Eric Smith | 8c66326 | 2007-08-25 02:26:07 +0000 | [diff] [blame^] | 1 | /* |
| 2 | string_format.h -- implementation of string.format(). |
| 3 | |
| 4 | It uses the Objects/stringlib conventions, so that it can be |
| 5 | compiled for both unicode and string objects. |
| 6 | */ |
| 7 | |
| 8 | |
| 9 | /* Defines for more efficiently reallocating the string buffer */ |
| 10 | #define INITIAL_SIZE_INCREMENT 100 |
| 11 | #define SIZE_MULTIPLIER 2 |
| 12 | #define MAX_SIZE_INCREMENT 3200 |
| 13 | |
| 14 | |
| 15 | /************************************************************************/ |
| 16 | /*********** Global data structures and forward declarations *********/ |
| 17 | /************************************************************************/ |
| 18 | |
| 19 | /* |
| 20 | A SubString consists of the characters between two string or |
| 21 | unicode pointers. |
| 22 | */ |
| 23 | typedef struct { |
| 24 | STRINGLIB_CHAR *ptr; |
| 25 | STRINGLIB_CHAR *end; |
| 26 | } SubString; |
| 27 | |
| 28 | |
| 29 | /* forward declaration for recursion */ |
| 30 | static PyObject * |
| 31 | build_string(SubString *input, PyObject *args, PyObject *kwargs, |
| 32 | int *recursion_level); |
| 33 | |
| 34 | |
| 35 | |
| 36 | /************************************************************************/ |
| 37 | /************************** Utility functions ************************/ |
| 38 | /************************************************************************/ |
| 39 | |
| 40 | /* fill in a SubString from a pointer and length */ |
| 41 | Py_LOCAL_INLINE(void) |
| 42 | SubString_init(SubString *str, STRINGLIB_CHAR *p, Py_ssize_t len) |
| 43 | { |
| 44 | str->ptr = p; |
| 45 | if (p == NULL) |
| 46 | str->end = NULL; |
| 47 | else |
| 48 | str->end = str->ptr + len; |
| 49 | } |
| 50 | |
| 51 | Py_LOCAL_INLINE(PyObject *) |
| 52 | SubString_new_object(SubString *str) |
| 53 | { |
| 54 | return STRINGLIB_NEW(str->ptr, str->end - str->ptr); |
| 55 | } |
| 56 | |
| 57 | /************************************************************************/ |
| 58 | /*********** Error handling and exception generation **************/ |
| 59 | /************************************************************************/ |
| 60 | |
| 61 | /* |
| 62 | Most of our errors are value errors, because to Python, the |
| 63 | format string is a "value". Also, it's convenient to return |
| 64 | a NULL when we are erroring out. |
| 65 | |
| 66 | XXX: need better error handling, per PEP 3101. |
| 67 | */ |
| 68 | static void * |
| 69 | SetError(const char *s) |
| 70 | { |
| 71 | /* PyErr_Format always returns NULL */ |
| 72 | return PyErr_Format(PyExc_ValueError, "%s in format string", s); |
| 73 | } |
| 74 | |
| 75 | /* |
| 76 | check_input returns True if we still have characters |
| 77 | left in the input string. |
| 78 | |
| 79 | XXX: make this function go away when better error handling is |
| 80 | implemented. |
| 81 | */ |
| 82 | Py_LOCAL_INLINE(int) |
| 83 | check_input(SubString *input) |
| 84 | { |
| 85 | if (input->ptr < input->end) |
| 86 | return 1; |
| 87 | PyErr_SetString(PyExc_ValueError, |
| 88 | "unterminated replacement field"); |
| 89 | return 0; |
| 90 | } |
| 91 | |
| 92 | /************************************************************************/ |
| 93 | /*********** Output string management functions ****************/ |
| 94 | /************************************************************************/ |
| 95 | |
| 96 | typedef struct { |
| 97 | STRINGLIB_CHAR *ptr; |
| 98 | STRINGLIB_CHAR *end; |
| 99 | PyObject *obj; |
| 100 | Py_ssize_t size_increment; |
| 101 | } OutputString; |
| 102 | |
| 103 | /* initialize an OutputString object, reserving size characters */ |
| 104 | static int |
| 105 | output_initialize(OutputString *output, Py_ssize_t size) |
| 106 | { |
| 107 | output->obj = STRINGLIB_NEW(NULL, size); |
| 108 | if (output->obj == NULL) |
| 109 | return 0; |
| 110 | |
| 111 | output->ptr = STRINGLIB_STR(output->obj); |
| 112 | output->end = STRINGLIB_LEN(output->obj) + output->ptr; |
| 113 | output->size_increment = INITIAL_SIZE_INCREMENT; |
| 114 | |
| 115 | return 1; |
| 116 | } |
| 117 | |
| 118 | /* |
| 119 | output_extend reallocates the output string buffer. |
| 120 | It returns a status: 0 for a failed reallocation, |
| 121 | 1 for success. |
| 122 | */ |
| 123 | |
| 124 | static int |
| 125 | output_extend(OutputString *output, Py_ssize_t count) |
| 126 | { |
| 127 | STRINGLIB_CHAR *startptr = STRINGLIB_STR(output->obj); |
| 128 | Py_ssize_t curlen = output->ptr - startptr; |
| 129 | Py_ssize_t maxlen = curlen + count + output->size_increment; |
| 130 | |
| 131 | if (STRINGLIB_RESIZE(&output->obj, maxlen) < 0) |
| 132 | return 0; |
| 133 | startptr = STRINGLIB_STR(output->obj); |
| 134 | output->ptr = startptr + curlen; |
| 135 | output->end = startptr + maxlen; |
| 136 | if (output->size_increment < MAX_SIZE_INCREMENT) |
| 137 | output->size_increment *= SIZE_MULTIPLIER; |
| 138 | return 1; |
| 139 | } |
| 140 | |
| 141 | /* |
| 142 | output_data dumps characters into our output string |
| 143 | buffer. |
| 144 | |
| 145 | In some cases, it has to reallocate the string. |
| 146 | |
| 147 | It returns a status: 0 for a failed reallocation, |
| 148 | 1 for success. |
| 149 | */ |
| 150 | static int |
| 151 | output_data(OutputString *output, const STRINGLIB_CHAR *s, Py_ssize_t count) |
| 152 | { |
| 153 | if ((count > output->end - output->ptr) && !output_extend(output, count)) |
| 154 | return 0; |
| 155 | memcpy(output->ptr, s, count * sizeof(STRINGLIB_CHAR)); |
| 156 | output->ptr += count; |
| 157 | return 1; |
| 158 | } |
| 159 | |
| 160 | /************************************************************************/ |
| 161 | /*********** Format string parsing -- integers and identifiers *********/ |
| 162 | /************************************************************************/ |
| 163 | |
| 164 | /* |
| 165 | end_identifier returns true if a character marks |
| 166 | the end of an identifier string. |
| 167 | |
| 168 | Although the PEP specifies that identifiers are |
| 169 | numbers or valid Python identifiers, we just let |
| 170 | getattr/getitem handle that, so the implementation |
| 171 | is more flexible than the PEP would indicate. |
| 172 | */ |
| 173 | Py_LOCAL_INLINE(int) |
| 174 | end_identifier(STRINGLIB_CHAR c) |
| 175 | { |
| 176 | switch (c) { |
| 177 | case '.': case '[': case ']': |
| 178 | return 1; |
| 179 | default: |
| 180 | return 0; |
| 181 | } |
| 182 | } |
| 183 | |
| 184 | /* |
| 185 | get_integer consumes 0 or more decimal digit characters from an |
| 186 | input string, updates *result with the corresponding positive |
| 187 | integer, and returns the number of digits consumed. |
| 188 | |
| 189 | returns -1 on error. |
| 190 | */ |
| 191 | static int |
| 192 | get_integer(STRINGLIB_CHAR **ptr, STRINGLIB_CHAR *end, |
| 193 | Py_ssize_t *result) |
| 194 | { |
| 195 | Py_ssize_t accumulator, digitval, oldaccumulator; |
| 196 | int numdigits; |
| 197 | accumulator = numdigits = 0; |
| 198 | for (;;(*ptr)++, numdigits++) { |
| 199 | if (*ptr >= end) |
| 200 | break; |
| 201 | digitval = STRINGLIB_TODECIMAL(**ptr); |
| 202 | if (digitval < 0) |
| 203 | break; |
| 204 | /* |
| 205 | This trick was copied from old Unicode format code. It's cute, |
| 206 | but would really suck on an old machine with a slow divide |
| 207 | implementation. Fortunately, in the normal case we do not |
| 208 | expect too many digits. |
| 209 | */ |
| 210 | oldaccumulator = accumulator; |
| 211 | accumulator *= 10; |
| 212 | if ((accumulator+10)/10 != oldaccumulator+1) { |
| 213 | PyErr_Format(PyExc_ValueError, |
| 214 | "Too many decimal digits in format string"); |
| 215 | return -1; |
| 216 | } |
| 217 | accumulator += digitval; |
| 218 | } |
| 219 | *result = accumulator; |
| 220 | return numdigits; |
| 221 | } |
| 222 | |
| 223 | /* |
| 224 | get_identifier is a bit of a misnomer. It returns a value for use |
| 225 | with getattr or getindex. This value will a string/unicode |
| 226 | object. The input cannot be zero length. Continues until end of |
| 227 | input, or end_identifier() returns true. |
| 228 | */ |
| 229 | static PyObject * |
| 230 | get_identifier(SubString *input) |
| 231 | { |
| 232 | STRINGLIB_CHAR *start; |
| 233 | |
| 234 | for (start = input->ptr; |
| 235 | input->ptr < input->end && !end_identifier(*input->ptr); |
| 236 | input->ptr++) |
| 237 | ; |
| 238 | |
| 239 | return STRINGLIB_NEW(start, input->ptr - start); |
| 240 | |
| 241 | /* |
| 242 | We might want to add code here to check for invalid Python |
| 243 | identifiers. All identifiers are eventually passed to getattr |
| 244 | or getitem, so there is a check when used. However, we might |
| 245 | want to remove (or not) the ability to have strings like |
| 246 | "a/b" or " ab" or "-1" (which is not parsed as a number). |
| 247 | For now, this is left as an exercise for the first disgruntled |
| 248 | user... |
| 249 | |
| 250 | if (XXX -- need check function) { |
| 251 | Py_DECREF(result); |
| 252 | PyErr_SetString(PyExc_ValueError, |
| 253 | "Invalid embedded Python identifier"); |
| 254 | return NULL; |
| 255 | } |
| 256 | */ |
| 257 | } |
| 258 | |
| 259 | /************************************************************************/ |
| 260 | /******** Functions to get field objects and specification strings ******/ |
| 261 | /************************************************************************/ |
| 262 | |
| 263 | /* get_field_and_spec is the main function in this section. It parses |
| 264 | the format string well enough to return a field object to render along |
| 265 | with a field specification string. |
| 266 | */ |
| 267 | |
| 268 | /* |
| 269 | look up key in our keyword arguments |
| 270 | */ |
| 271 | static PyObject * |
| 272 | key_lookup(PyObject *kwargs, PyObject *key) |
| 273 | { |
| 274 | PyObject *result; |
| 275 | |
| 276 | if (kwargs && (result = PyDict_GetItem(kwargs, key)) != NULL) { |
| 277 | Py_INCREF(result); |
| 278 | return result; |
| 279 | } |
| 280 | return NULL; |
| 281 | } |
| 282 | |
| 283 | /* |
| 284 | get_field_object returns the object inside {}, before the |
| 285 | format_spec. It handles getindex and getattr lookups and consumes |
| 286 | the entire input string. |
| 287 | */ |
| 288 | static PyObject * |
| 289 | get_field_object(SubString *input, PyObject *args, PyObject *kwargs) |
| 290 | { |
| 291 | PyObject *myobj, *subobj, *newobj; |
| 292 | STRINGLIB_CHAR c; |
| 293 | Py_ssize_t index; |
| 294 | int isindex, isnumeric, isargument; |
| 295 | |
| 296 | index = isnumeric = 0; /* Just to shut up the compiler warnings */ |
| 297 | |
| 298 | myobj = args; |
| 299 | Py_INCREF(myobj); |
| 300 | |
| 301 | for (isindex=1, isargument=1;;) { |
| 302 | if (!check_input(input)) |
| 303 | break; |
| 304 | if (!isindex) { |
| 305 | if ((subobj = get_identifier(input)) == NULL) |
| 306 | break; |
| 307 | newobj = PyObject_GetAttr(myobj, subobj); |
| 308 | Py_DECREF(subobj); |
| 309 | } else { |
| 310 | isnumeric = (STRINGLIB_ISDECIMAL(*input->ptr)); |
| 311 | if (isnumeric) |
| 312 | /* XXX: add error checking */ |
| 313 | get_integer(&input->ptr, input->end, &index); |
| 314 | |
| 315 | if (isnumeric && PySequence_Check(myobj)) |
| 316 | newobj = PySequence_GetItem(myobj, index); |
| 317 | else { |
| 318 | /* XXX -- do we need PyLong_FromLongLong? |
| 319 | Using ssizet, not int... */ |
| 320 | subobj = isnumeric ? |
| 321 | PyInt_FromLong(index) : |
| 322 | get_identifier(input); |
| 323 | if (subobj == NULL) |
| 324 | break; |
| 325 | if (isargument) { |
| 326 | newobj = key_lookup(kwargs, subobj); |
| 327 | } else { |
| 328 | newobj = PyObject_GetItem(myobj, subobj); |
| 329 | } |
| 330 | Py_DECREF(subobj); |
| 331 | } |
| 332 | } |
| 333 | Py_DECREF(myobj); |
| 334 | myobj = newobj; |
| 335 | if (myobj == NULL) |
| 336 | break; |
| 337 | if (!isargument && isindex) |
| 338 | if ((!check_input(input)) || (*(input->ptr++) != ']')) { |
| 339 | SetError("Expected ]"); |
| 340 | break; |
| 341 | } |
| 342 | |
| 343 | /* if at the end of input, return with myobj */ |
| 344 | if (input->ptr >= input->end) |
| 345 | return myobj; |
| 346 | |
| 347 | c = *input->ptr; |
| 348 | input->ptr++; |
| 349 | isargument = 0; |
| 350 | isindex = (c == '['); |
| 351 | if (!isindex && (c != '.')) { |
| 352 | SetError("Expected ., [, :, !, or }"); |
| 353 | break; |
| 354 | } |
| 355 | } |
| 356 | if ((myobj == NULL) && isargument) { |
| 357 | /* XXX: include more useful error information, like which |
| 358 | * keyword not found or which index missing */ |
| 359 | PyErr_Clear(); |
| 360 | return SetError(isnumeric |
| 361 | ? "Not enough positional arguments" |
| 362 | : "Keyword argument not found"); |
| 363 | } |
| 364 | Py_XDECREF(myobj); |
| 365 | return NULL; |
| 366 | } |
| 367 | |
| 368 | /************************************************************************/ |
| 369 | /***************** Field rendering functions **************************/ |
| 370 | /************************************************************************/ |
| 371 | |
| 372 | /* |
| 373 | render_field() is the main function in this section. It takes the |
| 374 | field object and field specification string generated by |
| 375 | get_field_and_spec, and renders the field into the output string. |
| 376 | |
| 377 | format() does the actual calling of the objects __format__ method. |
| 378 | */ |
| 379 | |
| 380 | |
| 381 | /* returns fieldobj.__format__(format_spec) */ |
| 382 | static PyObject * |
| 383 | format(PyObject *fieldobj, SubString *format_spec) |
| 384 | { |
| 385 | static PyObject *format_str = NULL; |
| 386 | PyObject *meth; |
| 387 | PyObject *spec = NULL; |
| 388 | PyObject *result = NULL; |
| 389 | |
| 390 | /* Initialize cached value */ |
| 391 | if (format_str == NULL) { |
| 392 | /* Initialize static variable needed by _PyType_Lookup */ |
| 393 | format_str = PyUnicode_FromString("__format__"); |
| 394 | if (format_str == NULL) |
| 395 | return NULL; |
| 396 | } |
| 397 | |
| 398 | /* Make sure the type is initialized. float gets initialized late */ |
| 399 | if (Py_Type(fieldobj)->tp_dict == NULL) |
| 400 | if (PyType_Ready(Py_Type(fieldobj)) < 0) |
| 401 | return NULL; |
| 402 | |
| 403 | /* we need to create an object out of the pointers we have */ |
| 404 | spec = SubString_new_object(format_spec); |
| 405 | if (spec == NULL) |
| 406 | goto done; |
| 407 | |
| 408 | /* Find the (unbound!) __format__ method (a borrowed reference) */ |
| 409 | meth = _PyType_Lookup(Py_Type(fieldobj), format_str); |
| 410 | if (meth == NULL) { |
| 411 | PyErr_Format(PyExc_TypeError, |
| 412 | "Type %.100s doesn't define __format__", |
| 413 | Py_Type(fieldobj)->tp_name); |
| 414 | goto done; |
| 415 | } |
| 416 | |
| 417 | /* And call it, binding it to the value */ |
| 418 | result = PyObject_CallFunctionObjArgs(meth, fieldobj, spec, NULL); |
| 419 | if (result == NULL) |
| 420 | goto done; |
| 421 | |
| 422 | if (!STRINGLIB_CHECK(result)) { |
| 423 | PyErr_SetString(PyExc_TypeError, |
| 424 | "__format__ method did not return " |
| 425 | STRINGLIB_TYPE_NAME); |
| 426 | Py_DECREF(result); |
| 427 | result = NULL; |
| 428 | goto done; |
| 429 | } |
| 430 | |
| 431 | done: |
| 432 | Py_XDECREF(spec); |
| 433 | return result; |
| 434 | } |
| 435 | |
| 436 | /* |
| 437 | render_field calls fieldobj.__format__(format_spec) method, and |
| 438 | appends to the output. |
| 439 | */ |
| 440 | static int |
| 441 | render_field(PyObject *fieldobj, SubString *format_spec, OutputString *output) |
| 442 | { |
| 443 | int ok = 0; |
| 444 | PyObject *result = format(fieldobj, format_spec); |
| 445 | |
| 446 | if (result == NULL) |
| 447 | goto done; |
| 448 | |
| 449 | ok = output_data(output, |
| 450 | STRINGLIB_STR(result), STRINGLIB_LEN(result)); |
| 451 | done: |
| 452 | Py_XDECREF(result); |
| 453 | return ok; |
| 454 | } |
| 455 | |
| 456 | static int |
| 457 | parse_field(SubString *str, SubString *field_name, SubString *format_spec, |
| 458 | STRINGLIB_CHAR *conversion) |
| 459 | { |
| 460 | STRINGLIB_CHAR c = 0; |
| 461 | |
| 462 | /* initialize these, as they may be empty */ |
| 463 | *conversion = '\0'; |
| 464 | SubString_init(format_spec, NULL, 0); |
| 465 | |
| 466 | /* search for the field name. it's terminated by the end of the |
| 467 | string, or a ':' or '!' */ |
| 468 | field_name->ptr = str->ptr; |
| 469 | while (str->ptr < str->end) { |
| 470 | switch (c = *(str->ptr++)) { |
| 471 | case ':': |
| 472 | case '!': |
| 473 | break; |
| 474 | default: |
| 475 | continue; |
| 476 | } |
| 477 | break; |
| 478 | } |
| 479 | |
| 480 | if (c == '!' || c == ':') { |
| 481 | /* we have a format specifier and/or a conversion */ |
| 482 | /* don't include the last character */ |
| 483 | field_name->end = str->ptr-1; |
| 484 | |
| 485 | /* the format specifier is the rest of the string */ |
| 486 | format_spec->ptr = str->ptr; |
| 487 | format_spec->end = str->end; |
| 488 | |
| 489 | /* see if there's a conversion specifier */ |
| 490 | if (c == '!') { |
| 491 | /* there must be another character present */ |
| 492 | if (format_spec->ptr >= format_spec->end) { |
| 493 | PyErr_SetString(PyExc_ValueError, |
| 494 | "end of format while looking for conversion " |
| 495 | "specifier"); |
| 496 | return 0; |
| 497 | } |
| 498 | *conversion = *(format_spec->ptr++); |
| 499 | |
| 500 | /* if there is another character, it must be a colon */ |
| 501 | if (format_spec->ptr < format_spec->end) { |
| 502 | c = *(format_spec->ptr++); |
| 503 | if (c != ':') { |
| 504 | PyErr_SetString(PyExc_ValueError, |
| 505 | "expected ':' after format specifier"); |
| 506 | return 0; |
| 507 | } |
| 508 | } |
| 509 | } |
| 510 | |
| 511 | return 1; |
| 512 | |
| 513 | } else { |
| 514 | /* end of string, there's no format_spec or conversion */ |
| 515 | field_name->end = str->ptr; |
| 516 | return 1; |
| 517 | } |
| 518 | } |
| 519 | |
| 520 | /************************************************************************/ |
| 521 | /******* Output string allocation and escape-to-markup processing ******/ |
| 522 | /************************************************************************/ |
| 523 | |
| 524 | /* MarkupIterator breaks the string into pieces of either literal |
| 525 | text, or things inside {} that need to be marked up. it is |
| 526 | designed to make it easy to wrap a Python iterator around it, for |
| 527 | use with the Formatter class */ |
| 528 | |
| 529 | typedef struct { |
| 530 | SubString str; |
| 531 | int in_markup; |
| 532 | } MarkupIterator; |
| 533 | |
| 534 | static int |
| 535 | MarkupIterator_init(MarkupIterator *self, STRINGLIB_CHAR *ptr, Py_ssize_t len) |
| 536 | { |
| 537 | SubString_init(&self->str, ptr, len); |
| 538 | self->in_markup = 0; |
| 539 | return 1; |
| 540 | } |
| 541 | |
| 542 | /* returns 0 on error, 1 on non-error termination, and 2 if it got a |
| 543 | string (or something to be expanded) */ |
| 544 | static int |
| 545 | MarkupIterator_next(MarkupIterator *self, int *is_markup, SubString *literal, |
| 546 | SubString *field_name, SubString *format_spec, |
| 547 | STRINGLIB_CHAR *conversion, |
| 548 | int *format_spec_needs_expanding) |
| 549 | { |
| 550 | int at_end; |
| 551 | STRINGLIB_CHAR c = 0; |
| 552 | STRINGLIB_CHAR *start; |
| 553 | int count; |
| 554 | Py_ssize_t len; |
| 555 | |
| 556 | *format_spec_needs_expanding = 0; |
| 557 | |
| 558 | /* no more input, end of iterator */ |
| 559 | if (self->str.ptr >= self->str.end) |
| 560 | return 1; |
| 561 | |
| 562 | *is_markup = self->in_markup; |
| 563 | start = self->str.ptr; |
| 564 | |
| 565 | if (self->in_markup) { |
| 566 | |
| 567 | /* prepare for next iteration */ |
| 568 | self->in_markup = 0; |
| 569 | |
| 570 | /* this is markup, find the end of the string by counting nested |
| 571 | braces. note that this prohibits escaped braces, so that |
| 572 | format_specs cannot have braces in them. */ |
| 573 | count = 1; |
| 574 | |
| 575 | /* we know we can't have a zero length string, so don't worry |
| 576 | about that case */ |
| 577 | while (self->str.ptr < self->str.end) { |
| 578 | switch (c = *(self->str.ptr++)) { |
| 579 | case '{': |
| 580 | /* the format spec needs to be recursively expanded. |
| 581 | this is an optimization, and not strictly needed */ |
| 582 | *format_spec_needs_expanding = 1; |
| 583 | count++; |
| 584 | break; |
| 585 | case '}': |
| 586 | count--; |
| 587 | if (count <= 0) { |
| 588 | /* we're done. parse and get out */ |
| 589 | literal->ptr = start; |
| 590 | literal->end = self->str.ptr-1; |
| 591 | |
| 592 | if (parse_field(literal, field_name, format_spec, |
| 593 | conversion) == 0) |
| 594 | return 0; |
| 595 | |
| 596 | /* success */ |
| 597 | return 2; |
| 598 | } |
| 599 | break; |
| 600 | } |
| 601 | } |
| 602 | /* end of string while searching for matching '}' */ |
| 603 | PyErr_SetString(PyExc_ValueError, "unmatched '{' in format"); |
| 604 | return 0; |
| 605 | |
| 606 | } else { |
| 607 | /* literal text, read until the end of string, an escaped { or }, |
| 608 | or an unescaped { */ |
| 609 | while (self->str.ptr < self->str.end) { |
| 610 | switch (c = *(self->str.ptr++)) { |
| 611 | case '{': |
| 612 | case '}': |
| 613 | self->in_markup = 1; |
| 614 | break; |
| 615 | default: |
| 616 | continue; |
| 617 | } |
| 618 | break; |
| 619 | } |
| 620 | |
| 621 | at_end = self->str.ptr >= self->str.end; |
| 622 | len = self->str.ptr - start; |
| 623 | |
| 624 | if ((c == '}') && (at_end || (c != *self->str.ptr))) |
| 625 | return (int)SetError("Single } encountered"); |
| 626 | if (at_end && c == '{') |
| 627 | return (int)SetError("Single { encountered"); |
| 628 | if (!at_end) { |
| 629 | if (c == *self->str.ptr) { |
| 630 | /* escaped } or {, skip it in the input */ |
| 631 | self->str.ptr++; |
| 632 | self->in_markup = 0; |
| 633 | } else |
| 634 | len--; |
| 635 | } |
| 636 | |
| 637 | /* this is just plain text, return it */ |
| 638 | literal->ptr = start; |
| 639 | literal->end = start + len; |
| 640 | return 2; |
| 641 | } |
| 642 | } |
| 643 | |
| 644 | |
| 645 | /* do the !r or !s conversion on obj */ |
| 646 | static PyObject * |
| 647 | do_conversion(PyObject *obj, STRINGLIB_CHAR conversion) |
| 648 | { |
| 649 | /* XXX in pre-3.0, do we need to convert this to unicode, since it |
| 650 | might have returned a string? */ |
| 651 | switch (conversion) { |
| 652 | case 'r': |
| 653 | return PyObject_Repr(obj); |
| 654 | case 's': |
| 655 | return PyObject_Unicode(obj); |
| 656 | default: |
| 657 | PyErr_Format(PyExc_ValueError, |
| 658 | "Unknown converion specifier %c", |
| 659 | conversion); |
| 660 | return NULL; |
| 661 | } |
| 662 | } |
| 663 | |
| 664 | /* given: |
| 665 | |
| 666 | {field_name!conversion:format_spec} |
| 667 | |
| 668 | compute the result and write it to output. |
| 669 | format_spec_needs_expanding is an optimization. if it's false, |
| 670 | just output the string directly, otherwise recursively expand the |
| 671 | format_spec string. */ |
| 672 | |
| 673 | static int |
| 674 | output_markup(SubString *field_name, SubString *format_spec, |
| 675 | int format_spec_needs_expanding, STRINGLIB_CHAR conversion, |
| 676 | OutputString *output, PyObject *args, PyObject *kwargs, |
| 677 | int *recursion_level) |
| 678 | { |
| 679 | PyObject *tmp = NULL; |
| 680 | PyObject *fieldobj = NULL; |
| 681 | SubString expanded_format_spec; |
| 682 | SubString *actual_format_spec; |
| 683 | int result = 0; |
| 684 | |
| 685 | /* convert field_name to an object */ |
| 686 | fieldobj = get_field_object(field_name, args, kwargs); |
| 687 | if (fieldobj == NULL) |
| 688 | goto done; |
| 689 | |
| 690 | if (conversion != '\0') { |
| 691 | tmp = do_conversion(fieldobj, conversion); |
| 692 | if (tmp == NULL) |
| 693 | goto done; |
| 694 | |
| 695 | /* do the assignment, transferring ownership: fieldobj = tmp */ |
| 696 | Py_DECREF(fieldobj); |
| 697 | fieldobj = tmp; |
| 698 | tmp = NULL; |
| 699 | } |
| 700 | |
| 701 | /* if needed, recurively compute the format_spec */ |
| 702 | if (format_spec_needs_expanding) { |
| 703 | tmp = build_string(format_spec, args, kwargs, recursion_level); |
| 704 | if (tmp == NULL) |
| 705 | goto done; |
| 706 | |
| 707 | /* note that in the case we're expanding the format string, |
| 708 | tmp must be kept around until after the call to |
| 709 | render_field. */ |
| 710 | SubString_init(&expanded_format_spec, |
| 711 | STRINGLIB_STR(tmp), STRINGLIB_LEN(tmp)); |
| 712 | actual_format_spec = &expanded_format_spec; |
| 713 | } else |
| 714 | actual_format_spec = format_spec; |
| 715 | |
| 716 | if (render_field(fieldobj, actual_format_spec, output) == 0) |
| 717 | goto done; |
| 718 | |
| 719 | result = 1; |
| 720 | |
| 721 | done: |
| 722 | Py_XDECREF(fieldobj); |
| 723 | Py_XDECREF(tmp); |
| 724 | |
| 725 | return result; |
| 726 | } |
| 727 | |
| 728 | /* |
| 729 | do_markup is the top-level loop for the format() function. It |
| 730 | searches through the format string for escapes to markup codes, and |
| 731 | calls other functions to move non-markup text to the output, |
| 732 | and to perform the markup to the output. |
| 733 | */ |
| 734 | static int |
| 735 | do_markup(SubString *input, PyObject *args, PyObject *kwargs, |
| 736 | OutputString *output, int *recursion_level) |
| 737 | { |
| 738 | MarkupIterator iter; |
| 739 | int is_markup; |
| 740 | int format_spec_needs_expanding; |
| 741 | int result; |
| 742 | SubString str; |
| 743 | SubString field_name; |
| 744 | SubString format_spec; |
| 745 | STRINGLIB_CHAR conversion; |
| 746 | |
| 747 | MarkupIterator_init(&iter, input->ptr, input->end - input->ptr); |
| 748 | while ((result = MarkupIterator_next(&iter, &is_markup, &str, &field_name, |
| 749 | &format_spec, &conversion, |
| 750 | &format_spec_needs_expanding)) == 2) { |
| 751 | if (is_markup) { |
| 752 | if (!output_markup(&field_name, &format_spec, |
| 753 | format_spec_needs_expanding, conversion, output, |
| 754 | args, kwargs, recursion_level)) |
| 755 | return 0; |
| 756 | } else { |
| 757 | if (!output_data(output, str.ptr, str.end-str.ptr)) |
| 758 | return 0; |
| 759 | } |
| 760 | } |
| 761 | return result; |
| 762 | } |
| 763 | |
| 764 | |
| 765 | /* |
| 766 | build_string allocates the output string and then |
| 767 | calls do_markup to do the heavy lifting. |
| 768 | */ |
| 769 | static PyObject * |
| 770 | build_string(SubString *input, PyObject *args, PyObject *kwargs, |
| 771 | int *recursion_level) |
| 772 | { |
| 773 | OutputString output; |
| 774 | PyObject *result = NULL; |
| 775 | Py_ssize_t count; |
| 776 | |
| 777 | output.obj = NULL; /* needed so cleanup code always works */ |
| 778 | |
| 779 | /* check the recursion level */ |
| 780 | (*recursion_level)--; |
| 781 | if (*recursion_level < 0) { |
| 782 | PyErr_SetString(PyExc_ValueError, |
| 783 | "Max string recursion exceeded"); |
| 784 | goto done; |
| 785 | } |
| 786 | |
| 787 | /* initial size is the length of the format string, plus the size |
| 788 | increment. seems like a reasonable default */ |
| 789 | if (!output_initialize(&output, |
| 790 | input->end - input->ptr + |
| 791 | INITIAL_SIZE_INCREMENT)) |
| 792 | goto done; |
| 793 | |
| 794 | if (!do_markup(input, args, kwargs, &output, recursion_level)) { |
| 795 | goto done; |
| 796 | } |
| 797 | |
| 798 | count = output.ptr - STRINGLIB_STR(output.obj); |
| 799 | if (STRINGLIB_RESIZE(&output.obj, count) < 0) { |
| 800 | goto done; |
| 801 | } |
| 802 | |
| 803 | /* transfer ownership to result */ |
| 804 | result = output.obj; |
| 805 | output.obj = NULL; |
| 806 | |
| 807 | done: |
| 808 | (*recursion_level)++; |
| 809 | Py_XDECREF(output.obj); |
| 810 | return result; |
| 811 | } |
| 812 | |
| 813 | /************************************************************************/ |
| 814 | /*********** main routine ***********************************************/ |
| 815 | /************************************************************************/ |
| 816 | |
| 817 | /* this is the main entry point */ |
| 818 | static PyObject * |
| 819 | do_string_format(PyObject *self, PyObject *args, PyObject *kwargs) |
| 820 | { |
| 821 | SubString input; |
| 822 | |
| 823 | /* PEP 3101 says only 2 levels, so that |
| 824 | "{0:{1}}".format('abc', 's') # works |
| 825 | "{0:{1:{2}}}".format('abc', 's', '') # fails |
| 826 | */ |
| 827 | int recursion_level = 2; |
| 828 | |
| 829 | SubString_init(&input, STRINGLIB_STR(self), STRINGLIB_LEN(self)); |
| 830 | return build_string(&input, args, kwargs, &recursion_level); |
| 831 | } |