Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 1 | /* implements the string, long, and float formatters. that is, |
| 2 | string.__format__, etc. */ |
| 3 | |
| 4 | /* Before including this, you must include either: |
| 5 | stringlib/unicodedefs.h |
| 6 | stringlib/stringdefs.h |
| 7 | |
| 8 | Also, you should define the names: |
| 9 | FORMAT_STRING |
| 10 | FORMAT_LONG |
| 11 | FORMAT_FLOAT |
| 12 | to be whatever you want the public names of these functions to |
| 13 | be. These are the only non-static functions defined here. |
| 14 | */ |
| 15 | |
Eric Smith | b7f5ba1 | 2007-08-29 12:38:45 +0000 | [diff] [blame] | 16 | #define ALLOW_PARENS_FOR_SIGN 0 |
| 17 | |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 18 | /* |
| 19 | get_integer consumes 0 or more decimal digit characters from an |
| 20 | input string, updates *result with the corresponding positive |
| 21 | integer, and returns the number of digits consumed. |
| 22 | |
| 23 | returns -1 on error. |
| 24 | */ |
| 25 | static int |
| 26 | get_integer(STRINGLIB_CHAR **ptr, STRINGLIB_CHAR *end, |
| 27 | Py_ssize_t *result) |
| 28 | { |
| 29 | Py_ssize_t accumulator, digitval, oldaccumulator; |
| 30 | int numdigits; |
| 31 | accumulator = numdigits = 0; |
| 32 | for (;;(*ptr)++, numdigits++) { |
| 33 | if (*ptr >= end) |
| 34 | break; |
| 35 | digitval = STRINGLIB_TODECIMAL(**ptr); |
| 36 | if (digitval < 0) |
| 37 | break; |
| 38 | /* |
| 39 | This trick was copied from old Unicode format code. It's cute, |
| 40 | but would really suck on an old machine with a slow divide |
| 41 | implementation. Fortunately, in the normal case we do not |
| 42 | expect too many digits. |
| 43 | */ |
| 44 | oldaccumulator = accumulator; |
| 45 | accumulator *= 10; |
| 46 | if ((accumulator+10)/10 != oldaccumulator+1) { |
| 47 | PyErr_Format(PyExc_ValueError, |
| 48 | "Too many decimal digits in format string"); |
| 49 | return -1; |
| 50 | } |
| 51 | accumulator += digitval; |
| 52 | } |
| 53 | *result = accumulator; |
| 54 | return numdigits; |
| 55 | } |
| 56 | |
| 57 | /************************************************************************/ |
| 58 | /*********** standard format specifier parsing **************************/ |
| 59 | /************************************************************************/ |
| 60 | |
| 61 | /* returns true if this character is a specifier alignment token */ |
| 62 | Py_LOCAL_INLINE(int) |
| 63 | is_alignment_token(STRINGLIB_CHAR c) |
| 64 | { |
| 65 | switch (c) { |
| 66 | case '<': case '>': case '=': case '^': |
| 67 | return 1; |
| 68 | default: |
| 69 | return 0; |
| 70 | } |
| 71 | } |
| 72 | |
| 73 | /* returns true if this character is a sign element */ |
| 74 | Py_LOCAL_INLINE(int) |
| 75 | is_sign_element(STRINGLIB_CHAR c) |
| 76 | { |
| 77 | switch (c) { |
Eric Smith | b7f5ba1 | 2007-08-29 12:38:45 +0000 | [diff] [blame] | 78 | case ' ': case '+': case '-': |
Eric Smith | 4430095 | 2007-08-29 12:43:12 +0000 | [diff] [blame] | 79 | #if ALLOW_PARENS_FOR_SIGN |
Eric Smith | b7f5ba1 | 2007-08-29 12:38:45 +0000 | [diff] [blame] | 80 | case '(': |
Eric Smith | 4430095 | 2007-08-29 12:43:12 +0000 | [diff] [blame] | 81 | #endif |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 82 | return 1; |
| 83 | default: |
| 84 | return 0; |
| 85 | } |
| 86 | } |
| 87 | |
| 88 | |
| 89 | typedef struct { |
| 90 | STRINGLIB_CHAR fill_char; |
| 91 | STRINGLIB_CHAR align; |
| 92 | STRINGLIB_CHAR sign; |
| 93 | Py_ssize_t width; |
| 94 | Py_ssize_t precision; |
| 95 | STRINGLIB_CHAR type; |
| 96 | } InternalFormatSpec; |
| 97 | |
| 98 | /* |
| 99 | ptr points to the start of the format_spec, end points just past its end. |
| 100 | fills in format with the parsed information. |
| 101 | returns 1 on success, 0 on failure. |
| 102 | if failure, sets the exception |
| 103 | */ |
| 104 | static int |
| 105 | parse_internal_render_format_spec(PyObject *format_spec, |
| 106 | InternalFormatSpec *format, |
| 107 | char default_type) |
| 108 | { |
| 109 | STRINGLIB_CHAR *ptr = STRINGLIB_STR(format_spec); |
| 110 | STRINGLIB_CHAR *end = ptr + STRINGLIB_LEN(format_spec); |
| 111 | |
| 112 | /* end-ptr is used throughout this code to specify the length of |
| 113 | the input string */ |
| 114 | |
| 115 | Py_ssize_t specified_width; |
| 116 | |
| 117 | format->fill_char = '\0'; |
| 118 | format->align = '\0'; |
| 119 | format->sign = '\0'; |
| 120 | format->width = -1; |
| 121 | format->precision = -1; |
| 122 | format->type = default_type; |
| 123 | |
| 124 | /* If the second char is an alignment token, |
| 125 | then parse the fill char */ |
| 126 | if (end-ptr >= 2 && is_alignment_token(ptr[1])) { |
| 127 | format->align = ptr[1]; |
| 128 | format->fill_char = ptr[0]; |
| 129 | ptr += 2; |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 130 | } |
| 131 | else if (end-ptr >= 1 && is_alignment_token(ptr[0])) { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 132 | format->align = ptr[0]; |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 133 | ++ptr; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 134 | } |
| 135 | |
| 136 | /* Parse the various sign options */ |
| 137 | if (end-ptr >= 1 && is_sign_element(ptr[0])) { |
| 138 | format->sign = ptr[0]; |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 139 | ++ptr; |
Eric Smith | b7f5ba1 | 2007-08-29 12:38:45 +0000 | [diff] [blame] | 140 | #if ALLOW_PARENS_FOR_SIGN |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 141 | if (end-ptr >= 1 && ptr[0] == ')') { |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 142 | ++ptr; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 143 | } |
Eric Smith | b7f5ba1 | 2007-08-29 12:38:45 +0000 | [diff] [blame] | 144 | #endif |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 145 | } |
| 146 | |
| 147 | /* The special case for 0-padding (backwards compat) */ |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 148 | if (format->fill_char == '\0' && end-ptr >= 1 && ptr[0] == '0') { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 149 | format->fill_char = '0'; |
| 150 | if (format->align == '\0') { |
| 151 | format->align = '='; |
| 152 | } |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 153 | ++ptr; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 154 | } |
| 155 | |
| 156 | /* XXX add error checking */ |
| 157 | specified_width = get_integer(&ptr, end, &format->width); |
| 158 | |
| 159 | /* if specified_width is 0, we didn't consume any characters for |
| 160 | the width. in that case, reset the width to -1, because |
| 161 | get_integer() will have set it to zero */ |
| 162 | if (specified_width == 0) { |
| 163 | format->width = -1; |
| 164 | } |
| 165 | |
| 166 | /* Parse field precision */ |
| 167 | if (end-ptr && ptr[0] == '.') { |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 168 | ++ptr; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 169 | |
| 170 | /* XXX add error checking */ |
| 171 | specified_width = get_integer(&ptr, end, &format->precision); |
| 172 | |
| 173 | /* not having a precision after a dot is an error */ |
| 174 | if (specified_width == 0) { |
| 175 | PyErr_Format(PyExc_ValueError, |
| 176 | "Format specifier missing precision"); |
| 177 | return 0; |
| 178 | } |
| 179 | |
| 180 | } |
| 181 | |
| 182 | /* Finally, parse the type field */ |
| 183 | |
| 184 | if (end-ptr > 1) { |
| 185 | /* invalid conversion spec */ |
| 186 | PyErr_Format(PyExc_ValueError, "Invalid conversion specification"); |
| 187 | return 0; |
| 188 | } |
| 189 | |
| 190 | if (end-ptr == 1) { |
| 191 | format->type = ptr[0]; |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 192 | ++ptr; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 193 | } |
| 194 | |
| 195 | return 1; |
| 196 | } |
| 197 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 198 | #if defined FORMAT_FLOAT || defined FORMAT_LONG |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 199 | /************************************************************************/ |
| 200 | /*********** common routines for numeric formatting *********************/ |
| 201 | /************************************************************************/ |
| 202 | |
| 203 | /* describes the layout for an integer, see the comment in |
| 204 | _calc_integer_widths() for details */ |
| 205 | typedef struct { |
| 206 | Py_ssize_t n_lpadding; |
| 207 | Py_ssize_t n_spadding; |
| 208 | Py_ssize_t n_rpadding; |
| 209 | char lsign; |
| 210 | Py_ssize_t n_lsign; |
| 211 | char rsign; |
| 212 | Py_ssize_t n_rsign; |
| 213 | Py_ssize_t n_total; /* just a convenience, it's derivable from the |
| 214 | other fields */ |
| 215 | } NumberFieldWidths; |
| 216 | |
| 217 | /* not all fields of format are used. for example, precision is |
| 218 | unused. should this take discrete params in order to be more clear |
| 219 | about what it does? or is passing a single format parameter easier |
| 220 | and more efficient enough to justify a little obfuscation? */ |
| 221 | static void |
| 222 | calc_number_widths(NumberFieldWidths *r, STRINGLIB_CHAR actual_sign, |
| 223 | Py_ssize_t n_digits, const InternalFormatSpec *format) |
| 224 | { |
| 225 | r->n_lpadding = 0; |
| 226 | r->n_spadding = 0; |
| 227 | r->n_rpadding = 0; |
| 228 | r->lsign = '\0'; |
| 229 | r->n_lsign = 0; |
| 230 | r->rsign = '\0'; |
| 231 | r->n_rsign = 0; |
| 232 | |
| 233 | /* the output will look like: |
| 234 | | | |
| 235 | | <lpadding> <lsign> <spadding> <digits> <rsign> <rpadding> | |
| 236 | | | |
| 237 | |
| 238 | lsign and rsign are computed from format->sign and the actual |
| 239 | sign of the number |
| 240 | |
| 241 | digits is already known |
| 242 | |
| 243 | the total width is either given, or computed from the |
| 244 | actual digits |
| 245 | |
| 246 | only one of lpadding, spadding, and rpadding can be non-zero, |
| 247 | and it's calculated from the width and other fields |
| 248 | */ |
| 249 | |
| 250 | /* compute the various parts we're going to write */ |
| 251 | if (format->sign == '+') { |
| 252 | /* always put a + or - */ |
| 253 | r->n_lsign = 1; |
| 254 | r->lsign = (actual_sign == '-' ? '-' : '+'); |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 255 | } |
Eric Smith | b7f5ba1 | 2007-08-29 12:38:45 +0000 | [diff] [blame] | 256 | #if ALLOW_PARENS_FOR_SIGN |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 257 | else if (format->sign == '(') { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 258 | if (actual_sign == '-') { |
| 259 | r->n_lsign = 1; |
| 260 | r->lsign = '('; |
| 261 | r->n_rsign = 1; |
| 262 | r->rsign = ')'; |
| 263 | } |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 264 | } |
Eric Smith | b7f5ba1 | 2007-08-29 12:38:45 +0000 | [diff] [blame] | 265 | #endif |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 266 | else if (format->sign == ' ') { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 267 | r->n_lsign = 1; |
| 268 | r->lsign = (actual_sign == '-' ? '-' : ' '); |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 269 | } |
| 270 | else { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 271 | /* non specified, or the default (-) */ |
| 272 | if (actual_sign == '-') { |
| 273 | r->n_lsign = 1; |
| 274 | r->lsign = '-'; |
| 275 | } |
| 276 | } |
| 277 | |
| 278 | /* now the number of padding characters */ |
| 279 | if (format->width == -1) { |
| 280 | /* no padding at all, nothing to do */ |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 281 | } |
| 282 | else { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 283 | /* see if any padding is needed */ |
| 284 | if (r->n_lsign + n_digits + r->n_rsign >= format->width) { |
| 285 | /* no padding needed, we're already bigger than the |
| 286 | requested width */ |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 287 | } |
| 288 | else { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 289 | /* determine which of left, space, or right padding is |
| 290 | needed */ |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 291 | Py_ssize_t padding = format->width - |
| 292 | (r->n_lsign + n_digits + r->n_rsign); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 293 | if (format->align == '<') |
| 294 | r->n_rpadding = padding; |
| 295 | else if (format->align == '>') |
| 296 | r->n_lpadding = padding; |
| 297 | else if (format->align == '^') { |
| 298 | r->n_lpadding = padding / 2; |
| 299 | r->n_rpadding = padding - r->n_lpadding; |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 300 | } |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 301 | else if (format->align == '=') |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 302 | r->n_spadding = padding; |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 303 | else |
| 304 | r->n_lpadding = padding; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 305 | } |
| 306 | } |
| 307 | r->n_total = r->n_lpadding + r->n_lsign + r->n_spadding + |
| 308 | n_digits + r->n_rsign + r->n_rpadding; |
| 309 | } |
| 310 | |
| 311 | /* fill in the non-digit parts of a numbers's string representation, |
| 312 | as determined in _calc_integer_widths(). returns the pointer to |
| 313 | where the digits go. */ |
| 314 | static STRINGLIB_CHAR * |
| 315 | fill_number(STRINGLIB_CHAR *p_buf, const NumberFieldWidths *spec, |
| 316 | Py_ssize_t n_digits, STRINGLIB_CHAR fill_char) |
| 317 | { |
| 318 | STRINGLIB_CHAR* p_digits; |
| 319 | |
| 320 | if (spec->n_lpadding) { |
| 321 | STRINGLIB_FILL(p_buf, fill_char, spec->n_lpadding); |
| 322 | p_buf += spec->n_lpadding; |
| 323 | } |
| 324 | if (spec->n_lsign == 1) { |
| 325 | *p_buf++ = spec->lsign; |
| 326 | } |
| 327 | if (spec->n_spadding) { |
| 328 | STRINGLIB_FILL(p_buf, fill_char, spec->n_spadding); |
| 329 | p_buf += spec->n_spadding; |
| 330 | } |
| 331 | p_digits = p_buf; |
| 332 | p_buf += n_digits; |
| 333 | if (spec->n_rsign == 1) { |
| 334 | *p_buf++ = spec->rsign; |
| 335 | } |
| 336 | if (spec->n_rpadding) { |
| 337 | STRINGLIB_FILL(p_buf, fill_char, spec->n_rpadding); |
| 338 | p_buf += spec->n_rpadding; |
| 339 | } |
| 340 | return p_digits; |
| 341 | } |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 342 | #endif /* FORMAT_FLOAT || FORMAT_LONG */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 343 | |
| 344 | /************************************************************************/ |
| 345 | /*********** string formatting ******************************************/ |
| 346 | /************************************************************************/ |
| 347 | |
| 348 | static PyObject * |
| 349 | format_string_internal(PyObject *value, const InternalFormatSpec *format) |
| 350 | { |
| 351 | Py_ssize_t width; /* total field width */ |
| 352 | Py_ssize_t lpad; |
| 353 | STRINGLIB_CHAR *dst; |
| 354 | STRINGLIB_CHAR *src = STRINGLIB_STR(value); |
| 355 | Py_ssize_t len = STRINGLIB_LEN(value); |
| 356 | PyObject *result = NULL; |
| 357 | |
| 358 | /* sign is not allowed on strings */ |
| 359 | if (format->sign != '\0') { |
| 360 | PyErr_SetString(PyExc_ValueError, |
| 361 | "Sign not allowed in string format specifier"); |
| 362 | goto done; |
| 363 | } |
| 364 | |
| 365 | /* '=' alignment not allowed on strings */ |
| 366 | if (format->align == '=') { |
| 367 | PyErr_SetString(PyExc_ValueError, |
| 368 | "'=' alignment not allowed " |
| 369 | "in string format specifier"); |
| 370 | goto done; |
| 371 | } |
| 372 | |
| 373 | /* if precision is specified, output no more that format.precision |
| 374 | characters */ |
| 375 | if (format->precision >= 0 && len >= format->precision) { |
| 376 | len = format->precision; |
| 377 | } |
| 378 | |
| 379 | if (format->width >= 0) { |
| 380 | width = format->width; |
| 381 | |
| 382 | /* but use at least len characters */ |
| 383 | if (len > width) { |
| 384 | width = len; |
| 385 | } |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 386 | } |
| 387 | else { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 388 | /* not specified, use all of the chars and no more */ |
| 389 | width = len; |
| 390 | } |
| 391 | |
| 392 | /* allocate the resulting string */ |
| 393 | result = STRINGLIB_NEW(NULL, width); |
| 394 | if (result == NULL) |
| 395 | goto done; |
| 396 | |
| 397 | /* now write into that space */ |
| 398 | dst = STRINGLIB_STR(result); |
| 399 | |
| 400 | /* figure out how much leading space we need, based on the |
| 401 | aligning */ |
| 402 | if (format->align == '>') |
| 403 | lpad = width - len; |
| 404 | else if (format->align == '^') |
| 405 | lpad = (width - len) / 2; |
| 406 | else |
| 407 | lpad = 0; |
| 408 | |
| 409 | /* if right aligning, increment the destination allow space on the |
| 410 | left */ |
| 411 | memcpy(dst + lpad, src, len * sizeof(STRINGLIB_CHAR)); |
| 412 | |
| 413 | /* do any padding */ |
| 414 | if (width > len) { |
| 415 | STRINGLIB_CHAR fill_char = format->fill_char; |
| 416 | if (fill_char == '\0') { |
| 417 | /* use the default, if not specified */ |
| 418 | fill_char = ' '; |
| 419 | } |
| 420 | |
| 421 | /* pad on left */ |
| 422 | if (lpad) |
| 423 | STRINGLIB_FILL(dst, fill_char, lpad); |
| 424 | |
| 425 | /* pad on right */ |
| 426 | if (width - len - lpad) |
| 427 | STRINGLIB_FILL(dst + len + lpad, fill_char, width - len - lpad); |
| 428 | } |
| 429 | |
| 430 | done: |
| 431 | return result; |
| 432 | } |
| 433 | |
| 434 | |
| 435 | /************************************************************************/ |
| 436 | /*********** long formatting ********************************************/ |
| 437 | /************************************************************************/ |
| 438 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 439 | #if defined FORMAT_LONG || defined FORMAT_INT |
| 440 | typedef PyObject* |
| 441 | (*IntOrLongToString)(PyObject *value, int base); |
| 442 | |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 443 | static PyObject * |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 444 | format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format, |
| 445 | IntOrLongToString tostring) |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 446 | { |
| 447 | PyObject *result = NULL; |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 448 | PyObject *tmp = NULL; |
| 449 | STRINGLIB_CHAR *pnumeric_chars; |
| 450 | STRINGLIB_CHAR numeric_char; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 451 | STRINGLIB_CHAR sign = '\0'; |
| 452 | STRINGLIB_CHAR *p; |
| 453 | Py_ssize_t n_digits; /* count of digits need from the computed |
| 454 | string */ |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 455 | Py_ssize_t n_leading_chars; |
Eric Smith | 5807c41 | 2008-05-11 21:00:57 +0000 | [diff] [blame^] | 456 | Py_ssize_t n_grouping_chars = 0; /* Count of additional chars to |
| 457 | allocate, used for 'n' |
| 458 | formatting. */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 459 | NumberFieldWidths spec; |
| 460 | long x; |
| 461 | |
| 462 | /* no precision allowed on integers */ |
| 463 | if (format->precision != -1) { |
| 464 | PyErr_SetString(PyExc_ValueError, |
| 465 | "Precision not allowed in integer format specifier"); |
| 466 | goto done; |
| 467 | } |
| 468 | |
| 469 | |
| 470 | /* special case for character formatting */ |
| 471 | if (format->type == 'c') { |
| 472 | /* error to specify a sign */ |
| 473 | if (format->sign != '\0') { |
| 474 | PyErr_SetString(PyExc_ValueError, |
| 475 | "Sign not allowed with integer" |
| 476 | " format specifier 'c'"); |
| 477 | goto done; |
| 478 | } |
| 479 | |
| 480 | /* taken from unicodeobject.c formatchar() */ |
| 481 | /* Integer input truncated to a character */ |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 482 | /* XXX: won't work for int */ |
Christian Heimes | 217cfd1 | 2007-12-02 14:31:20 +0000 | [diff] [blame] | 483 | x = PyLong_AsLong(value); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 484 | if (x == -1 && PyErr_Occurred()) |
| 485 | goto done; |
| 486 | #ifdef Py_UNICODE_WIDE |
| 487 | if (x < 0 || x > 0x10ffff) { |
| 488 | PyErr_SetString(PyExc_OverflowError, |
| 489 | "%c arg not in range(0x110000) " |
| 490 | "(wide Python build)"); |
| 491 | goto done; |
| 492 | } |
| 493 | #else |
| 494 | if (x < 0 || x > 0xffff) { |
| 495 | PyErr_SetString(PyExc_OverflowError, |
| 496 | "%c arg not in range(0x10000) " |
| 497 | "(narrow Python build)"); |
| 498 | goto done; |
| 499 | } |
| 500 | #endif |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 501 | numeric_char = (STRINGLIB_CHAR)x; |
| 502 | pnumeric_chars = &numeric_char; |
| 503 | n_digits = 1; |
Eric Smith | 0cb431c | 2007-08-28 01:07:27 +0000 | [diff] [blame] | 504 | } |
| 505 | else { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 506 | int base; |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 507 | int leading_chars_to_skip; /* Number of characters added by |
| 508 | PyNumber_ToBase that we want to |
| 509 | skip over. */ |
| 510 | |
| 511 | /* Compute the base and how many characters will be added by |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 512 | PyNumber_ToBase */ |
| 513 | switch (format->type) { |
| 514 | case 'b': |
| 515 | base = 2; |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 516 | leading_chars_to_skip = 2; /* 0b */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 517 | break; |
| 518 | case 'o': |
| 519 | base = 8; |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 520 | leading_chars_to_skip = 2; /* 0o */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 521 | break; |
| 522 | case 'x': |
| 523 | case 'X': |
| 524 | base = 16; |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 525 | leading_chars_to_skip = 2; /* 0x */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 526 | break; |
| 527 | default: /* shouldn't be needed, but stops a compiler warning */ |
| 528 | case 'd': |
Eric Smith | 5807c41 | 2008-05-11 21:00:57 +0000 | [diff] [blame^] | 529 | case 'n': |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 530 | base = 10; |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 531 | leading_chars_to_skip = 0; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 532 | break; |
| 533 | } |
| 534 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 535 | /* Do the hard part, converting to a string in a given base */ |
| 536 | tmp = tostring(value, base); |
| 537 | if (tmp == NULL) |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 538 | goto done; |
| 539 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 540 | pnumeric_chars = STRINGLIB_STR(tmp); |
| 541 | n_digits = STRINGLIB_LEN(tmp); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 542 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 543 | /* Remember not to modify what pnumeric_chars points to. it |
| 544 | might be interned. Only modify it after we copy it into a |
| 545 | newly allocated output buffer. */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 546 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 547 | /* Is a sign character present in the output? If so, remember it |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 548 | and skip it */ |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 549 | sign = pnumeric_chars[0]; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 550 | if (sign == '-') { |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 551 | ++leading_chars_to_skip; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 552 | } |
| 553 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 554 | /* Skip over the leading chars (0x, 0b, etc.) */ |
| 555 | n_digits -= leading_chars_to_skip; |
| 556 | pnumeric_chars += leading_chars_to_skip; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 557 | } |
| 558 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 559 | /* Calculate the widths of the various leading and trailing parts */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 560 | calc_number_widths(&spec, sign, n_digits, format); |
| 561 | |
Eric Smith | 5807c41 | 2008-05-11 21:00:57 +0000 | [diff] [blame^] | 562 | if (format->type == 'n') |
| 563 | /* Compute how many additional chars we need to allocate |
| 564 | to hold the thousands grouping. */ |
| 565 | STRINGLIB_GROUPING(pnumeric_chars, n_digits, |
| 566 | pnumeric_chars+n_digits, |
| 567 | 0, &n_grouping_chars, 0); |
| 568 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 569 | /* Allocate a new string to hold the result */ |
Eric Smith | 5807c41 | 2008-05-11 21:00:57 +0000 | [diff] [blame^] | 570 | result = STRINGLIB_NEW(NULL, spec.n_total + n_grouping_chars); |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 571 | if (!result) |
| 572 | goto done; |
| 573 | p = STRINGLIB_STR(result); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 574 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 575 | /* Fill in the digit parts */ |
| 576 | n_leading_chars = spec.n_lpadding + spec.n_lsign + spec.n_spadding; |
| 577 | memmove(p + n_leading_chars, |
| 578 | pnumeric_chars, |
| 579 | n_digits * sizeof(STRINGLIB_CHAR)); |
| 580 | |
Eric Smith | 5807c41 | 2008-05-11 21:00:57 +0000 | [diff] [blame^] | 581 | /* If type is 'X', convert to uppercase */ |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 582 | if (format->type == 'X') { |
| 583 | Py_ssize_t t; |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 584 | for (t = 0; t < n_digits; ++t) |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 585 | p[t + n_leading_chars] = STRINGLIB_TOUPPER(p[t + n_leading_chars]); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 586 | } |
| 587 | |
Eric Smith | 5807c41 | 2008-05-11 21:00:57 +0000 | [diff] [blame^] | 588 | /* Insert the grouping, if any, after the uppercasing of 'X', so we can |
| 589 | ensure that grouping chars won't be affeted. */ |
| 590 | if (n_grouping_chars && format->type == 'n') { |
| 591 | /* We know this can't fail, since we've already |
| 592 | reserved enough space. */ |
| 593 | STRINGLIB_CHAR *pstart = p + n_leading_chars; |
| 594 | int r = STRINGLIB_GROUPING(pstart, n_digits, |
| 595 | pstart + n_digits, |
| 596 | spec.n_total+n_grouping_chars-n_leading_chars, |
| 597 | NULL, 0); |
| 598 | assert(r); |
| 599 | } |
| 600 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 601 | /* Fill in the non-digit parts */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 602 | fill_number(p, &spec, n_digits, |
| 603 | format->fill_char == '\0' ? ' ' : format->fill_char); |
| 604 | |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 605 | done: |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 606 | Py_XDECREF(tmp); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 607 | return result; |
| 608 | } |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 609 | #endif /* defined FORMAT_LONG || defined FORMAT_INT */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 610 | |
| 611 | /************************************************************************/ |
| 612 | /*********** float formatting *******************************************/ |
| 613 | /************************************************************************/ |
| 614 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 615 | #ifdef FORMAT_FLOAT |
| 616 | #if STRINGLIB_IS_UNICODE |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 617 | /* taken from unicodeobject.c */ |
| 618 | static Py_ssize_t |
| 619 | strtounicode(Py_UNICODE *buffer, const char *charbuffer) |
| 620 | { |
| 621 | register Py_ssize_t i; |
| 622 | Py_ssize_t len = strlen(charbuffer); |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 623 | for (i = len - 1; i >= 0; --i) |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 624 | buffer[i] = (Py_UNICODE) charbuffer[i]; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 625 | |
| 626 | return len; |
| 627 | } |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 628 | #endif |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 629 | |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 630 | /* see FORMATBUFLEN in unicodeobject.c */ |
| 631 | #define FLOAT_FORMATBUFLEN 120 |
| 632 | |
| 633 | /* much of this is taken from unicodeobject.c */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 634 | static PyObject * |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 635 | format_float_internal(PyObject *value, |
| 636 | const InternalFormatSpec *format) |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 637 | { |
| 638 | /* fmt = '%.' + `prec` + `type` + '%%' |
| 639 | worst case length = 2 + 10 (len of INT_MAX) + 1 + 2 = 15 (use 20)*/ |
| 640 | char fmt[20]; |
| 641 | |
| 642 | /* taken from unicodeobject.c */ |
| 643 | /* Worst case length calc to ensure no buffer overrun: |
| 644 | |
| 645 | 'g' formats: |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 646 | fmt = %#.<prec>g |
| 647 | buf = '-' + [0-9]*prec + '.' + 'e+' + (longest exp |
| 648 | for any double rep.) |
| 649 | len = 1 + prec + 1 + 2 + 5 = 9 + prec |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 650 | |
| 651 | 'f' formats: |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 652 | buf = '-' + [0-9]*x + '.' + [0-9]*prec (with x < 50) |
| 653 | len = 1 + 50 + 1 + prec = 52 + prec |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 654 | |
| 655 | If prec=0 the effective precision is 1 (the leading digit is |
| 656 | always given), therefore increase the length by one. |
| 657 | |
| 658 | */ |
| 659 | char charbuf[FLOAT_FORMATBUFLEN]; |
| 660 | Py_ssize_t n_digits; |
| 661 | double x; |
| 662 | Py_ssize_t precision = format->precision; |
| 663 | PyObject *result = NULL; |
| 664 | STRINGLIB_CHAR sign; |
| 665 | char* trailing = ""; |
| 666 | STRINGLIB_CHAR *p; |
| 667 | NumberFieldWidths spec; |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 668 | STRINGLIB_CHAR type = format->type; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 669 | |
| 670 | #if STRINGLIB_IS_UNICODE |
| 671 | Py_UNICODE unicodebuf[FLOAT_FORMATBUFLEN]; |
| 672 | #endif |
| 673 | |
| 674 | /* first, do the conversion as 8-bit chars, using the platform's |
| 675 | snprintf. then, if needed, convert to unicode. */ |
| 676 | |
| 677 | /* 'F' is the same as 'f', per the PEP */ |
| 678 | if (type == 'F') |
| 679 | type = 'f'; |
| 680 | |
| 681 | x = PyFloat_AsDouble(value); |
| 682 | |
| 683 | if (x == -1.0 && PyErr_Occurred()) |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 684 | goto done; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 685 | |
| 686 | if (type == '%') { |
| 687 | type = 'f'; |
| 688 | x *= 100; |
| 689 | trailing = "%"; |
| 690 | } |
| 691 | |
| 692 | if (precision < 0) |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 693 | precision = 6; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 694 | if (type == 'f' && (fabs(x) / 1e25) >= 1e25) |
Eric Smith | 185e30c | 2007-08-30 22:23:08 +0000 | [diff] [blame] | 695 | type = 'g'; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 696 | |
| 697 | /* cast "type", because if we're in unicode we need to pass a |
| 698 | 8-bit char. this is safe, because we've restricted what "type" |
| 699 | can be */ |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 700 | PyOS_snprintf(fmt, sizeof(fmt), "%%.%" PY_FORMAT_SIZE_T "d%c", precision, |
| 701 | (char)type); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 702 | |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 703 | /* do the actual formatting */ |
| 704 | PyOS_ascii_formatd(charbuf, sizeof(charbuf), fmt, x); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 705 | |
| 706 | /* adding trailing to fmt with PyOS_snprintf doesn't work, not |
| 707 | sure why. we'll just concatentate it here, no harm done. we |
| 708 | know we can't have a buffer overflow from the fmt size |
| 709 | analysis */ |
| 710 | strcat(charbuf, trailing); |
| 711 | |
| 712 | /* rather than duplicate the code for snprintf for both unicode |
| 713 | and 8 bit strings, we just use the 8 bit version and then |
| 714 | convert to unicode in a separate code path. that's probably |
| 715 | the lesser of 2 evils. */ |
| 716 | #if STRINGLIB_IS_UNICODE |
| 717 | n_digits = strtounicode(unicodebuf, charbuf); |
| 718 | p = unicodebuf; |
| 719 | #else |
| 720 | /* compute the length. I believe this is done because the return |
| 721 | value from snprintf above is unreliable */ |
| 722 | n_digits = strlen(charbuf); |
| 723 | p = charbuf; |
| 724 | #endif |
| 725 | |
| 726 | /* is a sign character present in the output? if so, remember it |
| 727 | and skip it */ |
| 728 | sign = p[0]; |
| 729 | if (sign == '-') { |
Christian Heimes | c3f30c4 | 2008-02-22 16:37:40 +0000 | [diff] [blame] | 730 | ++p; |
| 731 | --n_digits; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 732 | } |
| 733 | |
| 734 | calc_number_widths(&spec, sign, n_digits, format); |
| 735 | |
| 736 | /* allocate a string with enough space */ |
| 737 | result = STRINGLIB_NEW(NULL, spec.n_total); |
| 738 | if (result == NULL) |
| 739 | goto done; |
| 740 | |
| 741 | /* fill in the non-digit parts */ |
| 742 | fill_number(STRINGLIB_STR(result), &spec, n_digits, |
| 743 | format->fill_char == '\0' ? ' ' : format->fill_char); |
| 744 | |
| 745 | /* fill in the digit parts */ |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 746 | memmove(STRINGLIB_STR(result) + |
| 747 | (spec.n_lpadding + spec.n_lsign + spec.n_spadding), |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 748 | p, |
| 749 | n_digits * sizeof(STRINGLIB_CHAR)); |
| 750 | |
| 751 | done: |
| 752 | return result; |
| 753 | } |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 754 | #endif /* FORMAT_FLOAT */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 755 | |
| 756 | /************************************************************************/ |
| 757 | /*********** built in formatters ****************************************/ |
| 758 | /************************************************************************/ |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 759 | #ifdef FORMAT_STRING |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 760 | PyObject * |
| 761 | FORMAT_STRING(PyObject* value, PyObject* args) |
| 762 | { |
| 763 | PyObject *format_spec; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 764 | PyObject *result = NULL; |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 765 | #if PY_VERSION_HEX < 0x03000000 |
| 766 | PyObject *tmp = NULL; |
| 767 | #endif |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 768 | InternalFormatSpec format; |
| 769 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 770 | /* If 2.x, we accept either str or unicode, and try to convert it |
| 771 | to the right type. In 3.x, we insist on only unicode */ |
| 772 | #if PY_VERSION_HEX >= 0x03000000 |
| 773 | if (!PyArg_ParseTuple(args, STRINGLIB_PARSE_CODE ":__format__", |
| 774 | &format_spec)) |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 775 | goto done; |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 776 | #else |
| 777 | /* If 2.x, convert format_spec to the same type as value */ |
| 778 | /* This is to allow things like u''.format('') */ |
| 779 | if (!PyArg_ParseTuple(args, "O:__format__", &format_spec)) |
| 780 | goto done; |
| 781 | if (!(PyString_Check(format_spec) || PyUnicode_Check(format_spec))) { |
| 782 | PyErr_Format(PyExc_TypeError, "__format__ arg must be str " |
| 783 | "or unicode, not %s", Py_TYPE(format_spec)->tp_name); |
| 784 | goto done; |
| 785 | } |
| 786 | tmp = STRINGLIB_TOSTR(format_spec); |
| 787 | if (tmp == NULL) |
| 788 | goto done; |
| 789 | format_spec = tmp; |
| 790 | #endif |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 791 | |
| 792 | /* check for the special case of zero length format spec, make |
| 793 | it equivalent to str(value) */ |
| 794 | if (STRINGLIB_LEN(format_spec) == 0) { |
| 795 | result = STRINGLIB_TOSTR(value); |
| 796 | goto done; |
| 797 | } |
| 798 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 799 | |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 800 | /* parse the format_spec */ |
| 801 | if (!parse_internal_render_format_spec(format_spec, &format, 's')) |
| 802 | goto done; |
| 803 | |
| 804 | /* type conversion? */ |
| 805 | switch (format.type) { |
| 806 | case 's': |
| 807 | /* no type conversion needed, already a string. do the formatting */ |
| 808 | result = format_string_internal(value, &format); |
| 809 | break; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 810 | default: |
| 811 | /* unknown */ |
Martin v. Löwis | 5a6f458 | 2008-04-07 03:22:07 +0000 | [diff] [blame] | 812 | #if STRINGLIB_IS_UNICODE |
| 813 | /* If STRINGLIB_CHAR is Py_UNICODE, %c might be out-of-range, |
| 814 | hence the two cases. If it is char, gcc complains that the |
| 815 | condition below is always true, hence the ifdef. */ |
| 816 | if (format.type > 32 && format.type <128) |
| 817 | #endif |
| 818 | PyErr_Format(PyExc_ValueError, "Unknown conversion type %c", |
| 819 | (char)format.type); |
| 820 | #if STRINGLIB_IS_UNICODE |
| 821 | else |
| 822 | PyErr_Format(PyExc_ValueError, "Unknown conversion type '\\x%x'", |
| 823 | (unsigned int)format.type); |
| 824 | #endif |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 825 | goto done; |
| 826 | } |
| 827 | |
| 828 | done: |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 829 | #if PY_VERSION_HEX < 0x03000000 |
| 830 | Py_XDECREF(tmp); |
| 831 | #endif |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 832 | return result; |
| 833 | } |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 834 | #endif /* FORMAT_STRING */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 835 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 836 | #if defined FORMAT_LONG || defined FORMAT_INT |
| 837 | static PyObject* |
| 838 | format_int_or_long(PyObject* value, PyObject* args, IntOrLongToString tostring) |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 839 | { |
| 840 | PyObject *format_spec; |
| 841 | PyObject *result = NULL; |
| 842 | PyObject *tmp = NULL; |
| 843 | InternalFormatSpec format; |
| 844 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 845 | if (!PyArg_ParseTuple(args, STRINGLIB_PARSE_CODE ":__format__", |
| 846 | &format_spec)) |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 847 | goto done; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 848 | |
| 849 | /* check for the special case of zero length format spec, make |
| 850 | it equivalent to str(value) */ |
| 851 | if (STRINGLIB_LEN(format_spec) == 0) { |
| 852 | result = STRINGLIB_TOSTR(value); |
| 853 | goto done; |
| 854 | } |
| 855 | |
| 856 | /* parse the format_spec */ |
| 857 | if (!parse_internal_render_format_spec(format_spec, &format, 'd')) |
| 858 | goto done; |
| 859 | |
| 860 | /* type conversion? */ |
| 861 | switch (format.type) { |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 862 | case 'b': |
| 863 | case 'c': |
| 864 | case 'd': |
| 865 | case 'o': |
| 866 | case 'x': |
| 867 | case 'X': |
Eric Smith | 5807c41 | 2008-05-11 21:00:57 +0000 | [diff] [blame^] | 868 | case 'n': |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 869 | /* no type conversion needed, already an int (or long). do |
| 870 | the formatting */ |
| 871 | result = format_int_or_long_internal(value, &format, tostring); |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 872 | break; |
| 873 | |
Eric Smith | fa767ef | 2008-01-28 10:59:27 +0000 | [diff] [blame] | 874 | case 'e': |
| 875 | case 'E': |
| 876 | case 'f': |
| 877 | case 'F': |
| 878 | case 'g': |
| 879 | case 'G': |
Eric Smith | fa767ef | 2008-01-28 10:59:27 +0000 | [diff] [blame] | 880 | case '%': |
| 881 | /* convert to float */ |
| 882 | tmp = PyNumber_Float(value); |
| 883 | if (tmp == NULL) |
| 884 | goto done; |
| 885 | result = format_float_internal(value, &format); |
| 886 | break; |
| 887 | |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 888 | default: |
| 889 | /* unknown */ |
| 890 | PyErr_Format(PyExc_ValueError, "Unknown conversion type %c", |
| 891 | format.type); |
| 892 | goto done; |
| 893 | } |
| 894 | |
| 895 | done: |
| 896 | Py_XDECREF(tmp); |
| 897 | return result; |
| 898 | } |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 899 | #endif /* FORMAT_LONG || defined FORMAT_INT */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 900 | |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 901 | #ifdef FORMAT_LONG |
| 902 | /* Need to define long_format as a function that will convert a long |
| 903 | to a string. In 3.0, _PyLong_Format has the correct signature. In |
| 904 | 2.x, we need to fudge a few parameters */ |
| 905 | #if PY_VERSION_HEX >= 0x03000000 |
| 906 | #define long_format _PyLong_Format |
| 907 | #else |
| 908 | static PyObject* |
| 909 | long_format(PyObject* value, int base) |
| 910 | { |
| 911 | /* Convert to base, don't add trailing 'L', and use the new octal |
| 912 | format. We already know this is a long object */ |
| 913 | assert(PyLong_Check(value)); |
| 914 | /* convert to base, don't add 'L', and use the new octal format */ |
| 915 | return _PyLong_Format(value, base, 0, 1); |
| 916 | } |
| 917 | #endif |
| 918 | |
| 919 | PyObject * |
| 920 | FORMAT_LONG(PyObject* value, PyObject* args) |
| 921 | { |
| 922 | return format_int_or_long(value, args, long_format); |
| 923 | } |
| 924 | #endif /* FORMAT_LONG */ |
| 925 | |
| 926 | #ifdef FORMAT_INT |
| 927 | /* this is only used for 2.x, not 3.0 */ |
| 928 | static PyObject* |
| 929 | int_format(PyObject* value, int base) |
| 930 | { |
| 931 | /* Convert to base, and use the new octal format. We already |
| 932 | know this is an int object */ |
| 933 | assert(PyInt_Check(value)); |
| 934 | return _PyInt_Format((PyIntObject*)value, base, 1); |
| 935 | } |
| 936 | |
| 937 | PyObject * |
| 938 | FORMAT_INT(PyObject* value, PyObject* args) |
| 939 | { |
| 940 | return format_int_or_long(value, args, int_format); |
| 941 | } |
| 942 | #endif /* FORMAT_INT */ |
| 943 | |
| 944 | #ifdef FORMAT_FLOAT |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 945 | PyObject * |
| 946 | FORMAT_FLOAT(PyObject *value, PyObject *args) |
| 947 | { |
| 948 | PyObject *format_spec; |
| 949 | PyObject *result = NULL; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 950 | InternalFormatSpec format; |
| 951 | |
Eric Smith | 37f1038 | 2007-09-01 10:56:01 +0000 | [diff] [blame] | 952 | if (!PyArg_ParseTuple(args, STRINGLIB_PARSE_CODE ":__format__", &format_spec)) |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 953 | goto done; |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 954 | |
| 955 | /* check for the special case of zero length format spec, make |
| 956 | it equivalent to str(value) */ |
| 957 | if (STRINGLIB_LEN(format_spec) == 0) { |
| 958 | result = STRINGLIB_TOSTR(value); |
| 959 | goto done; |
| 960 | } |
| 961 | |
| 962 | /* parse the format_spec */ |
Christian Heimes | b186d00 | 2008-03-18 15:15:01 +0000 | [diff] [blame] | 963 | if (!parse_internal_render_format_spec(format_spec, &format, '\0')) |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 964 | goto done; |
| 965 | |
| 966 | /* type conversion? */ |
| 967 | switch (format.type) { |
Christian Heimes | b186d00 | 2008-03-18 15:15:01 +0000 | [diff] [blame] | 968 | case '\0': |
| 969 | /* 'Z' means like 'g', but with at least one decimal. See |
| 970 | PyOS_ascii_formatd */ |
| 971 | format.type = 'Z'; |
| 972 | /* Deliberate fall through to the next case statement */ |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 973 | case 'e': |
| 974 | case 'E': |
| 975 | case 'f': |
| 976 | case 'F': |
| 977 | case 'g': |
| 978 | case 'G': |
| 979 | case 'n': |
| 980 | case '%': |
| 981 | /* no conversion, already a float. do the formatting */ |
| 982 | result = format_float_internal(value, &format); |
| 983 | break; |
| 984 | |
| 985 | default: |
| 986 | /* unknown */ |
| 987 | PyErr_Format(PyExc_ValueError, "Unknown conversion type %c", |
| 988 | format.type); |
| 989 | goto done; |
| 990 | } |
| 991 | |
| 992 | done: |
Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame] | 993 | return result; |
| 994 | } |
Eric Smith | 8fd3eba | 2008-02-17 19:48:00 +0000 | [diff] [blame] | 995 | #endif /* FORMAT_FLOAT */ |