include/ruby/encoding.h

   1 /**********************************************************************
   2
   3   encoding.h -
   4
   5   $Author: matz $
   6   created at: Thu May 24 11:49:41 JST 2007
   7
   8   Copyright (C) 2007 Yukihiro Matsumoto
   9
  10 **********************************************************************/
  11
  12 #ifndef RUBY_ENCODING_H
  13 #define RUBY_ENCODING_H 1
  14
  15 #include <stdarg.h>
  16 #include "ruby/oniguruma.h"
  17
  18 #define ENCODING_INLINE_MAX 1023
  19 #define ENCODING_SHIFT (FL_USHIFT+10)
  20 #define ENCODING_MASK (ENCODING_INLINE_MAX<<ENCODING_SHIFT)
  21
  22 #define ENCODING_SET_INLINED(obj,i) do {\
  23     RBASIC(obj)->flags &= ~ENCODING_MASK;\
  24     RBASIC(obj)->flags |= (i) << ENCODING_SHIFT;\
  25 } while (0)
  26 #define ENCODING_SET(obj,i) do {\
  27     VALUE rb_encoding_set_obj = (obj); \
  28     int encoding_set_enc_index = (i); \
  29     if (encoding_set_enc_index < ENCODING_INLINE_MAX) \
  30         ENCODING_SET_INLINED(rb_encoding_set_obj, encoding_set_enc_index); \
  31     else \
  32         rb_enc_set_index(rb_encoding_set_obj, encoding_set_enc_index); \
  33 } while (0)
  34
  35 #define ENCODING_GET_INLINED(obj) ((RBASIC(obj)->flags & ENCODING_MASK)>>ENCODING_SHIFT)
  36 #define ENCODING_GET(obj) \
  37     (ENCODING_GET_INLINED(obj) != ENCODING_INLINE_MAX ? \
  38      ENCODING_GET_INLINED(obj) : \
  39      rb_enc_get_index(obj))
  40
  41 #define ENCODING_IS_ASCII8BIT(obj) (ENCODING_GET_INLINED(obj) == 0)
  42
  43 #define ENCODING_MAXNAMELEN 42
  44
  45 #define ENC_CODERANGE_MASK      (FL_USER8|FL_USER9)
  46 #define ENC_CODERANGE_UNKNOWN   0
  47 #define ENC_CODERANGE_7BIT      FL_USER8
  48 #define ENC_CODERANGE_VALID     FL_USER9
  49 #define ENC_CODERANGE_BROKEN    (FL_USER8|FL_USER9)
  50 #define ENC_CODERANGE(obj) (RBASIC(obj)->flags & ENC_CODERANGE_MASK)
  51 #define ENC_CODERANGE_ASCIIONLY(obj) (ENC_CODERANGE(obj) == ENC_CODERANGE_7BIT)
  52 #define ENC_CODERANGE_SET(obj,cr) (RBASIC(obj)->flags = \
  53                                    (RBASIC(obj)->flags & ~ENC_CODERANGE_MASK) | (cr))
  54 #define ENC_CODERANGE_CLEAR(obj) ENC_CODERANGE_SET(obj,0)
  55
  56 /* assumed ASCII compatiblity */
  57 #define ENC_CODERANGE_AND(a, b) \
  58     (a == ENC_CODERANGE_7BIT ? b : \
  59      a == ENC_CODERANGE_VALID ? (b == ENC_CODERANGE_7BIT ? ENC_CODERANGE_VALID : b) : \
  60      ENC_CODERANGE_UNKNOWN)
  61
  62 #define ENCODING_CODERANGE_SET(obj, encindex, cr) \
  63     do { \
  64         VALUE rb_encoding_coderange_obj = (obj); \
  65         ENCODING_SET(rb_encoding_coderange_obj, (encindex)); \
  66         ENC_CODERANGE_SET(rb_encoding_coderange_obj, (cr)); \
  67     } while (0)
  68
  69 typedef OnigEncodingType rb_encoding;
  70
  71 int rb_enc_replicate(const char *, rb_encoding *);
  72 int rb_define_dummy_encoding(const char *);
  73 #define rb_enc_to_index(enc) ((enc) ? ENC_TO_ENCINDEX(enc) : 0)
  74 int rb_enc_get_index(VALUE obj);
  75 void rb_enc_set_index(VALUE obj, int encindex);
  76 int rb_enc_find_index(const char *name);
  77 int rb_to_encoding_index(VALUE);
  78 rb_encoding* rb_to_encoding(VALUE);
  79 rb_encoding* rb_enc_get(VALUE);
  80 rb_encoding* rb_enc_compatible(VALUE,VALUE);
  81 rb_encoding* rb_enc_check(VALUE,VALUE);
  82 VALUE rb_enc_associate_index(VALUE, int);
  83 VALUE rb_enc_associate(VALUE, rb_encoding*);
  84 void rb_enc_copy(VALUE dst, VALUE src);
  85
  86 VALUE rb_enc_str_new(const char*, long, rb_encoding*);
  87 VALUE rb_enc_reg_new(const char*, long, rb_encoding*, int);
  88 PRINTF_ARGS(VALUE rb_enc_sprintf(rb_encoding *, const char*, ...), 2, 3);
  89 VALUE rb_enc_vsprintf(rb_encoding *, const char*, va_list);
  90 long rb_enc_strlen(const char*, const char*, rb_encoding*);
  91 char* rb_enc_nth(const char*, const char*, int, rb_encoding*);
  92 VALUE rb_obj_encoding(VALUE);
  93 VALUE rb_enc_str_buf_cat(VALUE str, const char *ptr, long len, rb_encoding *enc);
  94
  95 /* index -> rb_encoding */
  96 rb_encoding* rb_enc_from_index(int idx);
  97
  98 /* name -> rb_encoding */
  99 rb_encoding * rb_enc_find(const char *name);
 100
 101 /* encoding -> name */
 102 #define rb_enc_name(enc) (enc)->name
 103
 104 /* encoding -> minlen/maxlen */
 105 #define rb_enc_mbminlen(enc) (enc)->min_enc_len
 106 #define rb_enc_mbmaxlen(enc) (enc)->max_enc_len
 107
 108 /* -> mbclen (no error notification: 0 < ret <= e-p, no exception) */
 109 int rb_enc_mbclen(const char *p, const char *e, rb_encoding *enc);
 110
 111 /* -> chlen, invalid or needmore */
 112 int rb_enc_precise_mbclen(const char *p, const char *e, rb_encoding *enc);
 113 #define MBCLEN_CHARFOUND_P(ret)     ONIGENC_MBCLEN_CHARFOUND_P(ret)
 114 #define MBCLEN_CHARFOUND_LEN(ret)     ONIGENC_MBCLEN_CHARFOUND_LEN(ret)
 115 #define MBCLEN_INVALID_P(ret)       ONIGENC_MBCLEN_INVALID_P(ret)
 116 #define MBCLEN_NEEDMORE_P(ret)      ONIGENC_MBCLEN_NEEDMORE_P(ret)
 117 #define MBCLEN_NEEDMORE_LEN(ret)      ONIGENC_MBCLEN_NEEDMORE_LEN(ret)
 118
 119 /* -> 0x00..0x7f, -1 */
 120 int rb_enc_ascget(const char *p, const char *e, int *len, rb_encoding *enc);
 121
 122 /* -> code or raise exception */
 123 int rb_enc_codepoint(const char *p, const char *e, rb_encoding *enc);
 124 #define rb_enc_mbc_to_codepoint(p, e, enc) ONIGENC_MBC_TO_CODE(enc,(UChar*)(p),(UChar*)(e))
 125
 126 /* -> codelen>0 or raise exception */
 127 int rb_enc_codelen(int code, rb_encoding *enc);
 128
 129 /* code,ptr,encoding -> write buf */
 130 #define rb_enc_mbcput(c,buf,enc) ONIGENC_CODE_TO_MBC(enc,c,(UChar*)(buf))
 131
 132 /* ptr, ptr, encoding -> prev_char */
 133 #define rb_enc_prev_char(s,p,enc) (char *)onigenc_get_prev_char_head(enc,(UChar*)(s),(UChar*)(p))
 134 /* ptr, ptr, encoding -> next_char */
 135 #define rb_enc_left_char_head(s,p,enc) (char *)onigenc_get_left_adjust_char_head(enc,(UChar*)(s),(UChar*)(p))
 136 #define rb_enc_right_char_head(s,p,enc) (char *)onigenc_get_right_adjust_char_head(enc,(UChar*)(s),(UChar*)(p))
 137
 138 /* ptr, ptr, encoding -> newline_or_not */
 139 #define rb_enc_is_newline(p,end,enc)  ONIGENC_IS_MBC_NEWLINE(enc,(UChar*)(p),(UChar*)(end))
 140
 141 #define rb_enc_isctype(c,t,enc) ONIGENC_IS_CODE_CTYPE(enc,c,t)
 142 #define rb_enc_isascii(c,enc) ONIGENC_IS_CODE_ASCII(c)
 143 #define rb_enc_isalpha(c,enc) ONIGENC_IS_CODE_ALPHA(enc,c)
 144 #define rb_enc_islower(c,enc) ONIGENC_IS_CODE_LOWER(enc,c)
 145 #define rb_enc_isupper(c,enc) ONIGENC_IS_CODE_UPPER(enc,c)
 146 #define rb_enc_ispunct(c,enc) ONIGENC_IS_CODE_PUNCT(enc,c)
 147 #define rb_enc_isalnum(c,enc) ONIGENC_IS_CODE_ALNUM(enc,c)
 148 #define rb_enc_isprint(c,enc) ONIGENC_IS_CODE_PRINT(enc,c)
 149 #define rb_enc_isspace(c,enc) ONIGENC_IS_CODE_SPACE(enc,c)
 150 #define rb_enc_isdigit(c,enc) ONIGENC_IS_CODE_DIGIT(enc,c)
 151
 152 #define rb_enc_asciicompat(enc) (rb_enc_mbminlen(enc)==1 && !rb_enc_dummy_p(enc))
 153
 154 int rb_enc_casefold(char *to, const char *p, const char *e, rb_encoding *enc);
 155 int rb_enc_toupper(int c, rb_encoding *enc);
 156 int rb_enc_tolower(int c, rb_encoding *enc);
 157 ID rb_intern3(const char*, long, rb_encoding*);
 158 ID rb_interned_id_p(const char *, long, rb_encoding *);
 159 int rb_enc_symname_p(const char*, rb_encoding*);
 160 int rb_enc_str_coderange(VALUE);
 161 long rb_str_coderange_scan_restartable(const char*, const char*, rb_encoding*, int*);
 162 int rb_enc_str_asciionly_p(VALUE);
 163 #define rb_enc_str_asciicompat_p(str) rb_enc_asciicompat(rb_enc_get(str))
 164 VALUE rb_enc_from_encoding(rb_encoding *enc);
 165 rb_encoding *rb_ascii8bit_encoding(void);
 166 rb_encoding *rb_utf8_encoding(void);
 167 rb_encoding *rb_usascii_encoding(void);
 168 rb_encoding *rb_locale_encoding(void);
 169 rb_encoding *rb_filesystem_encoding(void);
 170 rb_encoding *rb_default_external_encoding(void);
 171 int rb_ascii8bit_encindex(void);
 172 int rb_utf8_encindex(void);
 173 int rb_usascii_encindex(void);
 174 VALUE rb_enc_default_external(void);
 175 void rb_enc_set_default_external(VALUE encoding);
 176 VALUE rb_locale_charmap(VALUE klass);
 177 long rb_memsearch(const void*,long,const void*,long,rb_encoding*);
 178
 179 RUBY_EXTERN VALUE rb_cEncoding;
 180 #define enc_initialized_p(enc) ((enc)->ruby_encoding_index != ENC_UNINITIALIZED)
 181 #define ENC_DUMMY_FLAG (1<<24)
 182 #define ENC_INDEX_MASK (~(~0U<<24))
 183
 184 #define ENC_TO_ENCINDEX(enc)   ((enc)->ruby_encoding_index & ENC_INDEX_MASK)
 185 #define ENC_FROM_ENCINDEX(idx) (RARRAY_PTR(rb_encoding_list)[idx])
 186 #define ENC_FROM_ENCODING(enc) ENC_FROM_ENCINDEX(ENC_TO_ENCINDEX(enc))
 187
 188 #define ENC_DUMMY_P(enc) ((enc)->ruby_encoding_index & ENC_DUMMY_FLAG)
 189 #define ENC_SET_DUMMY(enc) ((enc)->ruby_encoding_index |= ENC_DUMMY_FLAG)
 190
 191 static inline int
 192 rb_enc_dummy_p(rb_encoding *enc)
 193 {
 194     return ENC_DUMMY_P(enc) != 0;
 195 }
 196
 197 VALUE rb_str_transcode(VALUE str, VALUE to);
 198
 199 /* econv stuff */
 200
 201 typedef enum {
 202     econv_invalid_byte_sequence,
 203     econv_undefined_conversion,
 204     econv_destination_buffer_full,
 205     econv_source_buffer_empty,
 206     econv_finished,
 207     econv_output_followed_by_input,
 208 } rb_econv_result_t;
 209
 210 typedef struct {
 211     const char *from;
 212     const char *to;
 213     struct rb_transcoding *tc;
 214     unsigned char *out_buf_start;
 215     unsigned char *out_data_start;
 216     unsigned char *out_data_end;
 217     unsigned char *out_buf_end;
 218     rb_econv_result_t last_result;
 219 } rb_econv_elem_t;
 220
 221 typedef struct {
 222     unsigned char *in_buf_start;
 223     unsigned char *in_data_start;
 224     unsigned char *in_data_end;
 225     unsigned char *in_buf_end;
 226     rb_econv_elem_t *elems;
 227     int num_trans;
 228     int num_finished;
 229     int last_trans_index; /* last trans, not including universal newline */
 230     struct rb_transcoding *last_tc;
 231
 232     /* last error */
 233     struct {
 234         rb_econv_result_t result;
 235         struct rb_transcoding *error_tc;
 236         const char *source_encoding;
 237         const char *destination_encoding;
 238         const unsigned char *error_bytes_start;
 239         size_t error_bytes_len;
 240         size_t readagain_len;
 241         int partial_input;
 242     } last_error;
 243
 244     /* The following fields are only for Encoding::Converter.
 245      * rb_econv_open set them NULL. */
 246     rb_encoding *source_encoding;
 247     rb_encoding *destination_encoding;
 248 } rb_econv_t;
 249
 250 rb_econv_t *rb_econv_open(const char *source_encoding, const char *destination_encoding, int flags);
 251 rb_econv_result_t rb_econv_convert(rb_econv_t *ec,
 252     const unsigned char **source_buffer_ptr, const unsigned char *source_buffer_end,
 253     unsigned char **destination_buffer_ptr, unsigned char *destination_buffer_end,
 254     int flags);
 255
 256
 257 /* result: 0:success -1:failure */
 258 int rb_econv_insert_output(rb_econv_t *ec,
 259     const unsigned char *str, size_t len, const char *str_encoding);
 260
 261 /* encoding that rb_econv_insert_output doesn't need conversion */
 262 const char *rb_econv_encoding_to_insert_output(rb_econv_t *ec);
 263
 264 void rb_econv_close(rb_econv_t *ec);
 265
 266 /* flags for rb_econv_open */
 267 #define ECONV_UNIVERSAL_NEWLINE_DECODER       0x100
 268 #define ECONV_CRLF_NEWLINE_ENCODER            0x200
 269 #define ECONV_CR_NEWLINE_ENCODER              0x400
 270
 271 /* flags for rb_econv_convert */
 272 #define ECONV_PARTIAL_INPUT                   0x10000
 273 #define ECONV_OUTPUT_FOLLOWED_BY_INPUT        0x20000
 274
 275 #endif /* RUBY_ENCODING_H */