00001
00002
00003
00004
00005
00006
00007
00008
00009
00010
00011
00012 #ifndef RUBY_ENCODING_H
00013 #define RUBY_ENCODING_H 1
00014
00015 #if defined(__cplusplus)
00016 extern "C" {
00017 #if 0
00018 }
00019 #endif
00020 #endif
00021
00022 #include <stdarg.h>
00023 #include "ruby/oniguruma.h"
00024
00025 #define ENCODING_INLINE_MAX 1023
00026 #define ENCODING_SHIFT (FL_USHIFT+10)
00027 #define ENCODING_MASK (((VALUE)ENCODING_INLINE_MAX)<<ENCODING_SHIFT)
00028
00029 #define ENCODING_SET_INLINED(obj,i) do {\
00030 RBASIC(obj)->flags &= ~ENCODING_MASK;\
00031 RBASIC(obj)->flags |= (VALUE)(i) << ENCODING_SHIFT;\
00032 } while (0)
00033 #define ENCODING_SET(obj,i) do {\
00034 VALUE rb_encoding_set_obj = (obj); \
00035 int encoding_set_enc_index = (i); \
00036 if (encoding_set_enc_index < ENCODING_INLINE_MAX) \
00037 ENCODING_SET_INLINED(rb_encoding_set_obj, encoding_set_enc_index); \
00038 else \
00039 rb_enc_set_index(rb_encoding_set_obj, encoding_set_enc_index); \
00040 } while (0)
00041
00042 #define ENCODING_GET_INLINED(obj) (int)((RBASIC(obj)->flags & ENCODING_MASK)>>ENCODING_SHIFT)
00043 #define ENCODING_GET(obj) \
00044 (ENCODING_GET_INLINED(obj) != ENCODING_INLINE_MAX ? \
00045 ENCODING_GET_INLINED(obj) : \
00046 rb_enc_get_index(obj))
00047
00048 #define ENCODING_IS_ASCII8BIT(obj) (ENCODING_GET_INLINED(obj) == 0)
00049
00050 #define ENCODING_MAXNAMELEN 42
00051
00052 #define ENC_CODERANGE_MASK ((int)(FL_USER8|FL_USER9))
00053 #define ENC_CODERANGE_UNKNOWN 0
00054 #define ENC_CODERANGE_7BIT ((int)FL_USER8)
00055 #define ENC_CODERANGE_VALID ((int)FL_USER9)
00056 #define ENC_CODERANGE_BROKEN ((int)(FL_USER8|FL_USER9))
00057 #define ENC_CODERANGE(obj) ((int)RBASIC(obj)->flags & ENC_CODERANGE_MASK)
00058 #define ENC_CODERANGE_ASCIIONLY(obj) (ENC_CODERANGE(obj) == ENC_CODERANGE_7BIT)
00059 #define ENC_CODERANGE_SET(obj,cr) (RBASIC(obj)->flags = \
00060 (RBASIC(obj)->flags & ~ENC_CODERANGE_MASK) | (cr))
00061 #define ENC_CODERANGE_CLEAR(obj) ENC_CODERANGE_SET(obj,0)
00062
00063
00064 #define ENC_CODERANGE_AND(a, b) \
00065 (a == ENC_CODERANGE_7BIT ? b : \
00066 a == ENC_CODERANGE_VALID ? (b == ENC_CODERANGE_7BIT ? ENC_CODERANGE_VALID : b) : \
00067 ENC_CODERANGE_UNKNOWN)
00068
00069 #define ENCODING_CODERANGE_SET(obj, encindex, cr) \
00070 do { \
00071 VALUE rb_encoding_coderange_obj = (obj); \
00072 ENCODING_SET(rb_encoding_coderange_obj, (encindex)); \
00073 ENC_CODERANGE_SET(rb_encoding_coderange_obj, (cr)); \
00074 } while (0)
00075
00076 typedef OnigEncodingType rb_encoding;
00077
00078 int rb_enc_replicate(const char *, rb_encoding *);
00079 int rb_define_dummy_encoding(const char *);
00080 #define rb_enc_to_index(enc) ((enc) ? ENC_TO_ENCINDEX(enc) : 0)
00081 int rb_enc_get_index(VALUE obj);
00082 void rb_enc_set_index(VALUE obj, int encindex);
00083 int rb_enc_find_index(const char *name);
00084 int rb_to_encoding_index(VALUE);
00085 rb_encoding* rb_to_encoding(VALUE);
00086 rb_encoding* rb_enc_get(VALUE);
00087 rb_encoding* rb_enc_compatible(VALUE,VALUE);
00088 rb_encoding* rb_enc_check(VALUE,VALUE);
00089 VALUE rb_enc_associate_index(VALUE, int);
00090 VALUE rb_enc_associate(VALUE, rb_encoding*);
00091 void rb_enc_copy(VALUE dst, VALUE src);
00092
00093 VALUE rb_enc_str_new(const char*, long, rb_encoding*);
00094 VALUE rb_enc_reg_new(const char*, long, rb_encoding*, int);
00095 PRINTF_ARGS(VALUE rb_enc_sprintf(rb_encoding *, const char*, ...), 2, 3);
00096 VALUE rb_enc_vsprintf(rb_encoding *, const char*, va_list);
00097 long rb_enc_strlen(const char*, const char*, rb_encoding*);
00098 char* rb_enc_nth(const char*, const char*, long, rb_encoding*);
00099 VALUE rb_obj_encoding(VALUE);
00100 VALUE rb_enc_str_buf_cat(VALUE str, const char *ptr, long len, rb_encoding *enc);
00101 VALUE rb_enc_uint_chr(unsigned int code, rb_encoding *enc);
00102
00103 VALUE rb_external_str_new_with_enc(const char *ptr, long len, rb_encoding *);
00104 VALUE rb_str_export_to_enc(VALUE, rb_encoding *);
00105 VALUE rb_str_conv_enc(VALUE str, rb_encoding *from, rb_encoding *to);
00106 VALUE rb_str_conv_enc_opts(VALUE str, rb_encoding *from, rb_encoding *to, int ecflags, VALUE ecopts);
00107
00108
00109 rb_encoding* rb_enc_from_index(int idx);
00110
00111
00112 rb_encoding * rb_enc_find(const char *name);
00113
00114
00115 #define rb_enc_name(enc) (enc)->name
00116
00117
00118 #define rb_enc_mbminlen(enc) (enc)->min_enc_len
00119 #define rb_enc_mbmaxlen(enc) (enc)->max_enc_len
00120
00121
00122 int rb_enc_mbclen(const char *p, const char *e, rb_encoding *enc);
00123
00124
00125 int rb_enc_fast_mbclen(const char *p, const char *e, rb_encoding *enc);
00126
00127
00128 int rb_enc_precise_mbclen(const char *p, const char *e, rb_encoding *enc);
00129 #define MBCLEN_CHARFOUND_P(ret) ONIGENC_MBCLEN_CHARFOUND_P(ret)
00130 #define MBCLEN_CHARFOUND_LEN(ret) ONIGENC_MBCLEN_CHARFOUND_LEN(ret)
00131 #define MBCLEN_INVALID_P(ret) ONIGENC_MBCLEN_INVALID_P(ret)
00132 #define MBCLEN_NEEDMORE_P(ret) ONIGENC_MBCLEN_NEEDMORE_P(ret)
00133 #define MBCLEN_NEEDMORE_LEN(ret) ONIGENC_MBCLEN_NEEDMORE_LEN(ret)
00134
00135
00136 int rb_enc_ascget(const char *p, const char *e, int *len, rb_encoding *enc);
00137
00138
00139
00140 unsigned int rb_enc_codepoint_len(const char *p, const char *e, int *len, rb_encoding *enc);
00141
00142
00143 unsigned int rb_enc_codepoint(const char *p, const char *e, rb_encoding *enc);
00144
00145 #define rb_enc_codepoint(p,e,enc) rb_enc_codepoint_len((p),(e),0,(enc))
00146 #define rb_enc_mbc_to_codepoint(p, e, enc) ONIGENC_MBC_TO_CODE(enc,(UChar*)(p),(UChar*)(e))
00147
00148
00149 int rb_enc_codelen(int code, rb_encoding *enc);
00150
00151
00152 #define rb_enc_mbcput(c,buf,enc) ONIGENC_CODE_TO_MBC(enc,c,(UChar*)(buf))
00153
00154
00155 #define rb_enc_prev_char(s,p,e,enc) (char *)onigenc_get_prev_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
00156
00157 #define rb_enc_left_char_head(s,p,e,enc) (char *)onigenc_get_left_adjust_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
00158 #define rb_enc_right_char_head(s,p,e,enc) (char *)onigenc_get_right_adjust_char_head(enc,(UChar*)(s),(UChar*)(p),(UChar*)(e))
00159
00160
00161 #define rb_enc_is_newline(p,end,enc) ONIGENC_IS_MBC_NEWLINE(enc,(UChar*)(p),(UChar*)(end))
00162
00163 #define rb_enc_isctype(c,t,enc) ONIGENC_IS_CODE_CTYPE(enc,c,t)
00164 #define rb_enc_isascii(c,enc) ONIGENC_IS_CODE_ASCII(c)
00165 #define rb_enc_isalpha(c,enc) ONIGENC_IS_CODE_ALPHA(enc,c)
00166 #define rb_enc_islower(c,enc) ONIGENC_IS_CODE_LOWER(enc,c)
00167 #define rb_enc_isupper(c,enc) ONIGENC_IS_CODE_UPPER(enc,c)
00168 #define rb_enc_ispunct(c,enc) ONIGENC_IS_CODE_PUNCT(enc,c)
00169 #define rb_enc_isalnum(c,enc) ONIGENC_IS_CODE_ALNUM(enc,c)
00170 #define rb_enc_isprint(c,enc) ONIGENC_IS_CODE_PRINT(enc,c)
00171 #define rb_enc_isspace(c,enc) ONIGENC_IS_CODE_SPACE(enc,c)
00172 #define rb_enc_isdigit(c,enc) ONIGENC_IS_CODE_DIGIT(enc,c)
00173
00174 #define rb_enc_asciicompat(enc) (rb_enc_mbminlen(enc)==1 && !rb_enc_dummy_p(enc))
00175
00176 int rb_enc_casefold(char *to, const char *p, const char *e, rb_encoding *enc);
00177 int rb_enc_toupper(int c, rb_encoding *enc);
00178 int rb_enc_tolower(int c, rb_encoding *enc);
00179 ID rb_intern3(const char*, long, rb_encoding*);
00180 ID rb_interned_id_p(const char *, long, rb_encoding *);
00181 int rb_enc_symname_p(const char*, rb_encoding*);
00182 int rb_enc_symname2_p(const char*, long, rb_encoding*);
00183 int rb_enc_str_coderange(VALUE);
00184 long rb_str_coderange_scan_restartable(const char*, const char*, rb_encoding*, int*);
00185 int rb_enc_str_asciionly_p(VALUE);
00186 #define rb_enc_str_asciicompat_p(str) rb_enc_asciicompat(rb_enc_get(str))
00187 VALUE rb_enc_from_encoding(rb_encoding *enc);
00188 int rb_enc_unicode_p(rb_encoding *enc);
00189 rb_encoding *rb_ascii8bit_encoding(void);
00190 rb_encoding *rb_utf8_encoding(void);
00191 rb_encoding *rb_usascii_encoding(void);
00192 rb_encoding *rb_locale_encoding(void);
00193 rb_encoding *rb_filesystem_encoding(void);
00194 rb_encoding *rb_default_external_encoding(void);
00195 rb_encoding *rb_default_internal_encoding(void);
00196 int rb_ascii8bit_encindex(void);
00197 int rb_utf8_encindex(void);
00198 int rb_usascii_encindex(void);
00199 int rb_locale_encindex(void);
00200 int rb_filesystem_encindex(void);
00201 VALUE rb_enc_default_external(void);
00202 VALUE rb_enc_default_internal(void);
00203 void rb_enc_set_default_external(VALUE encoding);
00204 void rb_enc_set_default_internal(VALUE encoding);
00205 VALUE rb_locale_charmap(VALUE klass);
00206 long rb_memsearch(const void*,long,const void*,long,rb_encoding*);
00207
00208 RUBY_EXTERN VALUE rb_cEncoding;
00209 #define ENC_DUMMY_FLAG (1<<24)
00210 #define ENC_INDEX_MASK (~(~0U<<24))
00211
00212 #define ENC_TO_ENCINDEX(enc) (int)((enc)->ruby_encoding_index & ENC_INDEX_MASK)
00213
00214 #define ENC_DUMMY_P(enc) ((enc)->ruby_encoding_index & ENC_DUMMY_FLAG)
00215 #define ENC_SET_DUMMY(enc) ((enc)->ruby_encoding_index |= ENC_DUMMY_FLAG)
00216
00217 static inline int
00218 rb_enc_dummy_p(rb_encoding *enc)
00219 {
00220 return ENC_DUMMY_P(enc) != 0;
00221 }
00222
00223
00224
00225 typedef enum {
00226 econv_invalid_byte_sequence,
00227 econv_undefined_conversion,
00228 econv_destination_buffer_full,
00229 econv_source_buffer_empty,
00230 econv_finished,
00231 econv_after_output,
00232 econv_incomplete_input
00233 } rb_econv_result_t;
00234
00235 typedef struct rb_econv_t rb_econv_t;
00236
00237 VALUE rb_str_encode(VALUE str, VALUE to, int ecflags, VALUE ecopts);
00238 int rb_econv_has_convpath_p(const char* from_encoding, const char* to_encoding);
00239
00240 int rb_econv_prepare_opts(VALUE opthash, VALUE *ecopts);
00241
00242 rb_econv_t *rb_econv_open(const char *source_encoding, const char *destination_encoding, int ecflags);
00243 rb_econv_t *rb_econv_open_opts(const char *source_encoding, const char *destination_encoding, int ecflags, VALUE ecopts);
00244
00245 rb_econv_result_t rb_econv_convert(rb_econv_t *ec,
00246 const unsigned char **source_buffer_ptr, const unsigned char *source_buffer_end,
00247 unsigned char **destination_buffer_ptr, unsigned char *destination_buffer_end,
00248 int flags);
00249 void rb_econv_close(rb_econv_t *ec);
00250
00251
00252 int rb_econv_set_replacement(rb_econv_t *ec, const unsigned char *str, size_t len, const char *encname);
00253
00254
00255 int rb_econv_decorate_at_first(rb_econv_t *ec, const char *decorator_name);
00256 int rb_econv_decorate_at_last(rb_econv_t *ec, const char *decorator_name);
00257
00258 VALUE rb_econv_open_exc(const char *senc, const char *denc, int ecflags);
00259
00260
00261 int rb_econv_insert_output(rb_econv_t *ec,
00262 const unsigned char *str, size_t len, const char *str_encoding);
00263
00264
00265 const char *rb_econv_encoding_to_insert_output(rb_econv_t *ec);
00266
00267
00268 void rb_econv_check_error(rb_econv_t *ec);
00269
00270
00271 VALUE rb_econv_make_exception(rb_econv_t *ec);
00272
00273 int rb_econv_putbackable(rb_econv_t *ec);
00274 void rb_econv_putback(rb_econv_t *ec, unsigned char *p, int n);
00275
00276
00277
00278 const char *rb_econv_asciicompat_encoding(const char *encname);
00279
00280 VALUE rb_econv_str_convert(rb_econv_t *ec, VALUE src, int flags);
00281 VALUE rb_econv_substr_convert(rb_econv_t *ec, VALUE src, long byteoff, long bytesize, int flags);
00282 VALUE rb_econv_str_append(rb_econv_t *ec, VALUE src, VALUE dst, int flags);
00283 VALUE rb_econv_substr_append(rb_econv_t *ec, VALUE src, long byteoff, long bytesize, VALUE dst, int flags);
00284
00285 void rb_econv_binmode(rb_econv_t *ec);
00286
00287
00288
00289 #define ECONV_ERROR_HANDLER_MASK 0x000000ff
00290
00291 #define ECONV_INVALID_MASK 0x0000000f
00292 #define ECONV_INVALID_REPLACE 0x00000002
00293
00294 #define ECONV_UNDEF_MASK 0x000000f0
00295 #define ECONV_UNDEF_REPLACE 0x00000020
00296 #define ECONV_UNDEF_HEX_CHARREF 0x00000030
00297
00298 #define ECONV_DECORATOR_MASK 0x0000ff00
00299
00300 #define ECONV_UNIVERSAL_NEWLINE_DECORATOR 0x00000100
00301 #define ECONV_CRLF_NEWLINE_DECORATOR 0x00001000
00302 #define ECONV_CR_NEWLINE_DECORATOR 0x00002000
00303 #define ECONV_XML_TEXT_DECORATOR 0x00004000
00304 #define ECONV_XML_ATTR_CONTENT_DECORATOR 0x00008000
00305
00306 #define ECONV_STATEFUL_DECORATOR_MASK 0x00f00000
00307 #define ECONV_XML_ATTR_QUOTE_DECORATOR 0x00100000
00308
00309
00310
00311
00312 #define ECONV_PARTIAL_INPUT 0x00010000
00313 #define ECONV_AFTER_OUTPUT 0x00020000
00314
00315
00316 #if defined(__cplusplus)
00317 #if 0
00318 {
00319 #endif
00320 }
00321 #endif
00322
00323 #endif
00324