + + cur = xmlGetUTF8Char(content, &len); + if ((charref_len != 0) && (c_out < charref_len)) { + /* + * We attempted to insert a character reference and failed. + * Undo what was written and skip the remaining charref. + */ + xmlBufErase(out, c_out); + writtentot -= c_out; + xmlBufShrink(in, charref_len - c_out); + charref_len = 0; + + ret = -1; + break; + } else if (cur > 0) { + xmlChar charref[20]; + +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "handling output conversion error\n"); + xmlGenericError(xmlGenericErrorContext, + "Bytes: 0x%02X 0x%02X 0x%02X 0x%02X\n", + content[0], content[1], + content[2], content[3]); +#endif + /* + * Removes the UTF8 sequence, and replace it by a charref + * and continue the transcoding phase, hoping the error + * did not mangle the encoder state. + */ + charref_len = snprintf((char *) &charref[0], sizeof(charref), + "&#%d;", cur); + xmlBufShrink(in, len); + xmlBufAddHead(in, charref, -1); + + goto retry; + } else { + char buf[50]; + + snprintf(&buf[0], 49, "0x%02X 0x%02X 0x%02X 0x%02X", + content[0], content[1], + content[2], content[3]); + buf[49] = 0; + xmlEncodingErr(XML_I18N_CONV_FAILED, + "output conversion failed due to conv error, bytes %s\n", + buf); + if (xmlBufGetAllocationScheme(in) != XML_BUFFER_ALLOC_IMMUTABLE) + content[0] = ' '; + } + break; + } + } + return(ret); +} +#endif + +/** + * xmlCharEncOutFunc: + * @handler: char enconding transformation data structure + * @out: an xmlBuffer for the output. + * @in: an xmlBuffer for the input + * + * Generic front-end for the encoding handler output function + * a first call with @in == NULL has to be made firs to initiate the + * output in case of non-stateless encoding needing to initiate their + * state or the output (like the BOM in UTF16). + * In case of UTF8 sequence conversion errors for the given encoder, + * the content will be automatically remapped to a CharRef sequence. + * + * Returns the number of byte written if success, or + * -1 general error + * -2 if the transcoding fails (for *in is not valid utf8 string or + * the result of transformation can't fit into the encoding we want), or + */ +int +xmlCharEncOutFunc(xmlCharEncodingHandler *handler, xmlBufferPtr out, + xmlBufferPtr in) { + int ret = -2; + int written; + int writtentot = 0; + int toconv; + int output = 0; + int charref_len = 0; + + if (handler == NULL) return(-1); + if (out == NULL) return(-1); + +retry: + + written = out->size - out->use; + + if (written > 0) + written--; /* Gennady: count '/0' */ + + /* + * First specific handling of in = NULL, i.e. the initialization call + */ + if (in == NULL) { + toconv = 0; + if (handler->output != NULL) { + ret = handler->output(&out->content[out->use], &written, + NULL, &toconv); + if (ret >= 0) { /* Gennady: check return value */ + out->use += written; + out->content[out->use] = 0; + } + } +#ifdef LIBXML_ICONV_ENABLED + else if (handler->iconv_out != NULL) { + ret = xmlIconvWrapper(handler->iconv_out, &out->content[out->use], + &written, NULL, &toconv); + out->use += written; + out->content[out->use] = 0; + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + else if (handler->uconv_out != NULL) { + ret = xmlUconvWrapper(handler->uconv_out, 0, + &out->content[out->use], + &written, NULL, &toconv); + out->use += written; + out->content[out->use] = 0; + } +#endif /* LIBXML_ICU_ENABLED */ +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "initialized encoder\n"); +#endif + return(0); + } + + /* + * Conversion itself. + */ + toconv = in->use; + if (toconv == 0) + return(0); + if (toconv * 4 >= written) { + xmlBufferGrow(out, toconv * 4); + written = out->size - out->use - 1; + } + if (handler->output != NULL) { + ret = handler->output(&out->content[out->use], &written, + in->content, &toconv); + if (written > 0) { + xmlBufferShrink(in, toconv); + out->use += written; + writtentot += written; + } + out->content[out->use] = 0; + } +#ifdef LIBXML_ICONV_ENABLED + else if (handler->iconv_out != NULL) { + ret = xmlIconvWrapper(handler->iconv_out, &out->content[out->use], + &written, in->content, &toconv); + xmlBufferShrink(in, toconv); + out->use += written; + writtentot += written; + out->content[out->use] = 0; + if (ret == -1) { + if (written > 0) { + /* + * Can be a limitation of iconv + */ + charref_len = 0; + goto retry; + } + ret = -3; + } + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + else if (handler->uconv_out != NULL) { + ret = xmlUconvWrapper(handler->uconv_out, 0, + &out->content[out->use], + &written, in->content, &toconv); + xmlBufferShrink(in, toconv); + out->use += written; + writtentot += written; + out->content[out->use] = 0; + if (ret == -1) { + if (written > 0) { + /* + * Can be a limitation of iconv + */ + charref_len = 0; + goto retry; + } + ret = -3; + } + } +#endif /* LIBXML_ICU_ENABLED */ + else { + xmlEncodingErr(XML_I18N_NO_OUTPUT, + "xmlCharEncOutFunc: no output function !\n", NULL); + return(-1); + } + + if (ret >= 0) output += ret; + + /* + * Attempt to handle error cases + */ + switch (ret) { + case 0: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "converted %d bytes to %d bytes of output\n", + toconv, written); +#endif + break; + case -1: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "output conversion failed by lack of space\n"); +#endif + break; + case -3: +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext,"converted %d bytes to %d bytes of output %d left\n", + toconv, written, in->use); +#endif + break; + case -2: { + int len = in->use; + const xmlChar *utf = (const xmlChar *) in->content; + int cur; + + cur = xmlGetUTF8Char(utf, &len); + if ((charref_len != 0) && (written < charref_len)) { + /* + * We attempted to insert a character reference and failed. + * Undo what was written and skip the remaining charref. + */ + out->use -= written; + writtentot -= written; + xmlBufferShrink(in, charref_len - written); + charref_len = 0; + + ret = -1; + break; + } else if (cur > 0) { + xmlChar charref[20]; + +#ifdef DEBUG_ENCODING + xmlGenericError(xmlGenericErrorContext, + "handling output conversion error\n"); + xmlGenericError(xmlGenericErrorContext, + "Bytes: 0x%02X 0x%02X 0x%02X 0x%02X\n", + in->content[0], in->content[1], + in->content[2], in->content[3]); +#endif + /* + * Removes the UTF8 sequence, and replace it by a charref + * and continue the transcoding phase, hoping the error + * did not mangle the encoder state. + */ + charref_len = snprintf((char *) &charref[0], sizeof(charref), + "&#%d;", cur); + xmlBufferShrink(in, len); + xmlBufferAddHead(in, charref, -1); + + goto retry; + } else { + char buf[50]; + + snprintf(&buf[0], 49, "0x%02X 0x%02X 0x%02X 0x%02X", + in->content[0], in->content[1], + in->content[2], in->content[3]); + buf[49] = 0; + xmlEncodingErr(XML_I18N_CONV_FAILED, + "output conversion failed due to conv error, bytes %s\n", + buf); + if (in->alloc != XML_BUFFER_ALLOC_IMMUTABLE) + in->content[0] = ' '; + } + break; + } + } + return(ret); +} + +/** + * xmlCharEncCloseFunc: + * @handler: char enconding transformation data structure + * + * Generic front-end for encoding handler close function + * + * Returns 0 if success, or -1 in case of error + */ +int +xmlCharEncCloseFunc(xmlCharEncodingHandler *handler) { + int ret = 0; + int tofree = 0; + int i, handler_in_list = 0; + + if (handler == NULL) return(-1); + if (handler->name == NULL) return(-1); + if (handlers != NULL) { + for (i = 0;i < nbCharEncodingHandler; i++) { + if (handler == handlers[i]) { + handler_in_list = 1; + break; + } + } + } +#ifdef LIBXML_ICONV_ENABLED + /* + * Iconv handlers can be used only once, free the whole block. + * and the associated icon resources. + */ + if ((handler_in_list == 0) && + ((handler->iconv_out != NULL) || (handler->iconv_in != NULL))) { + tofree = 1; + if (handler->iconv_out != NULL) { + if (iconv_close(handler->iconv_out)) + ret = -1; + handler->iconv_out = NULL; + } + if (handler->iconv_in != NULL) { + if (iconv_close(handler->iconv_in)) + ret = -1; + handler->iconv_in = NULL; + } + } +#endif /* LIBXML_ICONV_ENABLED */ +#ifdef LIBXML_ICU_ENABLED + if ((handler_in_list == 0) && + ((handler->uconv_out != NULL) || (handler->uconv_in != NULL))) { + tofree = 1; + if (handler->uconv_out != NULL) { + closeIcuConverter(handler->uconv_out); + handler->uconv_out = NULL; + } + if (handler->uconv_in != NULL) { + closeIcuConverter(handler->uconv_in); + handler->uconv_in = NULL; + } + } +#endif + if (tofree) { + /* free up only dynamic handlers iconv/uconv */ + if (handler->name != NULL) + xmlFree(handler->name); + handler->name = NULL; + xmlFree(handler); + } +#ifdef DEBUG_ENCODING + if (ret) + xmlGenericError(xmlGenericErrorContext, + "failed to close the encoding handler\n"); + else + xmlGenericError(xmlGenericErrorContext, + "closed the encoding handler\n"); +#endif + + return(ret); +} + +/** + * xmlByteConsumed: + * @ctxt: an XML parser context + * + * This function provides the current index of the parser relative + * to the start of the current entity. This function is computed in + * bytes from the beginning starting at zero and finishing at the + * size in byte of the file if parsing a file. The function is + * of constant cost if the input is UTF-8 but can be costly if run + * on non-UTF-8 input. + * + * Returns the index in bytes from the beginning of the entity or -1 + * in case the index could not be computed. + */ +long +xmlByteConsumed(xmlParserCtxtPtr ctxt) { + xmlParserInputPtr in; + + if (ctxt == NULL) return(-1); + in = ctxt->input; + if (in == NULL) return(-1); + if ((in->buf != NULL) && (in->buf->encoder != NULL)) { + unsigned int unused = 0; + xmlCharEncodingHandler * handler = in->buf->encoder; + /* + * Encoding conversion, compute the number of unused original + * bytes from the input not consumed and substract that from + * the raw consumed value, this is not a cheap operation + */ + if (in->end - in->cur > 0) { + unsigned char convbuf[32000]; + const unsigned char *cur = (const unsigned char *)in->cur; + int toconv = in->end - in->cur, written = 32000; + + int ret; + + if (handler->output != NULL) { + do { + toconv = in->end - cur; + written = 32000; + ret = handler->output(&convbuf[0], &written, + cur, &toconv); + if (ret == -1) return(-1); + unused += written; + cur += toconv; + } while (ret == -2); +#ifdef LIBXML_ICONV_ENABLED + } else if (handler->iconv_out != NULL) { + do { + toconv = in->end - cur; + written = 32000; + ret = xmlIconvWrapper(handler->iconv_out, &convbuf[0], + &written, cur, &toconv); + if (ret < 0) { + if (written > 0) + ret = -2; + else + return(-1); + } + unused += written; + cur += toconv; + } while (ret == -2); +#endif +#ifdef LIBXML_ICU_ENABLED + } else if (handler->uconv_out != NULL) { + do { + toconv = in->end - cur; + written = 32000; + ret = xmlUconvWrapper(handler->uconv_out, 0, &convbuf[0], + &written, cur, &toconv); + if (ret < 0) { + if (written > 0) + ret = -2; + else + return(-1); + } + unused += written; + cur += toconv; + } while (ret == -2); +#endif + } else { + /* could not find a converter */ + return(-1); + } + } + if (in->buf->rawconsumed < unused) + return(-1); + return(in->buf->rawconsumed - unused); + } + return(in->consumed + (in->cur - in->base)); +} + +#if !defined(LIBXML_ICONV_ENABLED) && !defined(LIBXML_ICU_ENABLED) +#ifdef LIBXML_ISO8859X_ENABLED + +/** + * UTF8ToISO8859x: + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @in: a pointer to an array of UTF-8 chars + * @inlen: the length of @in + * @xlattable: the 2-level transcoding table + * + * Take a block of UTF-8 chars in and try to convert it to an ISO 8859-* + * block of chars out. + * + * Returns 0 if success, -2 if the transcoding fails, or -1 otherwise + * The value of @inlen after return is the number of octets consumed + * as the return value is positive, else unpredictable. + * The value of @outlen after return is the number of ocetes consumed. + */ +static int +UTF8ToISO8859x(unsigned char* out, int *outlen, + const unsigned char* in, int *inlen, + unsigned char const *xlattable) { + const unsigned char* outstart = out; + const unsigned char* inend; + const unsigned char* instart = in; + const unsigned char* processed = in; + + if ((out == NULL) || (outlen == NULL) || (inlen == NULL) || + (xlattable == NULL)) + return(-1); + if (in == NULL) { + /* + * initialization nothing to do + */ + *outlen = 0; + *inlen = 0; + return(0); + } + inend = in + (*inlen); + while (in < inend) { + unsigned char d = *in++; + if (d < 0x80) { + *out++ = d; + } else if (d < 0xC0) { + /* trailing byte in leading position */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } else if (d < 0xE0) { + unsigned char c; + if (!(in < inend)) { + /* trailing byte not in input buffer */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-3); + } + c = *in++; + if ((c & 0xC0) != 0x80) { + /* not a trailing byte */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + c = c & 0x3F; + d = d & 0x1F; + d = xlattable [48 + c + xlattable [d] * 64]; + if (d == 0) { + /* not in character set */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + *out++ = d; + } else if (d < 0xF0) { + unsigned char c1; + unsigned char c2; + if (!(in < inend - 1)) { + /* trailing bytes not in input buffer */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-3); + } + c1 = *in++; + if ((c1 & 0xC0) != 0x80) { + /* not a trailing byte (c1) */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + c2 = *in++; + if ((c2 & 0xC0) != 0x80) { + /* not a trailing byte (c2) */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + c1 = c1 & 0x3F; + c2 = c2 & 0x3F; + d = d & 0x0F; + d = xlattable [48 + c2 + xlattable [48 + c1 + + xlattable [32 + d] * 64] * 64]; + if (d == 0) { + /* not in character set */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + *out++ = d; + } else { + /* cannot transcode >= U+010000 */ + *outlen = out - outstart; + *inlen = processed - instart; + return(-2); + } + processed = in; + } + *outlen = out - outstart; + *inlen = processed - instart; + return(*outlen); +} + +/** + * ISO8859xToUTF8 + * @out: a pointer to an array of bytes to store the result + * @outlen: the length of @out + * @in: a pointer to an array of ISO Latin 1 chars + * @inlen: the length of @in + * + * Take a block of ISO 8859-* chars in and try to convert it to an UTF-8 + * block of chars out. + * Returns 0 if success, or -1 otherwise + * The value of @inlen after return is the number of octets consumed + * The value of @outlen after return is the number of ocetes produced. + */ +static int +ISO8859xToUTF8(unsigned char* out, int *outlen, + const unsigned char* in, int *inlen, + unsigned short const *unicodetable) { + unsigned char* outstart = out; + unsigned char* outend; + const unsigned char* instart = in; + const unsigned char* inend; + const unsigned char* instop; + unsigned int c; + + if ((out == NULL) || (outlen == NULL) || (inlen == NULL) || + (in == NULL) || (unicodetable == NULL)) + return(-1); + outend = out + *outlen; + inend = in + *inlen; + instop = inend; + + while ((in < inend) && (out < outend - 2)) { + if (*in >= 0x80) { + c = unicodetable [*in - 0x80]; + if (c == 0) { + /* undefined code point */ + *outlen = out - outstart; + *inlen = in - instart; + return (-1); + } + if (c < 0x800) { + *out++ = ((c >> 6) & 0x1F) | 0xC0; + *out++ = (c & 0x3F) | 0x80; + } else { + *out++ = ((c >> 12) & 0x0F) | 0xE0; + *out++ = ((c >> 6) & 0x3F) | 0x80; + *out++ = (c & 0x3F) | 0x80; + } + ++in; + } + if (instop - in > outend - out) instop = in + (outend - out); + while ((*in < 0x80) && (in < instop)) { + *out++ = *in++; + } + } + if ((in < inend) && (out < outend) && (*in < 0x80)) { + *out++ = *in++; + } + if ((in < inend) && (out < outend) && (*in < 0x80)) { + *out++ = *in++; + } + *outlen = out - outstart; + *inlen = in - instart; + return (*outlen); +} + + +/************************************************************************ + * Lookup tables for ISO-8859-2..ISO-8859-16 transcoding * + ************************************************************************/ + +static unsigned short const xmlunicodetable_ISO8859_2 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0104, 0x02d8, 0x0141, 0x00a4, 0x013d, 0x015a, 0x00a7, + 0x00a8, 0x0160, 0x015e, 0x0164, 0x0179, 0x00ad, 0x017d, 0x017b, + 0x00b0, 0x0105, 0x02db, 0x0142, 0x00b4, 0x013e, 0x015b, 0x02c7, + 0x00b8, 0x0161, 0x015f, 0x0165, 0x017a, 0x02dd, 0x017e, 0x017c, + 0x0154, 0x00c1, 0x00c2, 0x0102, 0x00c4, 0x0139, 0x0106, 0x00c7, + 0x010c, 0x00c9, 0x0118, 0x00cb, 0x011a, 0x00cd, 0x00ce, 0x010e, + 0x0110, 0x0143, 0x0147, 0x00d3, 0x00d4, 0x0150, 0x00d6, 0x00d7, + 0x0158, 0x016e, 0x00da, 0x0170, 0x00dc, 0x00dd, 0x0162, 0x00df, + 0x0155, 0x00e1, 0x00e2, 0x0103, 0x00e4, 0x013a, 0x0107, 0x00e7, + 0x010d, 0x00e9, 0x0119, 0x00eb, 0x011b, 0x00ed, 0x00ee, 0x010f, + 0x0111, 0x0144, 0x0148, 0x00f3, 0x00f4, 0x0151, 0x00f6, 0x00f7, + 0x0159, 0x016f, 0x00fa, 0x0171, 0x00fc, 0x00fd, 0x0163, 0x02d9, +}; + +static unsigned char const xmltranscodetable_ISO8859_2 [48 + 6 * 64] = { + "\x00\x00\x01\x05\x02\x04\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\x00\xa4\x00\x00\xa7\xa8\x00\x00\x00\x00\xad\x00\x00" + "\xb0\x00\x00\x00\xb4\x00\x00\x00\xb8\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\xc3\xe3\xa1\xb1\xc6\xe6\x00\x00\x00\x00\xc8\xe8\xcf\xef" + "\xd0\xf0\x00\x00\x00\x00\x00\x00\xca\xea\xcc\xec\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\xc5\xe5\x00\x00\xa5\xb5\x00" + "\x00\x00\x00\x00\x00\x00\x00\xb7\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\xa2\xff\x00\xb2\x00\xbd\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\xa3\xb3\xd1\xf1\x00\x00\xd2\xf2\x00\x00\x00\x00\x00\x00\x00" + "\xd5\xf5\x00\x00\xc0\xe0\x00\x00\xd8\xf8\xa6\xb6\x00\x00\xaa\xba" + "\xa9\xb9\xde\xfe\xab\xbb\x00\x00\x00\x00\x00\x00\x00\x00\xd9\xf9" + "\xdb\xfb\x00\x00\x00\x00\x00\x00\x00\xac\xbc\xaf\xbf\xae\xbe\x00" + "\x00\xc1\xc2\x00\xc4\x00\x00\xc7\x00\xc9\x00\xcb\x00\xcd\xce\x00" + "\x00\x00\x00\xd3\xd4\x00\xd6\xd7\x00\x00\xda\x00\xdc\xdd\x00\xdf" + "\x00\xe1\xe2\x00\xe4\x00\x00\xe7\x00\xe9\x00\xeb\x00\xed\xee\x00" + "\x00\x00\x00\xf3\xf4\x00\xf6\xf7\x00\x00\xfa\x00\xfc\xfd\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_3 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0126, 0x02d8, 0x00a3, 0x00a4, 0x0000, 0x0124, 0x00a7, + 0x00a8, 0x0130, 0x015e, 0x011e, 0x0134, 0x00ad, 0x0000, 0x017b, + 0x00b0, 0x0127, 0x00b2, 0x00b3, 0x00b4, 0x00b5, 0x0125, 0x00b7, + 0x00b8, 0x0131, 0x015f, 0x011f, 0x0135, 0x00bd, 0x0000, 0x017c, + 0x00c0, 0x00c1, 0x00c2, 0x0000, 0x00c4, 0x010a, 0x0108, 0x00c7, + 0x00c8, 0x00c9, 0x00ca, 0x00cb, 0x00cc, 0x00cd, 0x00ce, 0x00cf, + 0x0000, 0x00d1, 0x00d2, 0x00d3, 0x00d4, 0x0120, 0x00d6, 0x00d7, + 0x011c, 0x00d9, 0x00da, 0x00db, 0x00dc, 0x016c, 0x015c, 0x00df, + 0x00e0, 0x00e1, 0x00e2, 0x0000, 0x00e4, 0x010b, 0x0109, 0x00e7, + 0x00e8, 0x00e9, 0x00ea, 0x00eb, 0x00ec, 0x00ed, 0x00ee, 0x00ef, + 0x0000, 0x00f1, 0x00f2, 0x00f3, 0x00f4, 0x0121, 0x00f6, 0x00f7, + 0x011d, 0x00f9, 0x00fa, 0x00fb, 0x00fc, 0x016d, 0x015d, 0x02d9, +}; + +static unsigned char const xmltranscodetable_ISO8859_3 [48 + 7 * 64] = { + "\x04\x00\x01\x06\x02\x05\x00\x00\x00\x00\x00\x03\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\xa3\xa4\x00\x00\xa7\xa8\x00\x00\x00\x00\xad\x00\x00" + "\xb0\x00\xb2\xb3\xb4\xb5\x00\xb7\xb8\x00\x00\x00\x00\xbd\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\xc6\xe6\xc5\xe5\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xd8\xf8\xab\xbb" + "\xd5\xf5\x00\x00\xa6\xb6\xa1\xb1\x00\x00\x00\x00\x00\x00\x00\x00" + "\xa9\xb9\x00\x00\xac\xbc\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\xa2\xff\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xf0\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xde\xfe\xaa\xba" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xdd\xfd\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xaf\xbf\x00\x00\x00" + "\xc0\xc1\xc2\x00\xc4\x00\x00\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\x00\xd1\xd2\xd3\xd4\x00\xd6\xd7\x00\xd9\xda\xdb\xdc\x00\x00\xdf" + "\xe0\xe1\xe2\x00\xe4\x00\x00\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\x00\xf1\xf2\xf3\xf4\x00\xf6\xf7\x00\xf9\xfa\xfb\xfc\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_4 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0104, 0x0138, 0x0156, 0x00a4, 0x0128, 0x013b, 0x00a7, + 0x00a8, 0x0160, 0x0112, 0x0122, 0x0166, 0x00ad, 0x017d, 0x00af, + 0x00b0, 0x0105, 0x02db, 0x0157, 0x00b4, 0x0129, 0x013c, 0x02c7, + 0x00b8, 0x0161, 0x0113, 0x0123, 0x0167, 0x014a, 0x017e, 0x014b, + 0x0100, 0x00c1, 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x00c6, 0x012e, + 0x010c, 0x00c9, 0x0118, 0x00cb, 0x0116, 0x00cd, 0x00ce, 0x012a, + 0x0110, 0x0145, 0x014c, 0x0136, 0x00d4, 0x00d5, 0x00d6, 0x00d7, + 0x00d8, 0x0172, 0x00da, 0x00db, 0x00dc, 0x0168, 0x016a, 0x00df, + 0x0101, 0x00e1, 0x00e2, 0x00e3, 0x00e4, 0x00e5, 0x00e6, 0x012f, + 0x010d, 0x00e9, 0x0119, 0x00eb, 0x0117, 0x00ed, 0x00ee, 0x012b, + 0x0111, 0x0146, 0x014d, 0x0137, 0x00f4, 0x00f5, 0x00f6, 0x00f7, + 0x00f8, 0x0173, 0x00fa, 0x00fb, 0x00fc, 0x0169, 0x016b, 0x02d9, +}; + +static unsigned char const xmltranscodetable_ISO8859_4 [48 + 6 * 64] = { + "\x00\x00\x01\x05\x02\x03\x00\x00\x00\x00\x00\x04\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\x00\xa4\x00\x00\xa7\xa8\x00\x00\x00\x00\xad\x00\xaf" + "\xb0\x00\x00\x00\xb4\x00\x00\x00\xb8\x00\x00\x00\x00\x00\x00\x00" + "\xc0\xe0\x00\x00\xa1\xb1\x00\x00\x00\x00\x00\x00\xc8\xe8\x00\x00" + "\xd0\xf0\xaa\xba\x00\x00\xcc\xec\xca\xea\x00\x00\x00\x00\x00\x00" + "\x00\x00\xab\xbb\x00\x00\x00\x00\xa5\xb5\xcf\xef\x00\x00\xc7\xe7" + "\x00\x00\x00\x00\x00\x00\xd3\xf3\xa2\x00\x00\xa6\xb6\x00\x00\x00" + "\x00\x00\x00\x00\x00\xd1\xf1\x00\x00\x00\xbd\xbf\xd2\xf2\x00\x00" + "\x00\x00\x00\x00\x00\x00\xa3\xb3\x00\x00\x00\x00\x00\x00\x00\x00" + "\xa9\xb9\x00\x00\x00\x00\xac\xbc\xdd\xfd\xde\xfe\x00\x00\x00\x00" + "\x00\x00\xd9\xf9\x00\x00\x00\x00\x00\x00\x00\x00\x00\xae\xbe\x00" + "\x00\x00\x00\x00\x00\x00\x00\xb7\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\xff\x00\xb2\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\xc1\xc2\xc3\xc4\xc5\xc6\x00\x00\xc9\x00\xcb\x00\xcd\xce\x00" + "\x00\x00\x00\x00\xd4\xd5\xd6\xd7\xd8\x00\xda\xdb\xdc\x00\x00\xdf" + "\x00\xe1\xe2\xe3\xe4\xe5\xe6\x00\x00\xe9\x00\xeb\x00\xed\xee\x00" + "\x00\x00\x00\x00\xf4\xf5\xf6\xf7\xf8\x00\xfa\xfb\xfc\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_5 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0401, 0x0402, 0x0403, 0x0404, 0x0405, 0x0406, 0x0407, + 0x0408, 0x0409, 0x040a, 0x040b, 0x040c, 0x00ad, 0x040e, 0x040f, + 0x0410, 0x0411, 0x0412, 0x0413, 0x0414, 0x0415, 0x0416, 0x0417, + 0x0418, 0x0419, 0x041a, 0x041b, 0x041c, 0x041d, 0x041e, 0x041f, + 0x0420, 0x0421, 0x0422, 0x0423, 0x0424, 0x0425, 0x0426, 0x0427, + 0x0428, 0x0429, 0x042a, 0x042b, 0x042c, 0x042d, 0x042e, 0x042f, + 0x0430, 0x0431, 0x0432, 0x0433, 0x0434, 0x0435, 0x0436, 0x0437, + 0x0438, 0x0439, 0x043a, 0x043b, 0x043c, 0x043d, 0x043e, 0x043f, + 0x0440, 0x0441, 0x0442, 0x0443, 0x0444, 0x0445, 0x0446, 0x0447, + 0x0448, 0x0449, 0x044a, 0x044b, 0x044c, 0x044d, 0x044e, 0x044f, + 0x2116, 0x0451, 0x0452, 0x0453, 0x0454, 0x0455, 0x0456, 0x0457, + 0x0458, 0x0459, 0x045a, 0x045b, 0x045c, 0x00a7, 0x045e, 0x045f, +}; + +static unsigned char const xmltranscodetable_ISO8859_5 [48 + 6 * 64] = { + "\x00\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x02\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\x00\x00\x00\x00\xfd\x00\x00\x00\x00\x00\xad\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\xa1\xa2\xa3\xa4\xa5\xa6\xa7\xa8\xa9\xaa\xab\xac\x00\xae\xaf" + "\xb0\xb1\xb2\xb3\xb4\xb5\xb6\xb7\xb8\xb9\xba\xbb\xbc\xbd\xbe\xbf" + "\xc0\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\xd0\xd1\xd2\xd3\xd4\xd5\xd6\xd7\xd8\xd9\xda\xdb\xdc\xdd\xde\xdf" + "\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\x00\xf1\xf2\xf3\xf4\xf5\xf6\xf7\xf8\xf9\xfa\xfb\xfc\x00\xfe\xff" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\xf0\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_6 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0000, 0x0000, 0x0000, 0x00a4, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x060c, 0x00ad, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x061b, 0x0000, 0x0000, 0x0000, 0x061f, + 0x0000, 0x0621, 0x0622, 0x0623, 0x0624, 0x0625, 0x0626, 0x0627, + 0x0628, 0x0629, 0x062a, 0x062b, 0x062c, 0x062d, 0x062e, 0x062f, + 0x0630, 0x0631, 0x0632, 0x0633, 0x0634, 0x0635, 0x0636, 0x0637, + 0x0638, 0x0639, 0x063a, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0640, 0x0641, 0x0642, 0x0643, 0x0644, 0x0645, 0x0646, 0x0647, + 0x0648, 0x0649, 0x064a, 0x064b, 0x064c, 0x064d, 0x064e, 0x064f, + 0x0650, 0x0651, 0x0652, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, +}; + +static unsigned char const xmltranscodetable_ISO8859_6 [48 + 5 * 64] = { + "\x02\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x03\x04\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\x00\xa4\x00\x00\x00\x00\x00\x00\x00\x00\xad\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xff\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xac\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xbb\x00\x00\x00\xbf" + "\x00\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\xd0\xd1\xd2\xd3\xd4\xd5\xd6\xd7\xd8\xd9\xda\x00\x00\x00\x00\x00" + "\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\xf0\xf1\xf2\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_7 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x2018, 0x2019, 0x00a3, 0x0000, 0x0000, 0x00a6, 0x00a7, + 0x00a8, 0x00a9, 0x0000, 0x00ab, 0x00ac, 0x00ad, 0x0000, 0x2015, + 0x00b0, 0x00b1, 0x00b2, 0x00b3, 0x0384, 0x0385, 0x0386, 0x00b7, + 0x0388, 0x0389, 0x038a, 0x00bb, 0x038c, 0x00bd, 0x038e, 0x038f, + 0x0390, 0x0391, 0x0392, 0x0393, 0x0394, 0x0395, 0x0396, 0x0397, + 0x0398, 0x0399, 0x039a, 0x039b, 0x039c, 0x039d, 0x039e, 0x039f, + 0x03a0, 0x03a1, 0x0000, 0x03a3, 0x03a4, 0x03a5, 0x03a6, 0x03a7, + 0x03a8, 0x03a9, 0x03aa, 0x03ab, 0x03ac, 0x03ad, 0x03ae, 0x03af, + 0x03b0, 0x03b1, 0x03b2, 0x03b3, 0x03b4, 0x03b5, 0x03b6, 0x03b7, + 0x03b8, 0x03b9, 0x03ba, 0x03bb, 0x03bc, 0x03bd, 0x03be, 0x03bf, + 0x03c0, 0x03c1, 0x03c2, 0x03c3, 0x03c4, 0x03c5, 0x03c6, 0x03c7, + 0x03c8, 0x03c9, 0x03ca, 0x03cb, 0x03cc, 0x03cd, 0x03ce, 0x0000, +}; + +static unsigned char const xmltranscodetable_ISO8859_7 [48 + 7 * 64] = { + "\x04\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x05\x06" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\xa3\x00\x00\xa6\xa7\xa8\xa9\x00\xab\xac\xad\x00\x00" + "\xb0\xb1\xb2\xb3\x00\x00\x00\xb7\x00\x00\x00\xbb\x00\xbd\x00\x00" + "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\xaf\x00\x00\xa1\xa2\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xff\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\xb4\xb5\xb6\x00\xb8\xb9\xba\x00\xbc\x00\xbe\xbf" + "\xc0\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\xd0\xd1\x00\xd3\xd4\xd5\xd6\xd7\xd8\xd9\xda\xdb\xdc\xdd\xde\xdf" + "\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\xf0\xf1\xf2\xf3\xf4\xf5\xf6\xf7\xf8\xf9\xfa\xfb\xfc\xfd\xfe\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_8 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0000, 0x00a2, 0x00a3, 0x00a4, 0x00a5, 0x00a6, 0x00a7, + 0x00a8, 0x00a9, 0x00d7, 0x00ab, 0x00ac, 0x00ad, 0x00ae, 0x00af, + 0x00b0, 0x00b1, 0x00b2, 0x00b3, 0x00b4, 0x00b5, 0x00b6, 0x00b7, + 0x00b8, 0x00b9, 0x00f7, 0x00bb, 0x00bc, 0x00bd, 0x00be, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, + 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x2017, + 0x05d0, 0x05d1, 0x05d2, 0x05d3, 0x05d4, 0x05d5, 0x05d6, 0x05d7, + 0x05d8, 0x05d9, 0x05da, 0x05db, 0x05dc, 0x05dd, 0x05de, 0x05df, + 0x05e0, 0x05e1, 0x05e2, 0x05e3, 0x05e4, 0x05e5, 0x05e6, 0x05e7, + 0x05e8, 0x05e9, 0x05ea, 0x0000, 0x0000, 0x200e, 0x200f, 0x0000, +}; + +static unsigned char const xmltranscodetable_ISO8859_8 [48 + 7 * 64] = { + "\x02\x00\x01\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x06\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\xa2\xa3\xa4\xa5\xa6\xa7\xa8\xa9\x00\xab\xac\xad\xae\xaf" + "\xb0\xb1\xb2\xb3\xb4\xb5\xb6\xb7\xb8\xb9\x00\xbb\xbc\xbd\xbe\x00" + "\xff\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\xaa\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\xba\x00\x00\x00\x00\x00\x00\x00\x00" + "\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xfd\xfe" + "\x00\x00\x00\x00\x00\x00\x00\xdf\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\xf0\xf1\xf2\xf3\xf4\xf5\xf6\xf7\xf8\xf9\xfa\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_9 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x00a1, 0x00a2, 0x00a3, 0x00a4, 0x00a5, 0x00a6, 0x00a7, + 0x00a8, 0x00a9, 0x00aa, 0x00ab, 0x00ac, 0x00ad, 0x00ae, 0x00af, + 0x00b0, 0x00b1, 0x00b2, 0x00b3, 0x00b4, 0x00b5, 0x00b6, 0x00b7, + 0x00b8, 0x00b9, 0x00ba, 0x00bb, 0x00bc, 0x00bd, 0x00be, 0x00bf, + 0x00c0, 0x00c1, 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x00c6, 0x00c7, + 0x00c8, 0x00c9, 0x00ca, 0x00cb, 0x00cc, 0x00cd, 0x00ce, 0x00cf, + 0x011e, 0x00d1, 0x00d2, 0x00d3, 0x00d4, 0x00d5, 0x00d6, 0x00d7, + 0x00d8, 0x00d9, 0x00da, 0x00db, 0x00dc, 0x0130, 0x015e, 0x00df, + 0x00e0, 0x00e1, 0x00e2, 0x00e3, 0x00e4, 0x00e5, 0x00e6, 0x00e7, + 0x00e8, 0x00e9, 0x00ea, 0x00eb, 0x00ec, 0x00ed, 0x00ee, 0x00ef, + 0x011f, 0x00f1, 0x00f2, 0x00f3, 0x00f4, 0x00f5, 0x00f6, 0x00f7, + 0x00f8, 0x00f9, 0x00fa, 0x00fb, 0x00fc, 0x0131, 0x015f, 0x00ff, +}; + +static unsigned char const xmltranscodetable_ISO8859_9 [48 + 5 * 64] = { + "\x00\x00\x01\x02\x03\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\xa1\xa2\xa3\xa4\xa5\xa6\xa7\xa8\xa9\xaa\xab\xac\xad\xae\xaf" + "\xb0\xb1\xb2\xb3\xb4\xb5\xb6\xb7\xb8\xb9\xba\xbb\xbc\xbd\xbe\xbf" + "\xc0\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\x00\xd1\xd2\xd3\xd4\xd5\xd6\xd7\xd8\xd9\xda\xdb\xdc\x00\x00\xdf" + "\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\x00\xf1\xf2\xf3\xf4\xf5\xf6\xf7\xf8\xf9\xfa\xfb\xfc\x00\x00\xff" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xd0\xf0" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xdd\xfd\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xde\xfe" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_10 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0104, 0x0112, 0x0122, 0x012a, 0x0128, 0x0136, 0x00a7, + 0x013b, 0x0110, 0x0160, 0x0166, 0x017d, 0x00ad, 0x016a, 0x014a, + 0x00b0, 0x0105, 0x0113, 0x0123, 0x012b, 0x0129, 0x0137, 0x00b7, + 0x013c, 0x0111, 0x0161, 0x0167, 0x017e, 0x2015, 0x016b, 0x014b, + 0x0100, 0x00c1, 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x00c6, 0x012e, + 0x010c, 0x00c9, 0x0118, 0x00cb, 0x0116, 0x00cd, 0x00ce, 0x00cf, + 0x00d0, 0x0145, 0x014c, 0x00d3, 0x00d4, 0x00d5, 0x00d6, 0x0168, + 0x00d8, 0x0172, 0x00da, 0x00db, 0x00dc, 0x00dd, 0x00de, 0x00df, + 0x0101, 0x00e1, 0x00e2, 0x00e3, 0x00e4, 0x00e5, 0x00e6, 0x012f, + 0x010d, 0x00e9, 0x0119, 0x00eb, 0x0117, 0x00ed, 0x00ee, 0x00ef, + 0x00f0, 0x0146, 0x014d, 0x00f3, 0x00f4, 0x00f5, 0x00f6, 0x0169, + 0x00f8, 0x0173, 0x00fa, 0x00fb, 0x00fc, 0x00fd, 0x00fe, 0x0138, +}; + +static unsigned char const xmltranscodetable_ISO8859_10 [48 + 7 * 64] = { + "\x00\x00\x01\x06\x02\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\x00\x00\x00\x00\xa7\x00\x00\x00\x00\x00\xad\x00\x00" + "\xb0\x00\x00\x00\x00\x00\x00\xb7\x00\x00\x00\x00\x00\x00\x00\x00" + "\xc0\xe0\x00\x00\xa1\xb1\x00\x00\x00\x00\x00\x00\xc8\xe8\x00\x00" + "\xa9\xb9\xa2\xb2\x00\x00\xcc\xec\xca\xea\x00\x00\x00\x00\x00\x00" + "\x00\x00\xa3\xb3\x00\x00\x00\x00\xa5\xb5\xa4\xb4\x00\x00\xc7\xe7" + "\x00\x00\x00\x00\x00\x00\xa6\xb6\xff\x00\x00\xa8\xb8\x00\x00\x00" + "\x00\x00\x00\x00\x00\xd1\xf1\x00\x00\x00\xaf\xbf\xd2\xf2\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xaa\xba\x00\x00\x00\x00\xab\xbb\xd7\xf7\xae\xbe\x00\x00\x00\x00" + "\x00\x00\xd9\xf9\x00\x00\x00\x00\x00\x00\x00\x00\x00\xac\xbc\x00" + "\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\xbd\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\xc1\xc2\xc3\xc4\xc5\xc6\x00\x00\xc9\x00\xcb\x00\xcd\xce\xcf" + "\xd0\x00\x00\xd3\xd4\xd5\xd6\x00\xd8\x00\xda\xdb\xdc\xdd\xde\xdf" + "\x00\xe1\xe2\xe3\xe4\xe5\xe6\x00\x00\xe9\x00\xeb\x00\xed\xee\xef" + "\xf0\x00\x00\xf3\xf4\xf5\xf6\x00\xf8\x00\xfa\xfb\xfc\xfd\xfe\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_11 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0e01, 0x0e02, 0x0e03, 0x0e04, 0x0e05, 0x0e06, 0x0e07, + 0x0e08, 0x0e09, 0x0e0a, 0x0e0b, 0x0e0c, 0x0e0d, 0x0e0e, 0x0e0f, + 0x0e10, 0x0e11, 0x0e12, 0x0e13, 0x0e14, 0x0e15, 0x0e16, 0x0e17, + 0x0e18, 0x0e19, 0x0e1a, 0x0e1b, 0x0e1c, 0x0e1d, 0x0e1e, 0x0e1f, + 0x0e20, 0x0e21, 0x0e22, 0x0e23, 0x0e24, 0x0e25, 0x0e26, 0x0e27, + 0x0e28, 0x0e29, 0x0e2a, 0x0e2b, 0x0e2c, 0x0e2d, 0x0e2e, 0x0e2f, + 0x0e30, 0x0e31, 0x0e32, 0x0e33, 0x0e34, 0x0e35, 0x0e36, 0x0e37, + 0x0e38, 0x0e39, 0x0e3a, 0x0000, 0x0000, 0x0000, 0x0000, 0x0e3f, + 0x0e40, 0x0e41, 0x0e42, 0x0e43, 0x0e44, 0x0e45, 0x0e46, 0x0e47, + 0x0e48, 0x0e49, 0x0e4a, 0x0e4b, 0x0e4c, 0x0e4d, 0x0e4e, 0x0e4f, + 0x0e50, 0x0e51, 0x0e52, 0x0e53, 0x0e54, 0x0e55, 0x0e56, 0x0e57, + 0x0e58, 0x0e59, 0x0e5a, 0x0e5b, 0x0000, 0x0000, 0x0000, 0x0000, +}; + +static unsigned char const xmltranscodetable_ISO8859_11 [48 + 6 * 64] = { + "\x04\x00\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x03\x05\x00\x00\x00\x00\x00\x00" + "\x00\xa1\xa2\xa3\xa4\xa5\xa6\xa7\xa8\xa9\xaa\xab\xac\xad\xae\xaf" + "\xb0\xb1\xb2\xb3\xb4\xb5\xb6\xb7\xb8\xb9\xba\xbb\xbc\xbd\xbe\xbf" + "\xc0\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\xd0\xd1\xd2\xd3\xd4\xd5\xd6\xd7\xd8\xd9\xda\x00\x00\x00\x00\xdf" + "\xff\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\xf0\xf1\xf2\xf3\xf4\xf5\xf6\xf7\xf8\xf9\xfa\xfb\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_13 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x201d, 0x00a2, 0x00a3, 0x00a4, 0x201e, 0x00a6, 0x00a7, + 0x00d8, 0x00a9, 0x0156, 0x00ab, 0x00ac, 0x00ad, 0x00ae, 0x00c6, + 0x00b0, 0x00b1, 0x00b2, 0x00b3, 0x201c, 0x00b5, 0x00b6, 0x00b7, + 0x00f8, 0x00b9, 0x0157, 0x00bb, 0x00bc, 0x00bd, 0x00be, 0x00e6, + 0x0104, 0x012e, 0x0100, 0x0106, 0x00c4, 0x00c5, 0x0118, 0x0112, + 0x010c, 0x00c9, 0x0179, 0x0116, 0x0122, 0x0136, 0x012a, 0x013b, + 0x0160, 0x0143, 0x0145, 0x00d3, 0x014c, 0x00d5, 0x00d6, 0x00d7, + 0x0172, 0x0141, 0x015a, 0x016a, 0x00dc, 0x017b, 0x017d, 0x00df, + 0x0105, 0x012f, 0x0101, 0x0107, 0x00e4, 0x00e5, 0x0119, 0x0113, + 0x010d, 0x00e9, 0x017a, 0x0117, 0x0123, 0x0137, 0x012b, 0x013c, + 0x0161, 0x0144, 0x0146, 0x00f3, 0x014d, 0x00f5, 0x00f6, 0x00f7, + 0x0173, 0x0142, 0x015b, 0x016b, 0x00fc, 0x017c, 0x017e, 0x2019, +}; + +static unsigned char const xmltranscodetable_ISO8859_13 [48 + 7 * 64] = { + "\x00\x00\x01\x04\x06\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\xa2\xa3\xa4\x00\xa6\xa7\x00\xa9\x00\xab\xac\xad\xae\x00" + "\xb0\xb1\xb2\xb3\x00\xb5\xb6\xb7\x00\xb9\x00\xbb\xbc\xbd\xbe\x00" + "\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\xff\x00\x00\xb4\xa1\xa5\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\xc4\xc5\xaf\x00\x00\xc9\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\xd3\x00\xd5\xd6\xd7\xa8\x00\x00\x00\xdc\x00\x00\xdf" + "\x00\x00\x00\x00\xe4\xe5\xbf\x00\x00\xe9\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\xf3\x00\xf5\xf6\xf7\xb8\x00\x00\x00\xfc\x00\x00\x00" + "\x00\xd9\xf9\xd1\xf1\xd2\xf2\x00\x00\x00\x00\x00\xd4\xf4\x00\x00" + "\x00\x00\x00\x00\x00\x00\xaa\xba\x00\x00\xda\xfa\x00\x00\x00\x00" + "\xd0\xf0\x00\x00\x00\x00\x00\x00\x00\x00\xdb\xfb\x00\x00\x00\x00" + "\x00\x00\xd8\xf8\x00\x00\x00\x00\x00\xca\xea\xdd\xfd\xde\xfe\x00" + "\xc2\xe2\x00\x00\xc0\xe0\xc3\xe3\x00\x00\x00\x00\xc8\xe8\x00\x00" + "\x00\x00\xc7\xe7\x00\x00\xcb\xeb\xc6\xe6\x00\x00\x00\x00\x00\x00" + "\x00\x00\xcc\xec\x00\x00\x00\x00\x00\x00\xce\xee\x00\x00\xc1\xe1" + "\x00\x00\x00\x00\x00\x00\xcd\xed\x00\x00\x00\xcf\xef\x00\x00\x00" +}; + +static unsigned short const xmlunicodetable_ISO8859_14 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x1e02, 0x1e03, 0x00a3, 0x010a, 0x010b, 0x1e0a, 0x00a7, + 0x1e80, 0x00a9, 0x1e82, 0x1e0b, 0x1ef2, 0x00ad, 0x00ae, 0x0178, + 0x1e1e, 0x1e1f, 0x0120, 0x0121, 0x1e40, 0x1e41, 0x00b6, 0x1e56, + 0x1e81, 0x1e57, 0x1e83, 0x1e60, 0x1ef3, 0x1e84, 0x1e85, 0x1e61, + 0x00c0, 0x00c1, 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x00c6, 0x00c7, + 0x00c8, 0x00c9, 0x00ca, 0x00cb, 0x00cc, 0x00cd, 0x00ce, 0x00cf, + 0x0174, 0x00d1, 0x00d2, 0x00d3, 0x00d4, 0x00d5, 0x00d6, 0x1e6a, + 0x00d8, 0x00d9, 0x00da, 0x00db, 0x00dc, 0x00dd, 0x0176, 0x00df, + 0x00e0, 0x00e1, 0x00e2, 0x00e3, 0x00e4, 0x00e5, 0x00e6, 0x00e7, + 0x00e8, 0x00e9, 0x00ea, 0x00eb, 0x00ec, 0x00ed, 0x00ee, 0x00ef, + 0x0175, 0x00f1, 0x00f2, 0x00f3, 0x00f4, 0x00f5, 0x00f6, 0x1e6b, + 0x00f8, 0x00f9, 0x00fa, 0x00fb, 0x00fc, 0x00fd, 0x0177, 0x00ff, +}; + +static unsigned char const xmltranscodetable_ISO8859_14 [48 + 10 * 64] = { + "\x00\x00\x01\x09\x04\x07\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\xa3\x00\x00\x00\xa7\x00\xa9\x00\x00\x00\xad\xae\x00" + "\x00\x00\x00\x00\x00\x00\xb6\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x03\x08\x05\x06\x00\x00\x00\x00" + "\x00\x00\xa1\xa2\x00\x00\x00\x00\x00\x00\xa6\xab\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xb0\xb1" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xa4\xa5\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xb2\xb3\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xa8\xb8\xaa\xba\xbd\xbe\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\xac\xbc\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\xd0\xf0\xde\xfe\xaf\x00\x00\x00\x00\x00\x00\x00" + "\xb4\xb5\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\xb7\xb9\x00\x00\x00\x00\x00\x00\x00\x00" + "\xbb\xbf\x00\x00\x00\x00\x00\x00\x00\x00\xd7\xf7\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xc0\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\x00\xd1\xd2\xd3\xd4\xd5\xd6\x00\xd8\xd9\xda\xdb\xdc\xdd\x00\xdf" + "\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\x00\xf1\xf2\xf3\xf4\xf5\xf6\x00\xf8\xf9\xfa\xfb\xfc\xfd\x00\xff" +}; + +static unsigned short const xmlunicodetable_ISO8859_15 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x00a1, 0x00a2, 0x00a3, 0x20ac, 0x00a5, 0x0160, 0x00a7, + 0x0161, 0x00a9, 0x00aa, 0x00ab, 0x00ac, 0x00ad, 0x00ae, 0x00af, + 0x00b0, 0x00b1, 0x00b2, 0x00b3, 0x017d, 0x00b5, 0x00b6, 0x00b7, + 0x017e, 0x00b9, 0x00ba, 0x00bb, 0x0152, 0x0153, 0x0178, 0x00bf, + 0x00c0, 0x00c1, 0x00c2, 0x00c3, 0x00c4, 0x00c5, 0x00c6, 0x00c7, + 0x00c8, 0x00c9, 0x00ca, 0x00cb, 0x00cc, 0x00cd, 0x00ce, 0x00cf, + 0x00d0, 0x00d1, 0x00d2, 0x00d3, 0x00d4, 0x00d5, 0x00d6, 0x00d7, + 0x00d8, 0x00d9, 0x00da, 0x00db, 0x00dc, 0x00dd, 0x00de, 0x00df, + 0x00e0, 0x00e1, 0x00e2, 0x00e3, 0x00e4, 0x00e5, 0x00e6, 0x00e7, + 0x00e8, 0x00e9, 0x00ea, 0x00eb, 0x00ec, 0x00ed, 0x00ee, 0x00ef, + 0x00f0, 0x00f1, 0x00f2, 0x00f3, 0x00f4, 0x00f5, 0x00f6, 0x00f7, + 0x00f8, 0x00f9, 0x00fa, 0x00fb, 0x00fc, 0x00fd, 0x00fe, 0x00ff, +}; + +static unsigned char const xmltranscodetable_ISO8859_15 [48 + 6 * 64] = { + "\x00\x00\x01\x05\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\xa1\xa2\xa3\x00\xa5\x00\xa7\x00\xa9\xaa\xab\xac\xad\xae\xaf" + "\xb0\xb1\xb2\xb3\x00\xb5\xb6\xb7\x00\xb9\xba\xbb\x00\x00\x00\xbf" + "\x00\x00\x03\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xa4\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\xbc\xbd\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xa6\xa8\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\xbe\x00\x00\x00\x00\xb4\xb8\x00" + "\xc0\xc1\xc2\xc3\xc4\xc5\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\xd0\xd1\xd2\xd3\xd4\xd5\xd6\xd7\xd8\xd9\xda\xdb\xdc\xdd\xde\xdf" + "\xe0\xe1\xe2\xe3\xe4\xe5\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\xf0\xf1\xf2\xf3\xf4\xf5\xf6\xf7\xf8\xf9\xfa\xfb\xfc\xfd\xfe\xff" +}; + +static unsigned short const xmlunicodetable_ISO8859_16 [128] = { + 0x0080, 0x0081, 0x0082, 0x0083, 0x0084, 0x0085, 0x0086, 0x0087, + 0x0088, 0x0089, 0x008a, 0x008b, 0x008c, 0x008d, 0x008e, 0x008f, + 0x0090, 0x0091, 0x0092, 0x0093, 0x0094, 0x0095, 0x0096, 0x0097, + 0x0098, 0x0099, 0x009a, 0x009b, 0x009c, 0x009d, 0x009e, 0x009f, + 0x00a0, 0x0104, 0x0105, 0x0141, 0x20ac, 0x201e, 0x0160, 0x00a7, + 0x0161, 0x00a9, 0x0218, 0x00ab, 0x0179, 0x00ad, 0x017a, 0x017b, + 0x00b0, 0x00b1, 0x010c, 0x0142, 0x017d, 0x201d, 0x00b6, 0x00b7, + 0x017e, 0x010d, 0x0219, 0x00bb, 0x0152, 0x0153, 0x0178, 0x017c, + 0x00c0, 0x00c1, 0x00c2, 0x0102, 0x00c4, 0x0106, 0x00c6, 0x00c7, + 0x00c8, 0x00c9, 0x00ca, 0x00cb, 0x00cc, 0x00cd, 0x00ce, 0x00cf, + 0x0110, 0x0143, 0x00d2, 0x00d3, 0x00d4, 0x0150, 0x00d6, 0x015a, + 0x0170, 0x00d9, 0x00da, 0x00db, 0x00dc, 0x0118, 0x021a, 0x00df, + 0x00e0, 0x00e1, 0x00e2, 0x0103, 0x00e4, 0x0107, 0x00e6, 0x00e7, + 0x00e8, 0x00e9, 0x00ea, 0x00eb, 0x00ec, 0x00ed, 0x00ee, 0x00ef, + 0x0111, 0x0144, 0x00f2, 0x00f3, 0x00f4, 0x0151, 0x00f6, 0x015b, + 0x0171, 0x00f9, 0x00fa, 0x00fb, 0x00fc, 0x0119, 0x021b, 0x00ff, +}; + +static unsigned char const xmltranscodetable_ISO8859_16 [48 + 9 * 64] = { + "\x00\x00\x01\x08\x02\x03\x00\x00\x07\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x04\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8a\x8b\x8c\x8d\x8e\x8f" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9a\x9b\x9c\x9d\x9e\x9f" + "\xa0\x00\x00\x00\x00\x00\x00\xa7\x00\xa9\x00\xab\x00\xad\x00\x00" + "\xb0\xb1\x00\x00\x00\x00\xb6\xb7\x00\x00\x00\xbb\x00\x00\x00\x00" + "\x00\x00\xc3\xe3\xa1\xa2\xc5\xe5\x00\x00\x00\x00\xb2\xb9\x00\x00" + "\xd0\xf0\x00\x00\x00\x00\x00\x00\xdd\xfd\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\xa3\xb3\xd1\xf1\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xd5\xf5\xbc\xbd\x00\x00\x00\x00\x00\x00\xd7\xf7\x00\x00\x00\x00" + "\xa6\xa8\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xd8\xf8\x00\x00\x00\x00\x00\x00\xbe\xac\xae\xaf\xbf\xb4\xb8\x00" + "\x06\x00\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xa4\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xb5\xa5\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\xaa\xba\xde\xfe\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\xc0\xc1\xc2\x00\xc4\x00\xc6\xc7\xc8\xc9\xca\xcb\xcc\xcd\xce\xcf" + "\x00\x00\xd2\xd3\xd4\x00\xd6\x00\x00\xd9\xda\xdb\xdc\x00\x00\xdf" + "\xe0\xe1\xe2\x00\xe4\x00\xe6\xe7\xe8\xe9\xea\xeb\xec\xed\xee\xef" + "\x00\x00\xf2\xf3\xf4\x00\xf6\x00\x00\xf9\xfa\xfb\xfc\x00\x00\xff" +}; + + +/* + * auto-generated functions for ISO-8859-2 .. ISO-8859-16 + */ + +static int ISO8859_2ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_2); +} +static int UTF8ToISO8859_2 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_2); +} + +static int ISO8859_3ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_3); +} +static int UTF8ToISO8859_3 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_3); +} + +static int ISO8859_4ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_4); +} +static int UTF8ToISO8859_4 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_4); +} + +static int ISO8859_5ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_5); +} +static int UTF8ToISO8859_5 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_5); +} + +static int ISO8859_6ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_6); +} +static int UTF8ToISO8859_6 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_6); +} + +static int ISO8859_7ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_7); +} +static int UTF8ToISO8859_7 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_7); +} + +static int ISO8859_8ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_8); +} +static int UTF8ToISO8859_8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_8); +} + +static int ISO8859_9ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_9); +} +static int UTF8ToISO8859_9 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_9); +} + +static int ISO8859_10ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_10); +} +static int UTF8ToISO8859_10 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_10); +} + +static int ISO8859_11ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_11); +} +static int UTF8ToISO8859_11 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_11); +} + +static int ISO8859_13ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_13); +} +static int UTF8ToISO8859_13 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_13); +} + +static int ISO8859_14ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_14); +} +static int UTF8ToISO8859_14 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_14); +} + +static int ISO8859_15ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_15); +} +static int UTF8ToISO8859_15 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_15); +} + +static int ISO8859_16ToUTF8 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return ISO8859xToUTF8 (out, outlen, in, inlen, xmlunicodetable_ISO8859_16); +} +static int UTF8ToISO8859_16 (unsigned char* out, int *outlen, + const unsigned char* in, int *inlen) { + return UTF8ToISO8859x (out, outlen, in, inlen, xmltranscodetable_ISO8859_16); +} + +static void +xmlRegisterCharEncodingHandlersISO8859x (void) { + xmlNewCharEncodingHandler ("ISO-8859-2", ISO8859_2ToUTF8, UTF8ToISO8859_2); + xmlNewCharEncodingHandler ("ISO-8859-3", ISO8859_3ToUTF8, UTF8ToISO8859_3); + xmlNewCharEncodingHandler ("ISO-8859-4", ISO8859_4ToUTF8, UTF8ToISO8859_4); + xmlNewCharEncodingHandler ("ISO-8859-5", ISO8859_5ToUTF8, UTF8ToISO8859_5); + xmlNewCharEncodingHandler ("ISO-8859-6", ISO8859_6ToUTF8, UTF8ToISO8859_6); + xmlNewCharEncodingHandler ("ISO-8859-7", ISO8859_7ToUTF8, UTF8ToISO8859_7); + xmlNewCharEncodingHandler ("ISO-8859-8", ISO8859_8ToUTF8, UTF8ToISO8859_8); + xmlNewCharEncodingHandler ("ISO-8859-9", ISO8859_9ToUTF8, UTF8ToISO8859_9); + xmlNewCharEncodingHandler ("ISO-8859-10", ISO8859_10ToUTF8, UTF8ToISO8859_10); + xmlNewCharEncodingHandler ("ISO-8859-11", ISO8859_11ToUTF8, UTF8ToISO8859_11); + xmlNewCharEncodingHandler ("ISO-8859-13", ISO8859_13ToUTF8, UTF8ToISO8859_13); + xmlNewCharEncodingHandler ("ISO-8859-14", ISO8859_14ToUTF8, UTF8ToISO8859_14); + xmlNewCharEncodingHandler ("ISO-8859-15", ISO8859_15ToUTF8, UTF8ToISO8859_15); + xmlNewCharEncodingHandler ("ISO-8859-16", ISO8859_16ToUTF8, UTF8ToISO8859_16); +} + +#endif +#endif + +#define bottom_encoding +#include "elfgcchack.h" |