Mercurial > cpython-lru_cache-weakref
annotate Python/formatter_unicode.c @ 72493:c4dcb68e0711
Check for PyUnicode_CopyCharacters() failure
| author | Victor Stinner <victor.stinner@haypocalc.com> |
|---|---|
| date | Wed, 28 Sep 2011 21:39:17 +0200 |
| parents | 8beaa9a37387 |
| children | cf3557b65ae1 |
| rev | line source |
|---|---|
|
42817
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
1 /* implements the unicode (as opposed to string) version of the |
|
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
2 built-in formatters for string, int, float. that is, the versions |
|
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
3 of int.__float__, etc., that take and return unicode objects */ |
|
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
4 |
|
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
5 #include "Python.h" |
| 72475 | 6 #include <locale.h> |
| 7 | |
| 8 /* Raises an exception about an unknown presentation type for this | |
| 9 * type. */ | |
| 10 | |
| 11 static void | |
| 12 unknown_presentation_type(Py_UCS4 presentation_type, | |
| 13 const char* type_name) | |
| 14 { | |
| 15 /* %c might be out-of-range, hence the two cases. */ | |
| 16 if (presentation_type > 32 && presentation_type < 128) | |
| 17 PyErr_Format(PyExc_ValueError, | |
| 18 "Unknown format code '%c' " | |
| 19 "for object of type '%.200s'", | |
| 20 (char)presentation_type, | |
| 21 type_name); | |
| 22 else | |
| 23 PyErr_Format(PyExc_ValueError, | |
| 24 "Unknown format code '\\x%x' " | |
| 25 "for object of type '%.200s'", | |
| 26 (unsigned int)presentation_type, | |
| 27 type_name); | |
| 28 } | |
| 29 | |
| 30 static void | |
| 31 invalid_comma_type(Py_UCS4 presentation_type) | |
| 32 { | |
| 33 if (presentation_type > 32 && presentation_type < 128) | |
| 34 PyErr_Format(PyExc_ValueError, | |
| 35 "Cannot specify ',' with '%c'.", | |
| 36 (char)presentation_type); | |
| 37 else | |
| 38 PyErr_Format(PyExc_ValueError, | |
| 39 "Cannot specify ',' with '\\x%x'.", | |
| 40 (unsigned int)presentation_type); | |
| 41 } | |
| 42 | |
| 43 /* | |
| 44 get_integer consumes 0 or more decimal digit characters from an | |
| 45 input string, updates *result with the corresponding positive | |
| 46 integer, and returns the number of digits consumed. | |
| 47 | |
| 48 returns -1 on error. | |
| 49 */ | |
| 50 static int | |
| 51 get_integer(PyObject *str, Py_ssize_t *pos, Py_ssize_t end, | |
| 52 Py_ssize_t *result) | |
| 53 { | |
| 54 Py_ssize_t accumulator, digitval, oldaccumulator; | |
| 55 int numdigits; | |
| 56 accumulator = numdigits = 0; | |
| 57 for (;;(*pos)++, numdigits++) { | |
| 58 if (*pos >= end) | |
| 59 break; | |
| 60 digitval = Py_UNICODE_TODECIMAL(PyUnicode_READ_CHAR(str, *pos)); | |
| 61 if (digitval < 0) | |
| 62 break; | |
| 63 /* | |
| 64 This trick was copied from old Unicode format code. It's cute, | |
| 65 but would really suck on an old machine with a slow divide | |
| 66 implementation. Fortunately, in the normal case we do not | |
| 67 expect too many digits. | |
| 68 */ | |
| 69 oldaccumulator = accumulator; | |
| 70 accumulator *= 10; | |
| 71 if ((accumulator+10)/10 != oldaccumulator+1) { | |
| 72 PyErr_Format(PyExc_ValueError, | |
| 73 "Too many decimal digits in format string"); | |
| 74 return -1; | |
| 75 } | |
| 76 accumulator += digitval; | |
| 77 } | |
| 78 *result = accumulator; | |
| 79 return numdigits; | |
| 80 } | |
| 81 | |
| 82 /************************************************************************/ | |
| 83 /*********** standard format specifier parsing **************************/ | |
| 84 /************************************************************************/ | |
| 85 | |
| 86 /* returns true if this character is a specifier alignment token */ | |
| 87 Py_LOCAL_INLINE(int) | |
| 88 is_alignment_token(Py_UCS4 c) | |
| 89 { | |
| 90 switch (c) { | |
| 91 case '<': case '>': case '=': case '^': | |
| 92 return 1; | |
| 93 default: | |
| 94 return 0; | |
| 95 } | |
| 96 } | |
| 97 | |
| 98 /* returns true if this character is a sign element */ | |
| 99 Py_LOCAL_INLINE(int) | |
| 100 is_sign_element(Py_UCS4 c) | |
| 101 { | |
| 102 switch (c) { | |
| 103 case ' ': case '+': case '-': | |
| 104 return 1; | |
| 105 default: | |
| 106 return 0; | |
| 107 } | |
| 108 } | |
| 109 | |
| 110 | |
| 111 typedef struct { | |
| 112 Py_UCS4 fill_char; | |
| 113 Py_UCS4 align; | |
| 114 int alternate; | |
| 115 Py_UCS4 sign; | |
| 116 Py_ssize_t width; | |
| 117 int thousands_separators; | |
| 118 Py_ssize_t precision; | |
| 119 Py_UCS4 type; | |
| 120 } InternalFormatSpec; | |
| 121 | |
| 122 #if 0 | |
| 123 /* Occassionally useful for debugging. Should normally be commented out. */ | |
| 124 static void | |
| 125 DEBUG_PRINT_FORMAT_SPEC(InternalFormatSpec *format) | |
| 126 { | |
| 127 printf("internal format spec: fill_char %d\n", format->fill_char); | |
| 128 printf("internal format spec: align %d\n", format->align); | |
| 129 printf("internal format spec: alternate %d\n", format->alternate); | |
| 130 printf("internal format spec: sign %d\n", format->sign); | |
| 131 printf("internal format spec: width %zd\n", format->width); | |
| 132 printf("internal format spec: thousands_separators %d\n", | |
| 133 format->thousands_separators); | |
| 134 printf("internal format spec: precision %zd\n", format->precision); | |
| 135 printf("internal format spec: type %c\n", format->type); | |
| 136 printf("\n"); | |
| 137 } | |
| 138 #endif | |
| 139 | |
| 140 | |
| 141 /* | |
| 142 ptr points to the start of the format_spec, end points just past its end. | |
| 143 fills in format with the parsed information. | |
| 144 returns 1 on success, 0 on failure. | |
| 145 if failure, sets the exception | |
| 146 */ | |
| 147 static int | |
| 148 parse_internal_render_format_spec(PyObject *format_spec, | |
| 149 Py_ssize_t start, Py_ssize_t end, | |
| 150 InternalFormatSpec *format, | |
| 151 char default_type, | |
| 152 char default_align) | |
| 153 { | |
| 154 Py_ssize_t pos = start; | |
| 155 /* end-pos is used throughout this code to specify the length of | |
| 156 the input string */ | |
| 157 #define READ_spec(index) PyUnicode_READ_CHAR(format_spec, index) | |
| 158 | |
| 159 Py_ssize_t consumed; | |
| 160 int align_specified = 0; | |
| 161 | |
| 162 format->fill_char = '\0'; | |
| 163 format->align = default_align; | |
| 164 format->alternate = 0; | |
| 165 format->sign = '\0'; | |
| 166 format->width = -1; | |
| 167 format->thousands_separators = 0; | |
| 168 format->precision = -1; | |
| 169 format->type = default_type; | |
| 170 | |
| 171 /* If the second char is an alignment token, | |
| 172 then parse the fill char */ | |
| 173 if (end-pos >= 2 && is_alignment_token(READ_spec(pos+1))) { | |
| 174 format->align = READ_spec(pos+1); | |
| 175 format->fill_char = READ_spec(pos); | |
| 176 align_specified = 1; | |
| 177 pos += 2; | |
| 178 } | |
| 179 else if (end-pos >= 1 && is_alignment_token(READ_spec(pos))) { | |
| 180 format->align = READ_spec(pos); | |
| 181 align_specified = 1; | |
| 182 ++pos; | |
| 183 } | |
| 184 | |
| 185 /* Parse the various sign options */ | |
| 186 if (end-pos >= 1 && is_sign_element(READ_spec(pos))) { | |
| 187 format->sign = READ_spec(pos); | |
| 188 ++pos; | |
| 189 } | |
| 190 | |
| 191 /* If the next character is #, we're in alternate mode. This only | |
| 192 applies to integers. */ | |
| 193 if (end-pos >= 1 && READ_spec(pos) == '#') { | |
| 194 format->alternate = 1; | |
| 195 ++pos; | |
| 196 } | |
| 197 | |
| 198 /* The special case for 0-padding (backwards compat) */ | |
| 199 if (format->fill_char == '\0' && end-pos >= 1 && READ_spec(pos) == '0') { | |
| 200 format->fill_char = '0'; | |
| 201 if (!align_specified) { | |
| 202 format->align = '='; | |
| 203 } | |
| 204 ++pos; | |
| 205 } | |
| 206 | |
| 207 consumed = get_integer(format_spec, &pos, end, &format->width); | |
| 208 if (consumed == -1) | |
| 209 /* Overflow error. Exception already set. */ | |
| 210 return 0; | |
| 211 | |
| 212 /* If consumed is 0, we didn't consume any characters for the | |
| 213 width. In that case, reset the width to -1, because | |
| 214 get_integer() will have set it to zero. -1 is how we record | |
| 215 that the width wasn't specified. */ | |
| 216 if (consumed == 0) | |
| 217 format->width = -1; | |
| 218 | |
| 219 /* Comma signifies add thousands separators */ | |
| 220 if (end-pos && READ_spec(pos) == ',') { | |
| 221 format->thousands_separators = 1; | |
| 222 ++pos; | |
| 223 } | |
| 224 | |
| 225 /* Parse field precision */ | |
| 226 if (end-pos && READ_spec(pos) == '.') { | |
| 227 ++pos; | |
| 228 | |
| 229 consumed = get_integer(format_spec, &pos, end, &format->precision); | |
| 230 if (consumed == -1) | |
| 231 /* Overflow error. Exception already set. */ | |
| 232 return 0; | |
| 233 | |
| 234 /* Not having a precision after a dot is an error. */ | |
| 235 if (consumed == 0) { | |
| 236 PyErr_Format(PyExc_ValueError, | |
| 237 "Format specifier missing precision"); | |
| 238 return 0; | |
| 239 } | |
| 240 | |
| 241 } | |
| 242 | |
| 243 /* Finally, parse the type field. */ | |
| 244 | |
| 245 if (end-pos > 1) { | |
| 246 /* More than one char remain, invalid conversion spec. */ | |
| 247 PyErr_Format(PyExc_ValueError, "Invalid conversion specification"); | |
| 248 return 0; | |
| 249 } | |
| 250 | |
| 251 if (end-pos == 1) { | |
| 252 format->type = READ_spec(pos); | |
| 253 ++pos; | |
| 254 } | |
| 255 | |
| 256 /* Do as much validating as we can, just by looking at the format | |
| 257 specifier. Do not take into account what type of formatting | |
| 258 we're doing (int, float, string). */ | |
| 259 | |
| 260 if (format->thousands_separators) { | |
| 261 switch (format->type) { | |
| 262 case 'd': | |
| 263 case 'e': | |
| 264 case 'f': | |
| 265 case 'g': | |
| 266 case 'E': | |
| 267 case 'G': | |
| 268 case '%': | |
| 269 case 'F': | |
| 270 case '\0': | |
| 271 /* These are allowed. See PEP 378.*/ | |
| 272 break; | |
| 273 default: | |
| 274 invalid_comma_type(format->type); | |
| 275 return 0; | |
| 276 } | |
| 277 } | |
| 278 | |
| 279 if (format->fill_char > 127 || format->align > 127 || | |
| 280 format->sign > 127) { | |
| 281 PyErr_SetString(PyExc_ValueError, "fill character too large"); | |
| 282 return 0; | |
| 283 } | |
| 284 | |
| 285 return 1; | |
| 286 } | |
| 287 | |
| 288 /* Calculate the padding needed. */ | |
| 289 static void | |
| 290 calc_padding(Py_ssize_t nchars, Py_ssize_t width, Py_UCS4 align, | |
| 291 Py_ssize_t *n_lpadding, Py_ssize_t *n_rpadding, | |
| 292 Py_ssize_t *n_total) | |
| 293 { | |
| 294 if (width >= 0) { | |
| 295 if (nchars > width) | |
| 296 *n_total = nchars; | |
| 297 else | |
| 298 *n_total = width; | |
| 299 } | |
| 300 else { | |
| 301 /* not specified, use all of the chars and no more */ | |
| 302 *n_total = nchars; | |
| 303 } | |
| 304 | |
| 305 /* Figure out how much leading space we need, based on the | |
| 306 aligning */ | |
| 307 if (align == '>') | |
| 308 *n_lpadding = *n_total - nchars; | |
| 309 else if (align == '^') | |
| 310 *n_lpadding = (*n_total - nchars) / 2; | |
| 311 else if (align == '<' || align == '=') | |
| 312 *n_lpadding = 0; | |
| 313 else { | |
| 314 /* We should never have an unspecified alignment. */ | |
| 315 *n_lpadding = 0; | |
| 316 assert(0); | |
| 317 } | |
| 318 | |
| 319 *n_rpadding = *n_total - nchars - *n_lpadding; | |
| 320 } | |
| 321 | |
| 322 static void | |
| 323 unicode_fill(PyObject *str, Py_ssize_t start, Py_ssize_t end, Py_UCS4 ch) | |
| 324 { | |
| 325 int kind = PyUnicode_KIND(str); | |
| 326 void *data = PyUnicode_DATA(str); | |
| 327 while (start < end) | |
| 328 PyUnicode_WRITE(kind, data, start++, ch); | |
| 329 } | |
| 330 | |
| 331 /* Do the padding, and return a pointer to where the caller-supplied | |
| 332 content goes. */ | |
| 333 static Py_ssize_t | |
| 334 fill_padding(PyObject *s, Py_ssize_t start, Py_ssize_t nchars, | |
| 335 Py_UCS4 fill_char, Py_ssize_t n_lpadding, | |
| 336 Py_ssize_t n_rpadding) | |
| 337 { | |
| 338 /* Pad on left. */ | |
| 339 if (n_lpadding) | |
| 340 unicode_fill(s, start, start + n_lpadding, fill_char); | |
| 341 | |
| 342 /* Pad on right. */ | |
| 343 if (n_rpadding) | |
| 344 unicode_fill(s, start + nchars + n_lpadding, | |
| 345 start + nchars + n_lpadding + n_rpadding, fill_char); | |
| 346 | |
| 347 /* Pointer to the user content. */ | |
| 348 return start + n_lpadding; | |
| 349 } | |
| 350 | |
| 351 /************************************************************************/ | |
| 352 /*********** common routines for numeric formatting *********************/ | |
| 353 /************************************************************************/ | |
| 354 | |
| 355 /* Locale type codes. */ | |
| 356 #define LT_CURRENT_LOCALE 0 | |
| 357 #define LT_DEFAULT_LOCALE 1 | |
| 358 #define LT_NO_LOCALE 2 | |
| 359 | |
| 360 /* Locale info needed for formatting integers and the part of floats | |
| 361 before and including the decimal. Note that locales only support | |
| 362 8-bit chars, not unicode. */ | |
| 363 typedef struct { | |
| 364 char *decimal_point; | |
| 365 char *thousands_sep; | |
| 366 char *grouping; | |
| 367 } LocaleInfo; | |
| 368 | |
| 369 /* describes the layout for an integer, see the comment in | |
| 370 calc_number_widths() for details */ | |
| 371 typedef struct { | |
| 372 Py_ssize_t n_lpadding; | |
| 373 Py_ssize_t n_prefix; | |
| 374 Py_ssize_t n_spadding; | |
| 375 Py_ssize_t n_rpadding; | |
| 376 char sign; | |
| 377 Py_ssize_t n_sign; /* number of digits needed for sign (0/1) */ | |
| 378 Py_ssize_t n_grouped_digits; /* Space taken up by the digits, including | |
| 379 any grouping chars. */ | |
| 380 Py_ssize_t n_decimal; /* 0 if only an integer */ | |
| 381 Py_ssize_t n_remainder; /* Digits in decimal and/or exponent part, | |
| 382 excluding the decimal itself, if | |
| 383 present. */ | |
| 384 | |
| 385 /* These 2 are not the widths of fields, but are needed by | |
| 386 STRINGLIB_GROUPING. */ | |
| 387 Py_ssize_t n_digits; /* The number of digits before a decimal | |
| 388 or exponent. */ | |
| 389 Py_ssize_t n_min_width; /* The min_width we used when we computed | |
| 390 the n_grouped_digits width. */ | |
| 391 } NumberFieldWidths; | |
|
42817
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
392 |
|
47393
0da27fdf0192
Refactor and clean up str.format() code (and helpers) in advance of optimizations.
Eric Smith <eric@trueblade.com>
parents:
42817
diff
changeset
|
393 |
| 72475 | 394 /* Given a number of the form: |
| 395 digits[remainder] | |
| 396 where ptr points to the start and end points to the end, find where | |
| 397 the integer part ends. This could be a decimal, an exponent, both, | |
| 398 or neither. | |
| 399 If a decimal point is present, set *has_decimal and increment | |
| 400 remainder beyond it. | |
| 401 Results are undefined (but shouldn't crash) for improperly | |
| 402 formatted strings. | |
| 403 */ | |
| 404 static void | |
| 405 parse_number(PyObject *s, Py_ssize_t pos, Py_ssize_t end, | |
| 406 Py_ssize_t *n_remainder, int *has_decimal) | |
| 407 { | |
| 408 Py_ssize_t remainder; | |
| 409 | |
| 410 while (pos<end && isdigit(PyUnicode_READ_CHAR(s, pos))) | |
| 411 ++pos; | |
| 412 remainder = pos; | |
| 413 | |
| 414 /* Does remainder start with a decimal point? */ | |
| 415 *has_decimal = pos<end && PyUnicode_READ_CHAR(s, remainder) == '.'; | |
| 416 | |
| 417 /* Skip the decimal point. */ | |
| 418 if (*has_decimal) | |
| 419 remainder++; | |
| 420 | |
| 421 *n_remainder = end - remainder; | |
| 422 } | |
| 423 | |
| 424 /* not all fields of format are used. for example, precision is | |
| 425 unused. should this take discrete params in order to be more clear | |
| 426 about what it does? or is passing a single format parameter easier | |
| 427 and more efficient enough to justify a little obfuscation? */ | |
| 428 static Py_ssize_t | |
| 429 calc_number_widths(NumberFieldWidths *spec, Py_ssize_t n_prefix, | |
| 430 Py_UCS4 sign_char, PyObject *number, Py_ssize_t n_start, | |
| 431 Py_ssize_t n_end, Py_ssize_t n_remainder, | |
| 432 int has_decimal, const LocaleInfo *locale, | |
| 433 const InternalFormatSpec *format) | |
| 434 { | |
| 435 Py_ssize_t n_non_digit_non_padding; | |
| 436 Py_ssize_t n_padding; | |
| 437 | |
| 438 spec->n_digits = n_end - n_start - n_remainder - (has_decimal?1:0); | |
| 439 spec->n_lpadding = 0; | |
| 440 spec->n_prefix = n_prefix; | |
| 441 spec->n_decimal = has_decimal ? strlen(locale->decimal_point) : 0; | |
| 442 spec->n_remainder = n_remainder; | |
| 443 spec->n_spadding = 0; | |
| 444 spec->n_rpadding = 0; | |
| 445 spec->sign = '\0'; | |
| 446 spec->n_sign = 0; | |
| 447 | |
| 448 /* the output will look like: | |
| 449 | | | |
| 450 | <lpadding> <sign> <prefix> <spadding> <grouped_digits> <decimal> <remainder> <rpadding> | | |
| 451 | | | |
| 452 | |
| 453 sign is computed from format->sign and the actual | |
| 454 sign of the number | |
| 455 | |
| 456 prefix is given (it's for the '0x' prefix) | |
| 457 | |
| 458 digits is already known | |
| 459 | |
| 460 the total width is either given, or computed from the | |
| 461 actual digits | |
| 462 | |
| 463 only one of lpadding, spadding, and rpadding can be non-zero, | |
| 464 and it's calculated from the width and other fields | |
| 465 */ | |
| 466 | |
| 467 /* compute the various parts we're going to write */ | |
| 468 switch (format->sign) { | |
| 469 case '+': | |
| 470 /* always put a + or - */ | |
| 471 spec->n_sign = 1; | |
| 472 spec->sign = (sign_char == '-' ? '-' : '+'); | |
| 473 break; | |
| 474 case ' ': | |
| 475 spec->n_sign = 1; | |
| 476 spec->sign = (sign_char == '-' ? '-' : ' '); | |
| 477 break; | |
| 478 default: | |
| 479 /* Not specified, or the default (-) */ | |
| 480 if (sign_char == '-') { | |
| 481 spec->n_sign = 1; | |
| 482 spec->sign = '-'; | |
| 483 } | |
| 484 } | |
| 485 | |
| 486 /* The number of chars used for non-digits and non-padding. */ | |
| 487 n_non_digit_non_padding = spec->n_sign + spec->n_prefix + spec->n_decimal + | |
| 488 spec->n_remainder; | |
| 489 | |
| 490 /* min_width can go negative, that's okay. format->width == -1 means | |
| 491 we don't care. */ | |
| 492 if (format->fill_char == '0' && format->align == '=') | |
| 493 spec->n_min_width = format->width - n_non_digit_non_padding; | |
| 494 else | |
| 495 spec->n_min_width = 0; | |
| 496 | |
| 497 if (spec->n_digits == 0) | |
| 498 /* This case only occurs when using 'c' formatting, we need | |
| 499 to special case it because the grouping code always wants | |
| 500 to have at least one character. */ | |
| 501 spec->n_grouped_digits = 0; | |
| 502 else | |
| 503 spec->n_grouped_digits = _PyUnicode_InsertThousandsGrouping( | |
| 504 PyUnicode_1BYTE_KIND, NULL, 0, NULL, | |
| 505 spec->n_digits, spec->n_min_width, | |
| 506 locale->grouping, locale->thousands_sep); | |
| 507 | |
| 508 /* Given the desired width and the total of digit and non-digit | |
| 509 space we consume, see if we need any padding. format->width can | |
| 510 be negative (meaning no padding), but this code still works in | |
| 511 that case. */ | |
| 512 n_padding = format->width - | |
| 513 (n_non_digit_non_padding + spec->n_grouped_digits); | |
| 514 if (n_padding > 0) { | |
| 515 /* Some padding is needed. Determine if it's left, space, or right. */ | |
| 516 switch (format->align) { | |
| 517 case '<': | |
| 518 spec->n_rpadding = n_padding; | |
| 519 break; | |
| 520 case '^': | |
| 521 spec->n_lpadding = n_padding / 2; | |
| 522 spec->n_rpadding = n_padding - spec->n_lpadding; | |
| 523 break; | |
| 524 case '=': | |
| 525 spec->n_spadding = n_padding; | |
| 526 break; | |
| 527 case '>': | |
| 528 spec->n_lpadding = n_padding; | |
| 529 break; | |
| 530 default: | |
| 531 /* Shouldn't get here, but treat it as '>' */ | |
| 532 spec->n_lpadding = n_padding; | |
| 533 assert(0); | |
| 534 break; | |
| 535 } | |
| 536 } | |
| 537 return spec->n_lpadding + spec->n_sign + spec->n_prefix + | |
| 538 spec->n_spadding + spec->n_grouped_digits + spec->n_decimal + | |
| 539 spec->n_remainder + spec->n_rpadding; | |
| 540 } | |
| 541 | |
| 542 /* Fill in the digit parts of a numbers's string representation, | |
| 543 as determined in calc_number_widths(). | |
| 544 No error checking, since we know the buffer is the correct size. */ | |
| 545 static void | |
| 546 fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, | |
| 547 PyObject *digits, Py_ssize_t d_start, Py_ssize_t d_end, | |
| 548 PyObject *prefix, Py_ssize_t p_start, Py_UCS4 fill_char, | |
| 549 LocaleInfo *locale, int toupper) | |
| 550 { | |
| 551 /* Used to keep track of digits, decimal, and remainder. */ | |
| 552 Py_ssize_t d_pos = d_start; | |
| 553 unsigned int kind = PyUnicode_KIND(out); | |
| 554 void *data = PyUnicode_DATA(out); | |
| 555 | |
| 556 #ifndef NDEBUG | |
| 557 Py_ssize_t r; | |
| 558 #endif | |
| 559 | |
| 560 if (spec->n_lpadding) { | |
| 561 unicode_fill(out, pos, pos + spec->n_lpadding, fill_char); | |
| 562 pos += spec->n_lpadding; | |
| 563 } | |
| 564 if (spec->n_sign == 1) { | |
| 565 PyUnicode_WRITE(kind, data, pos++, spec->sign); | |
| 566 } | |
| 567 if (spec->n_prefix) { | |
| 568 PyUnicode_CopyCharacters(out, pos, prefix, p_start, spec->n_prefix); | |
| 569 if (toupper) { | |
| 570 Py_ssize_t t; | |
| 571 /* XXX if the upper-case prefix is wider than the target | |
| 572 buffer, the caller should have allocated a wider string, | |
| 573 but currently doesn't. */ | |
| 574 for (t = 0; t < spec->n_prefix; ++t) | |
| 575 PyUnicode_WRITE(kind, data, pos + t, | |
| 576 Py_UNICODE_TOUPPER( | |
| 577 PyUnicode_READ(kind, data, pos + t))); | |
| 578 } | |
| 579 pos += spec->n_prefix; | |
| 580 } | |
| 581 if (spec->n_spadding) { | |
| 582 unicode_fill(out, pos, pos + spec->n_spadding, fill_char); | |
| 583 pos += spec->n_spadding; | |
| 584 } | |
| 585 | |
| 586 /* Only for type 'c' special case, it has no digits. */ | |
| 587 if (spec->n_digits != 0) { | |
| 588 /* Fill the digits with InsertThousandsGrouping. */ | |
| 589 char *pdigits = PyUnicode_DATA(digits); | |
| 590 if (PyUnicode_KIND(digits) < kind) { | |
| 591 pdigits = _PyUnicode_AsKind(digits, kind); | |
| 592 if (pdigits == NULL) { | |
| 593 /* XXX report exception */ | |
| 594 Py_FatalError("out of memory"); | |
| 595 return; | |
| 596 } | |
| 597 } | |
| 598 #ifndef NDEBUG | |
| 599 r = | |
| 600 #endif | |
| 601 _PyUnicode_InsertThousandsGrouping( | |
| 602 kind, | |
| 603 (char*)data + PyUnicode_KIND_SIZE(kind, pos), | |
| 604 spec->n_grouped_digits, | |
| 605 pdigits + PyUnicode_KIND_SIZE(kind, d_pos), | |
| 606 spec->n_digits, spec->n_min_width, | |
| 607 locale->grouping, locale->thousands_sep); | |
| 608 #ifndef NDEBUG | |
| 609 assert(r == spec->n_grouped_digits); | |
| 610 #endif | |
| 611 if (PyUnicode_KIND(digits) < kind) | |
| 612 PyMem_Free(pdigits); | |
| 613 d_pos += spec->n_digits; | |
| 614 } | |
| 615 if (toupper) { | |
| 616 Py_ssize_t t; | |
| 617 for (t = 0; t < spec->n_grouped_digits; ++t) | |
| 618 PyUnicode_WRITE(kind, data, pos + t, | |
| 619 Py_UNICODE_TOUPPER( | |
| 620 PyUnicode_READ(kind, data, pos + t))); | |
| 621 } | |
| 622 pos += spec->n_grouped_digits; | |
| 623 | |
| 624 if (spec->n_decimal) { | |
| 625 Py_ssize_t t; | |
| 626 for (t = 0; t < spec->n_decimal; ++t) | |
| 627 PyUnicode_WRITE(kind, data, pos + t, | |
| 628 locale->decimal_point[t]); | |
| 629 pos += spec->n_decimal; | |
| 630 d_pos += 1; | |
| 631 } | |
| 632 | |
| 633 if (spec->n_remainder) { | |
| 634 PyUnicode_CopyCharacters(out, pos, digits, d_pos, spec->n_remainder); | |
| 635 pos += spec->n_remainder; | |
| 636 d_pos += spec->n_remainder; | |
| 637 } | |
| 638 | |
| 639 if (spec->n_rpadding) { | |
| 640 unicode_fill(out, pos, pos + spec->n_rpadding, fill_char); | |
| 641 pos += spec->n_rpadding; | |
| 642 } | |
| 643 } | |
| 644 | |
| 645 static char no_grouping[1] = {CHAR_MAX}; | |
| 646 | |
| 647 /* Find the decimal point character(s?), thousands_separator(s?), and | |
| 648 grouping description, either for the current locale if type is | |
| 649 LT_CURRENT_LOCALE, a hard-coded locale if LT_DEFAULT_LOCALE, or | |
| 650 none if LT_NO_LOCALE. */ | |
| 651 static void | |
| 652 get_locale_info(int type, LocaleInfo *locale_info) | |
| 653 { | |
| 654 switch (type) { | |
| 655 case LT_CURRENT_LOCALE: { | |
| 656 struct lconv *locale_data = localeconv(); | |
| 657 locale_info->decimal_point = locale_data->decimal_point; | |
| 658 locale_info->thousands_sep = locale_data->thousands_sep; | |
| 659 locale_info->grouping = locale_data->grouping; | |
| 660 break; | |
| 661 } | |
| 662 case LT_DEFAULT_LOCALE: | |
| 663 locale_info->decimal_point = "."; | |
| 664 locale_info->thousands_sep = ","; | |
| 665 locale_info->grouping = "\3"; /* Group every 3 characters. The | |
| 666 (implicit) trailing 0 means repeat | |
| 667 infinitely. */ | |
| 668 break; | |
| 669 case LT_NO_LOCALE: | |
| 670 locale_info->decimal_point = "."; | |
| 671 locale_info->thousands_sep = ""; | |
| 672 locale_info->grouping = no_grouping; | |
| 673 break; | |
| 674 default: | |
| 675 assert(0); | |
| 676 } | |
| 677 } | |
| 678 | |
| 679 /************************************************************************/ | |
| 680 /*********** string formatting ******************************************/ | |
| 681 /************************************************************************/ | |
| 682 | |
| 683 static PyObject * | |
| 684 format_string_internal(PyObject *value, const InternalFormatSpec *format) | |
| 685 { | |
| 686 Py_ssize_t lpad; | |
| 687 Py_ssize_t rpad; | |
| 688 Py_ssize_t total; | |
| 689 Py_ssize_t pos; | |
| 690 Py_ssize_t len = PyUnicode_GET_SIZE(value); | |
| 691 PyObject *result = NULL; | |
| 692 int maxchar = 127; | |
| 693 | |
| 694 /* sign is not allowed on strings */ | |
| 695 if (format->sign != '\0') { | |
| 696 PyErr_SetString(PyExc_ValueError, | |
| 697 "Sign not allowed in string format specifier"); | |
| 698 goto done; | |
| 699 } | |
| 700 | |
| 701 /* alternate is not allowed on strings */ | |
| 702 if (format->alternate) { | |
| 703 PyErr_SetString(PyExc_ValueError, | |
| 704 "Alternate form (#) not allowed in string format " | |
| 705 "specifier"); | |
| 706 goto done; | |
| 707 } | |
| 708 | |
| 709 /* '=' alignment not allowed on strings */ | |
| 710 if (format->align == '=') { | |
| 711 PyErr_SetString(PyExc_ValueError, | |
| 712 "'=' alignment not allowed " | |
| 713 "in string format specifier"); | |
| 714 goto done; | |
| 715 } | |
| 716 | |
| 717 /* if precision is specified, output no more that format.precision | |
| 718 characters */ | |
| 719 if (format->precision >= 0 && len >= format->precision) { | |
| 720 len = format->precision; | |
| 721 } | |
| 722 | |
| 723 calc_padding(len, format->width, format->align, &lpad, &rpad, &total); | |
| 724 | |
| 725 /* allocate the resulting string */ | |
| 726 result = PyUnicode_New(total, maxchar); | |
| 727 if (result == NULL) | |
| 728 goto done; | |
| 729 | |
| 730 /* Write into that space. First the padding. */ | |
| 731 pos = fill_padding(result, 0, len, | |
| 732 format->fill_char=='\0'?' ':format->fill_char, | |
| 733 lpad, rpad); | |
| 734 | |
| 735 /* Then the source string. */ | |
| 736 PyUnicode_CopyCharacters(result, pos, value, 0, len); | |
| 737 | |
| 738 done: | |
| 739 return result; | |
| 740 } | |
| 741 | |
| 742 | |
| 743 /************************************************************************/ | |
| 744 /*********** long formatting ********************************************/ | |
| 745 /************************************************************************/ | |
| 746 | |
| 747 typedef PyObject* | |
| 748 (*IntOrLongToString)(PyObject *value, int base); | |
| 749 | |
| 750 static PyObject * | |
| 751 format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format, | |
| 752 IntOrLongToString tostring) | |
| 753 { | |
| 754 PyObject *result = NULL; | |
| 755 int maxchar = 127; | |
| 756 PyObject *tmp = NULL; | |
| 757 Py_ssize_t inumeric_chars; | |
| 758 Py_UCS4 sign_char = '\0'; | |
| 759 Py_ssize_t n_digits; /* count of digits need from the computed | |
| 760 string */ | |
| 761 Py_ssize_t n_remainder = 0; /* Used only for 'c' formatting, which | |
| 762 produces non-digits */ | |
| 763 Py_ssize_t n_prefix = 0; /* Count of prefix chars, (e.g., '0x') */ | |
| 764 Py_ssize_t n_total; | |
| 765 Py_ssize_t prefix; | |
| 766 NumberFieldWidths spec; | |
| 767 long x; | |
| 768 | |
| 769 /* Locale settings, either from the actual locale or | |
| 770 from a hard-code pseudo-locale */ | |
| 771 LocaleInfo locale; | |
| 772 | |
| 773 /* no precision allowed on integers */ | |
| 774 if (format->precision != -1) { | |
| 775 PyErr_SetString(PyExc_ValueError, | |
| 776 "Precision not allowed in integer format specifier"); | |
| 777 goto done; | |
| 778 } | |
| 779 | |
| 780 /* special case for character formatting */ | |
| 781 if (format->type == 'c') { | |
| 782 /* error to specify a sign */ | |
| 783 if (format->sign != '\0') { | |
| 784 PyErr_SetString(PyExc_ValueError, | |
| 785 "Sign not allowed with integer" | |
| 786 " format specifier 'c'"); | |
| 787 goto done; | |
| 788 } | |
| 789 | |
| 790 /* taken from unicodeobject.c formatchar() */ | |
| 791 /* Integer input truncated to a character */ | |
| 792 /* XXX: won't work for int */ | |
| 793 x = PyLong_AsLong(value); | |
| 794 if (x == -1 && PyErr_Occurred()) | |
| 795 goto done; | |
| 796 if (x < 0 || x > 0x10ffff) { | |
| 797 PyErr_SetString(PyExc_OverflowError, | |
| 798 "%c arg not in range(0x110000) " | |
| 799 "(wide Python build)"); | |
| 800 goto done; | |
| 801 } | |
| 802 tmp = PyUnicode_FromOrdinal(x); | |
| 803 inumeric_chars = 0; | |
| 804 n_digits = 1; | |
| 805 if (x > maxchar) | |
| 806 maxchar = x; | |
| 807 | |
| 808 /* As a sort-of hack, we tell calc_number_widths that we only | |
| 809 have "remainder" characters. calc_number_widths thinks | |
| 810 these are characters that don't get formatted, only copied | |
| 811 into the output string. We do this for 'c' formatting, | |
| 812 because the characters are likely to be non-digits. */ | |
| 813 n_remainder = 1; | |
| 814 } | |
| 815 else { | |
| 816 int base; | |
| 817 int leading_chars_to_skip = 0; /* Number of characters added by | |
| 818 PyNumber_ToBase that we want to | |
| 819 skip over. */ | |
| 820 | |
| 821 /* Compute the base and how many characters will be added by | |
| 822 PyNumber_ToBase */ | |
| 823 switch (format->type) { | |
| 824 case 'b': | |
| 825 base = 2; | |
| 826 leading_chars_to_skip = 2; /* 0b */ | |
| 827 break; | |
| 828 case 'o': | |
| 829 base = 8; | |
| 830 leading_chars_to_skip = 2; /* 0o */ | |
| 831 break; | |
| 832 case 'x': | |
| 833 case 'X': | |
| 834 base = 16; | |
| 835 leading_chars_to_skip = 2; /* 0x */ | |
| 836 break; | |
| 837 default: /* shouldn't be needed, but stops a compiler warning */ | |
| 838 case 'd': | |
| 839 case 'n': | |
| 840 base = 10; | |
| 841 break; | |
| 842 } | |
| 843 | |
| 844 /* The number of prefix chars is the same as the leading | |
| 845 chars to skip */ | |
| 846 if (format->alternate) | |
| 847 n_prefix = leading_chars_to_skip; | |
| 848 | |
| 849 /* Do the hard part, converting to a string in a given base */ | |
| 850 tmp = tostring(value, base); | |
| 851 if (tmp == NULL || PyUnicode_READY(tmp) == -1) | |
| 852 goto done; | |
| 853 | |
| 854 inumeric_chars = 0; | |
| 855 n_digits = PyUnicode_GET_LENGTH(tmp); | |
| 856 | |
| 857 prefix = inumeric_chars; | |
| 858 | |
| 859 /* Is a sign character present in the output? If so, remember it | |
| 860 and skip it */ | |
| 861 if (PyUnicode_READ_CHAR(tmp, inumeric_chars) == '-') { | |
| 862 sign_char = '-'; | |
| 863 ++prefix; | |
| 864 ++leading_chars_to_skip; | |
| 865 } | |
| 866 | |
| 867 /* Skip over the leading chars (0x, 0b, etc.) */ | |
| 868 n_digits -= leading_chars_to_skip; | |
| 869 inumeric_chars += leading_chars_to_skip; | |
| 870 } | |
| 871 | |
| 872 /* Determine the grouping, separator, and decimal point, if any. */ | |
| 873 get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : | |
| 874 (format->thousands_separators ? | |
| 875 LT_DEFAULT_LOCALE : | |
| 876 LT_NO_LOCALE), | |
| 877 &locale); | |
| 878 | |
| 879 /* Calculate how much memory we'll need. */ | |
| 880 n_total = calc_number_widths(&spec, n_prefix, sign_char, tmp, inumeric_chars, | |
| 881 inumeric_chars + n_digits, n_remainder, 0, &locale, format); | |
| 882 | |
| 883 /* Allocate the memory. */ | |
| 884 result = PyUnicode_New(n_total, maxchar); | |
| 885 if (!result) | |
| 886 goto done; | |
| 887 | |
| 888 /* Populate the memory. */ | |
| 889 fill_number(result, 0, &spec, tmp, inumeric_chars, inumeric_chars + n_digits, | |
| 890 tmp, prefix, | |
| 891 format->fill_char == '\0' ? ' ' : format->fill_char, | |
| 892 &locale, format->type == 'X'); | |
| 893 | |
| 894 done: | |
| 895 Py_XDECREF(tmp); | |
| 896 return result; | |
| 897 } | |
| 898 | |
| 899 /************************************************************************/ | |
| 900 /*********** float formatting *******************************************/ | |
| 901 /************************************************************************/ | |
| 902 | |
| 903 static PyObject* | |
| 904 strtounicode(char *charbuffer, Py_ssize_t len) | |
| 905 { | |
| 906 return PyUnicode_FromKindAndData(PyUnicode_1BYTE_KIND, charbuffer, len); | |
| 907 } | |
|
47393
0da27fdf0192
Refactor and clean up str.format() code (and helpers) in advance of optimizations.
Eric Smith <eric@trueblade.com>
parents:
42817
diff
changeset
|
908 |
| 72475 | 909 /* much of this is taken from unicodeobject.c */ |
| 910 static PyObject * | |
| 911 format_float_internal(PyObject *value, | |
| 912 const InternalFormatSpec *format) | |
| 913 { | |
| 914 char *buf = NULL; /* buffer returned from PyOS_double_to_string */ | |
| 915 Py_ssize_t n_digits; | |
| 916 Py_ssize_t n_remainder; | |
| 917 Py_ssize_t n_total; | |
| 918 int has_decimal; | |
| 919 double val; | |
| 920 Py_ssize_t precision = format->precision; | |
| 921 Py_ssize_t default_precision = 6; | |
| 922 Py_UCS4 type = format->type; | |
| 923 int add_pct = 0; | |
| 924 Py_ssize_t index; | |
| 925 NumberFieldWidths spec; | |
| 926 int flags = 0; | |
| 927 PyObject *result = NULL; | |
| 928 int maxchar = 127; | |
| 929 Py_UCS4 sign_char = '\0'; | |
| 930 int float_type; /* Used to see if we have a nan, inf, or regular float. */ | |
| 931 PyObject *unicode_tmp = NULL; | |
| 932 | |
| 933 /* Locale settings, either from the actual locale or | |
| 934 from a hard-code pseudo-locale */ | |
| 935 LocaleInfo locale; | |
| 936 | |
| 937 if (format->alternate) | |
| 938 flags |= Py_DTSF_ALT; | |
| 939 | |
| 940 if (type == '\0') { | |
| 941 /* Omitted type specifier. Behaves in the same way as repr(x) | |
| 942 and str(x) if no precision is given, else like 'g', but with | |
| 943 at least one digit after the decimal point. */ | |
| 944 flags |= Py_DTSF_ADD_DOT_0; | |
| 945 type = 'r'; | |
| 946 default_precision = 0; | |
| 947 } | |
| 948 | |
| 949 if (type == 'n') | |
| 950 /* 'n' is the same as 'g', except for the locale used to | |
| 951 format the result. We take care of that later. */ | |
| 952 type = 'g'; | |
| 953 | |
| 954 val = PyFloat_AsDouble(value); | |
| 955 if (val == -1.0 && PyErr_Occurred()) | |
| 956 goto done; | |
| 957 | |
| 958 if (type == '%') { | |
| 959 type = 'f'; | |
| 960 val *= 100; | |
| 961 add_pct = 1; | |
| 962 } | |
| 963 | |
| 964 if (precision < 0) | |
| 965 precision = default_precision; | |
| 966 else if (type == 'r') | |
| 967 type = 'g'; | |
| 968 | |
| 969 /* Cast "type", because if we're in unicode we need to pass a | |
| 970 8-bit char. This is safe, because we've restricted what "type" | |
| 971 can be. */ | |
| 972 buf = PyOS_double_to_string(val, (char)type, precision, flags, | |
| 973 &float_type); | |
| 974 if (buf == NULL) | |
| 975 goto done; | |
| 976 n_digits = strlen(buf); | |
| 977 | |
| 978 if (add_pct) { | |
| 979 /* We know that buf has a trailing zero (since we just called | |
| 980 strlen() on it), and we don't use that fact any more. So we | |
| 981 can just write over the trailing zero. */ | |
| 982 buf[n_digits] = '%'; | |
| 983 n_digits += 1; | |
| 984 } | |
| 985 | |
| 986 /* Since there is no unicode version of PyOS_double_to_string, | |
| 987 just use the 8 bit version and then convert to unicode. */ | |
| 988 unicode_tmp = strtounicode(buf, n_digits); | |
| 989 if (unicode_tmp == NULL) | |
| 990 goto done; | |
| 991 index = 0; | |
| 992 | |
| 993 /* Is a sign character present in the output? If so, remember it | |
| 994 and skip it */ | |
| 995 if (PyUnicode_READ_CHAR(unicode_tmp, index) == '-') { | |
| 996 sign_char = '-'; | |
| 997 ++index; | |
| 998 --n_digits; | |
| 999 } | |
| 1000 | |
| 1001 /* Determine if we have any "remainder" (after the digits, might include | |
| 1002 decimal or exponent or both (or neither)) */ | |
| 1003 parse_number(unicode_tmp, index, index + n_digits, &n_remainder, &has_decimal); | |
| 1004 | |
| 1005 /* Determine the grouping, separator, and decimal point, if any. */ | |
| 1006 get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : | |
| 1007 (format->thousands_separators ? | |
| 1008 LT_DEFAULT_LOCALE : | |
| 1009 LT_NO_LOCALE), | |
| 1010 &locale); | |
| 1011 | |
| 1012 /* Calculate how much memory we'll need. */ | |
| 1013 n_total = calc_number_widths(&spec, 0, sign_char, unicode_tmp, index, | |
| 1014 index + n_digits, n_remainder, has_decimal, | |
| 1015 &locale, format); | |
| 1016 | |
| 1017 /* Allocate the memory. */ | |
| 1018 result = PyUnicode_New(n_total, maxchar); | |
| 1019 if (result == NULL) | |
| 1020 goto done; | |
| 1021 | |
| 1022 /* Populate the memory. */ | |
| 1023 fill_number(result, 0, &spec, unicode_tmp, index, index + n_digits, | |
| 1024 NULL, 0, | |
| 1025 format->fill_char == '\0' ? ' ' : format->fill_char, &locale, | |
| 1026 0); | |
| 1027 | |
| 1028 done: | |
| 1029 PyMem_Free(buf); | |
| 1030 Py_DECREF(unicode_tmp); | |
| 1031 return result; | |
| 1032 } | |
| 1033 | |
| 1034 /************************************************************************/ | |
| 1035 /*********** complex formatting *****************************************/ | |
| 1036 /************************************************************************/ | |
| 1037 | |
| 1038 static PyObject * | |
| 1039 format_complex_internal(PyObject *value, | |
| 1040 const InternalFormatSpec *format) | |
| 1041 { | |
| 1042 double re; | |
| 1043 double im; | |
| 1044 char *re_buf = NULL; /* buffer returned from PyOS_double_to_string */ | |
| 1045 char *im_buf = NULL; /* buffer returned from PyOS_double_to_string */ | |
| 1046 | |
| 1047 InternalFormatSpec tmp_format = *format; | |
| 1048 Py_ssize_t n_re_digits; | |
| 1049 Py_ssize_t n_im_digits; | |
| 1050 Py_ssize_t n_re_remainder; | |
| 1051 Py_ssize_t n_im_remainder; | |
| 1052 Py_ssize_t n_re_total; | |
| 1053 Py_ssize_t n_im_total; | |
| 1054 int re_has_decimal; | |
| 1055 int im_has_decimal; | |
| 1056 Py_ssize_t precision = format->precision; | |
| 1057 Py_ssize_t default_precision = 6; | |
| 1058 Py_UCS4 type = format->type; | |
| 1059 Py_ssize_t i_re; | |
| 1060 Py_ssize_t i_im; | |
| 1061 NumberFieldWidths re_spec; | |
| 1062 NumberFieldWidths im_spec; | |
| 1063 int flags = 0; | |
| 1064 PyObject *result = NULL; | |
| 1065 int maxchar = 127; | |
| 1066 int rkind; | |
| 1067 void *rdata; | |
| 1068 Py_ssize_t index; | |
| 1069 Py_UCS4 re_sign_char = '\0'; | |
| 1070 Py_UCS4 im_sign_char = '\0'; | |
| 1071 int re_float_type; /* Used to see if we have a nan, inf, or regular float. */ | |
| 1072 int im_float_type; | |
| 1073 int add_parens = 0; | |
| 1074 int skip_re = 0; | |
| 1075 Py_ssize_t lpad; | |
| 1076 Py_ssize_t rpad; | |
| 1077 Py_ssize_t total; | |
| 1078 PyObject *re_unicode_tmp = NULL; | |
| 1079 PyObject *im_unicode_tmp = NULL; | |
| 1080 | |
| 1081 /* Locale settings, either from the actual locale or | |
| 1082 from a hard-code pseudo-locale */ | |
| 1083 LocaleInfo locale; | |
| 1084 | |
| 1085 /* Zero padding is not allowed. */ | |
| 1086 if (format->fill_char == '0') { | |
| 1087 PyErr_SetString(PyExc_ValueError, | |
| 1088 "Zero padding is not allowed in complex format " | |
| 1089 "specifier"); | |
| 1090 goto done; | |
| 1091 } | |
| 1092 | |
| 1093 /* Neither is '=' alignment . */ | |
| 1094 if (format->align == '=') { | |
| 1095 PyErr_SetString(PyExc_ValueError, | |
| 1096 "'=' alignment flag is not allowed in complex format " | |
| 1097 "specifier"); | |
| 1098 goto done; | |
| 1099 } | |
| 1100 | |
| 1101 re = PyComplex_RealAsDouble(value); | |
| 1102 if (re == -1.0 && PyErr_Occurred()) | |
| 1103 goto done; | |
| 1104 im = PyComplex_ImagAsDouble(value); | |
| 1105 if (im == -1.0 && PyErr_Occurred()) | |
| 1106 goto done; | |
| 1107 | |
| 1108 if (format->alternate) | |
| 1109 flags |= Py_DTSF_ALT; | |
| 1110 | |
| 1111 if (type == '\0') { | |
| 1112 /* Omitted type specifier. Should be like str(self). */ | |
| 1113 type = 'r'; | |
| 1114 default_precision = 0; | |
| 1115 if (re == 0.0 && copysign(1.0, re) == 1.0) | |
| 1116 skip_re = 1; | |
| 1117 else | |
| 1118 add_parens = 1; | |
| 1119 } | |
| 1120 | |
| 1121 if (type == 'n') | |
| 1122 /* 'n' is the same as 'g', except for the locale used to | |
| 1123 format the result. We take care of that later. */ | |
| 1124 type = 'g'; | |
| 1125 | |
| 1126 if (precision < 0) | |
| 1127 precision = default_precision; | |
| 1128 else if (type == 'r') | |
| 1129 type = 'g'; | |
| 1130 | |
| 1131 /* Cast "type", because if we're in unicode we need to pass a | |
| 1132 8-bit char. This is safe, because we've restricted what "type" | |
| 1133 can be. */ | |
| 1134 re_buf = PyOS_double_to_string(re, (char)type, precision, flags, | |
| 1135 &re_float_type); | |
| 1136 if (re_buf == NULL) | |
| 1137 goto done; | |
| 1138 im_buf = PyOS_double_to_string(im, (char)type, precision, flags, | |
| 1139 &im_float_type); | |
| 1140 if (im_buf == NULL) | |
| 1141 goto done; | |
| 1142 | |
| 1143 n_re_digits = strlen(re_buf); | |
| 1144 n_im_digits = strlen(im_buf); | |
| 1145 | |
| 1146 /* Since there is no unicode version of PyOS_double_to_string, | |
| 1147 just use the 8 bit version and then convert to unicode. */ | |
| 1148 re_unicode_tmp = strtounicode(re_buf, n_re_digits); | |
| 1149 if (re_unicode_tmp == NULL) | |
| 1150 goto done; | |
| 1151 i_re = 0; | |
| 1152 | |
| 1153 im_unicode_tmp = strtounicode(im_buf, n_im_digits); | |
| 1154 if (im_unicode_tmp == NULL) | |
| 1155 goto done; | |
| 1156 i_im = 0; | |
| 1157 | |
| 1158 /* Is a sign character present in the output? If so, remember it | |
| 1159 and skip it */ | |
| 1160 if (PyUnicode_READ_CHAR(re_unicode_tmp, i_re) == '-') { | |
| 1161 re_sign_char = '-'; | |
| 1162 ++i_re; | |
| 1163 --n_re_digits; | |
| 1164 } | |
| 1165 if (PyUnicode_READ_CHAR(im_unicode_tmp, i_im) == '-') { | |
| 1166 im_sign_char = '-'; | |
| 1167 ++i_im; | |
| 1168 --n_im_digits; | |
| 1169 } | |
| 1170 | |
| 1171 /* Determine if we have any "remainder" (after the digits, might include | |
| 1172 decimal or exponent or both (or neither)) */ | |
| 1173 parse_number(re_unicode_tmp, i_re, i_re + n_re_digits, | |
| 1174 &n_re_remainder, &re_has_decimal); | |
| 1175 parse_number(im_unicode_tmp, i_im, i_im + n_im_digits, | |
| 1176 &n_im_remainder, &im_has_decimal); | |
| 1177 | |
| 1178 /* Determine the grouping, separator, and decimal point, if any. */ | |
| 1179 get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : | |
| 1180 (format->thousands_separators ? | |
| 1181 LT_DEFAULT_LOCALE : | |
| 1182 LT_NO_LOCALE), | |
| 1183 &locale); | |
| 1184 | |
| 1185 /* Turn off any padding. We'll do it later after we've composed | |
| 1186 the numbers without padding. */ | |
| 1187 tmp_format.fill_char = '\0'; | |
| 1188 tmp_format.align = '<'; | |
| 1189 tmp_format.width = -1; | |
| 1190 | |
| 1191 /* Calculate how much memory we'll need. */ | |
| 1192 n_re_total = calc_number_widths(&re_spec, 0, re_sign_char, re_unicode_tmp, | |
| 1193 i_re, i_re + n_re_digits, n_re_remainder, | |
| 1194 re_has_decimal, &locale, &tmp_format); | |
| 1195 | |
| 1196 /* Same formatting, but always include a sign, unless the real part is | |
| 1197 * going to be omitted, in which case we use whatever sign convention was | |
| 1198 * requested by the original format. */ | |
| 1199 if (!skip_re) | |
| 1200 tmp_format.sign = '+'; | |
| 1201 n_im_total = calc_number_widths(&im_spec, 0, im_sign_char, im_unicode_tmp, | |
| 1202 i_im, i_im + n_im_digits, n_im_remainder, | |
| 1203 im_has_decimal, &locale, &tmp_format); | |
| 1204 | |
| 1205 if (skip_re) | |
| 1206 n_re_total = 0; | |
| 1207 | |
| 1208 /* Add 1 for the 'j', and optionally 2 for parens. */ | |
| 1209 calc_padding(n_re_total + n_im_total + 1 + add_parens * 2, | |
| 1210 format->width, format->align, &lpad, &rpad, &total); | |
| 1211 | |
| 1212 result = PyUnicode_New(total, maxchar); | |
| 1213 if (result == NULL) | |
| 1214 goto done; | |
| 1215 rkind = PyUnicode_KIND(result); | |
| 1216 rdata = PyUnicode_DATA(result); | |
| 1217 | |
| 1218 /* Populate the memory. First, the padding. */ | |
| 1219 index = fill_padding(result, 0, | |
| 1220 n_re_total + n_im_total + 1 + add_parens * 2, | |
| 1221 format->fill_char=='\0' ? ' ' : format->fill_char, | |
| 1222 lpad, rpad); | |
| 1223 | |
| 1224 if (add_parens) | |
| 1225 PyUnicode_WRITE(rkind, rdata, index++, '('); | |
| 1226 | |
| 1227 if (!skip_re) { | |
| 1228 fill_number(result, index, &re_spec, re_unicode_tmp, | |
| 1229 i_re, i_re + n_re_digits, NULL, 0, 0, &locale, 0); | |
| 1230 index += n_re_total; | |
| 1231 } | |
| 1232 fill_number(result, index, &im_spec, im_unicode_tmp, | |
| 1233 i_im, i_im + n_im_digits, NULL, 0, 0, &locale, 0); | |
| 1234 index += n_im_total; | |
| 1235 PyUnicode_WRITE(rkind, rdata, index++, 'j'); | |
| 1236 | |
| 1237 if (add_parens) | |
| 1238 PyUnicode_WRITE(rkind, rdata, index++, ')'); | |
| 1239 | |
| 1240 done: | |
| 1241 PyMem_Free(re_buf); | |
| 1242 PyMem_Free(im_buf); | |
| 1243 Py_XDECREF(re_unicode_tmp); | |
| 1244 Py_XDECREF(im_unicode_tmp); | |
| 1245 return result; | |
| 1246 } | |
| 1247 | |
| 1248 /************************************************************************/ | |
| 1249 /*********** built in formatters ****************************************/ | |
| 1250 /************************************************************************/ | |
| 1251 PyObject * | |
| 1252 _PyUnicode_FormatAdvanced(PyObject *obj, | |
| 1253 PyObject *format_spec, | |
| 1254 Py_ssize_t start, Py_ssize_t end) | |
| 1255 { | |
| 1256 InternalFormatSpec format; | |
| 1257 PyObject *result = NULL; | |
| 1258 | |
| 1259 /* check for the special case of zero length format spec, make | |
| 1260 it equivalent to str(obj) */ | |
| 1261 if (start == end) { | |
| 1262 result = PyObject_Str(obj); | |
| 1263 goto done; | |
| 1264 } | |
| 1265 | |
| 1266 /* parse the format_spec */ | |
| 1267 if (!parse_internal_render_format_spec(format_spec, start, end, | |
| 1268 &format, 's', '<')) | |
| 1269 goto done; | |
| 1270 | |
| 1271 /* type conversion? */ | |
| 1272 switch (format.type) { | |
| 1273 case 's': | |
| 1274 /* no type conversion needed, already a string. do the formatting */ | |
| 1275 result = format_string_internal(obj, &format); | |
| 1276 break; | |
| 1277 default: | |
| 1278 /* unknown */ | |
| 1279 unknown_presentation_type(format.type, obj->ob_type->tp_name); | |
| 1280 goto done; | |
| 1281 } | |
| 1282 | |
| 1283 done: | |
| 1284 return result; | |
| 1285 } | |
| 1286 | |
| 1287 static PyObject* | |
| 1288 format_int_or_long(PyObject* obj, PyObject* format_spec, | |
| 1289 Py_ssize_t start, Py_ssize_t end, | |
| 1290 IntOrLongToString tostring) | |
| 1291 { | |
| 1292 PyObject *result = NULL; | |
| 1293 PyObject *tmp = NULL; | |
| 1294 InternalFormatSpec format; | |
| 1295 | |
| 1296 /* check for the special case of zero length format spec, make | |
| 1297 it equivalent to str(obj) */ | |
| 1298 if (start == end) { | |
| 1299 result = PyObject_Str(obj); | |
| 1300 goto done; | |
| 1301 } | |
| 1302 | |
| 1303 /* parse the format_spec */ | |
| 1304 if (!parse_internal_render_format_spec(format_spec, start, end, | |
| 1305 &format, 'd', '>')) | |
| 1306 goto done; | |
| 1307 | |
| 1308 /* type conversion? */ | |
| 1309 switch (format.type) { | |
| 1310 case 'b': | |
| 1311 case 'c': | |
| 1312 case 'd': | |
| 1313 case 'o': | |
| 1314 case 'x': | |
| 1315 case 'X': | |
| 1316 case 'n': | |
| 1317 /* no type conversion needed, already an int (or long). do | |
| 1318 the formatting */ | |
| 1319 result = format_int_or_long_internal(obj, &format, tostring); | |
| 1320 break; | |
| 1321 | |
| 1322 case 'e': | |
| 1323 case 'E': | |
| 1324 case 'f': | |
| 1325 case 'F': | |
| 1326 case 'g': | |
| 1327 case 'G': | |
| 1328 case '%': | |
| 1329 /* convert to float */ | |
| 1330 tmp = PyNumber_Float(obj); | |
| 1331 if (tmp == NULL) | |
| 1332 goto done; | |
| 1333 result = format_float_internal(tmp, &format); | |
| 1334 break; | |
| 1335 | |
| 1336 default: | |
| 1337 /* unknown */ | |
| 1338 unknown_presentation_type(format.type, obj->ob_type->tp_name); | |
| 1339 goto done; | |
| 1340 } | |
| 1341 | |
| 1342 done: | |
| 1343 Py_XDECREF(tmp); | |
| 1344 return result; | |
| 1345 } | |
| 1346 | |
| 1347 /* Need to define long_format as a function that will convert a long | |
| 1348 to a string. In 3.0, _PyLong_Format has the correct signature. */ | |
| 1349 #define long_format _PyLong_Format | |
| 1350 | |
| 1351 PyObject * | |
| 1352 _PyLong_FormatAdvanced(PyObject *obj, | |
| 1353 PyObject *format_spec, | |
| 1354 Py_ssize_t start, Py_ssize_t end) | |
| 1355 { | |
| 1356 return format_int_or_long(obj, format_spec, start, end, | |
| 1357 long_format); | |
| 1358 } | |
| 1359 | |
| 1360 PyObject * | |
| 1361 _PyFloat_FormatAdvanced(PyObject *obj, | |
| 1362 PyObject *format_spec, | |
| 1363 Py_ssize_t start, Py_ssize_t end) | |
| 1364 { | |
| 1365 PyObject *result = NULL; | |
| 1366 InternalFormatSpec format; | |
| 1367 | |
| 1368 /* check for the special case of zero length format spec, make | |
| 1369 it equivalent to str(obj) */ | |
| 1370 if (start == end) { | |
| 1371 result = PyObject_Str(obj); | |
| 1372 goto done; | |
| 1373 } | |
| 1374 | |
| 1375 /* parse the format_spec */ | |
| 1376 if (!parse_internal_render_format_spec(format_spec, start, end, | |
| 1377 &format, '\0', '>')) | |
| 1378 goto done; | |
| 1379 | |
| 1380 /* type conversion? */ | |
| 1381 switch (format.type) { | |
| 1382 case '\0': /* No format code: like 'g', but with at least one decimal. */ | |
| 1383 case 'e': | |
| 1384 case 'E': | |
| 1385 case 'f': | |
| 1386 case 'F': | |
| 1387 case 'g': | |
| 1388 case 'G': | |
| 1389 case 'n': | |
| 1390 case '%': | |
| 1391 /* no conversion, already a float. do the formatting */ | |
| 1392 result = format_float_internal(obj, &format); | |
| 1393 break; | |
| 1394 | |
| 1395 default: | |
| 1396 /* unknown */ | |
| 1397 unknown_presentation_type(format.type, obj->ob_type->tp_name); | |
| 1398 goto done; | |
| 1399 } | |
| 1400 | |
| 1401 done: | |
| 1402 return result; | |
| 1403 } | |
| 1404 | |
| 1405 PyObject * | |
| 1406 _PyComplex_FormatAdvanced(PyObject *obj, | |
| 1407 PyObject *format_spec, | |
| 1408 Py_ssize_t start, Py_ssize_t end) | |
| 1409 { | |
| 1410 PyObject *result = NULL; | |
| 1411 InternalFormatSpec format; | |
| 1412 | |
| 1413 /* check for the special case of zero length format spec, make | |
| 1414 it equivalent to str(obj) */ | |
| 1415 if (start == end) { | |
| 1416 result = PyObject_Str(obj); | |
| 1417 goto done; | |
| 1418 } | |
| 1419 | |
| 1420 /* parse the format_spec */ | |
| 1421 if (!parse_internal_render_format_spec(format_spec, start, end, | |
| 1422 &format, '\0', '>')) | |
| 1423 goto done; | |
| 1424 | |
| 1425 /* type conversion? */ | |
| 1426 switch (format.type) { | |
| 1427 case '\0': /* No format code: like 'g', but with at least one decimal. */ | |
| 1428 case 'e': | |
| 1429 case 'E': | |
| 1430 case 'f': | |
| 1431 case 'F': | |
| 1432 case 'g': | |
| 1433 case 'G': | |
| 1434 case 'n': | |
| 1435 /* no conversion, already a complex. do the formatting */ | |
| 1436 result = format_complex_internal(obj, &format); | |
| 1437 break; | |
| 1438 | |
| 1439 default: | |
| 1440 /* unknown */ | |
| 1441 unknown_presentation_type(format.type, obj->ob_type->tp_name); | |
| 1442 goto done; | |
| 1443 } | |
| 1444 | |
| 1445 done: | |
| 1446 return result; | |
| 1447 } |