Home | History | Annotate | Download | only in stringlib
      1 /* implements the string, long, and float formatters.  that is,
      2    string.__format__, etc. */
      3 
      4 #include <locale.h>
      5 
      6 /* Before including this, you must include either:
      7    stringlib/unicodedefs.h
      8    stringlib/stringdefs.h
      9 
     10    Also, you should define the names:
     11    FORMAT_STRING
     12    FORMAT_LONG
     13    FORMAT_FLOAT
     14    FORMAT_COMPLEX
     15    to be whatever you want the public names of these functions to
     16    be.  These are the only non-static functions defined here.
     17 */
     18 
     19 /* Raises an exception about an unknown presentation type for this
     20  * type. */
     21 
     22 static void
     23 unknown_presentation_type(STRINGLIB_CHAR presentation_type,
     24                           const char* type_name)
     25 {
     26 #if STRINGLIB_IS_UNICODE
     27     /* If STRINGLIB_CHAR is Py_UNICODE, %c might be out-of-range,
     28        hence the two cases. If it is char, gcc complains that the
     29        condition below is always true, hence the ifdef. */
     30     if (presentation_type > 32 && presentation_type < 128)
     31 #endif
     32         PyErr_Format(PyExc_ValueError,
     33                      "Unknown format code '%c' "
     34                      "for object of type '%.200s'",
     35                      (char)presentation_type,
     36                      type_name);
     37 #if STRINGLIB_IS_UNICODE
     38     else
     39         PyErr_Format(PyExc_ValueError,
     40                      "Unknown format code '\\x%x' "
     41                      "for object of type '%.200s'",
     42                      (unsigned int)presentation_type,
     43                      type_name);
     44 #endif
     45 }
     46 
     47 static void
     48 invalid_comma_type(STRINGLIB_CHAR presentation_type)
     49 {
     50 #if STRINGLIB_IS_UNICODE
     51     /* See comment in unknown_presentation_type */
     52     if (presentation_type > 32 && presentation_type < 128)
     53 #endif
     54         PyErr_Format(PyExc_ValueError,
     55                      "Cannot specify ',' with '%c'.",
     56                      (char)presentation_type);
     57 #if STRINGLIB_IS_UNICODE
     58     else
     59         PyErr_Format(PyExc_ValueError,
     60                      "Cannot specify ',' with '\\x%x'.",
     61                      (unsigned int)presentation_type);
     62 #endif
     63 }
     64 
     65 /*
     66     get_integer consumes 0 or more decimal digit characters from an
     67     input string, updates *result with the corresponding positive
     68     integer, and returns the number of digits consumed.
     69 
     70     returns -1 on error.
     71 */
     72 static int
     73 get_integer(STRINGLIB_CHAR **ptr, STRINGLIB_CHAR *end,
     74                   Py_ssize_t *result)
     75 {
     76     Py_ssize_t accumulator, digitval;
     77     int numdigits;
     78     accumulator = numdigits = 0;
     79     for (;;(*ptr)++, numdigits++) {
     80         if (*ptr >= end)
     81             break;
     82         digitval = STRINGLIB_TODECIMAL(**ptr);
     83         if (digitval < 0)
     84             break;
     85         /*
     86            Detect possible overflow before it happens:
     87 
     88               accumulator * 10 + digitval > PY_SSIZE_T_MAX if and only if
     89               accumulator > (PY_SSIZE_T_MAX - digitval) / 10.
     90         */
     91         if (accumulator > (PY_SSIZE_T_MAX - digitval) / 10) {
     92             PyErr_Format(PyExc_ValueError,
     93                          "Too many decimal digits in format string");
     94             return -1;
     95         }
     96         accumulator = accumulator * 10 + digitval;
     97     }
     98     *result = accumulator;
     99     return numdigits;
    100 }
    101 
    102 /************************************************************************/
    103 /*********** standard format specifier parsing **************************/
    104 /************************************************************************/
    105 
    106 /* returns true if this character is a specifier alignment token */
    107 Py_LOCAL_INLINE(int)
    108 is_alignment_token(STRINGLIB_CHAR c)
    109 {
    110     switch (c) {
    111     case '<': case '>': case '=': case '^':
    112         return 1;
    113     default:
    114         return 0;
    115     }
    116 }
    117 
    118 /* returns true if this character is a sign element */
    119 Py_LOCAL_INLINE(int)
    120 is_sign_element(STRINGLIB_CHAR c)
    121 {
    122     switch (c) {
    123     case ' ': case '+': case '-':
    124         return 1;
    125     default:
    126         return 0;
    127     }
    128 }
    129 
    130 
    131 typedef struct {
    132     STRINGLIB_CHAR fill_char;
    133     STRINGLIB_CHAR align;
    134     int alternate;
    135     STRINGLIB_CHAR sign;
    136     Py_ssize_t width;
    137     int thousands_separators;
    138     Py_ssize_t precision;
    139     STRINGLIB_CHAR type;
    140 } InternalFormatSpec;
    141 
    142 
    143 #if 0
    144 /* Occasionally useful for debugging. Should normally be commented out. */
    145 static void
    146 DEBUG_PRINT_FORMAT_SPEC(InternalFormatSpec *format)
    147 {
    148     printf("internal format spec: fill_char %d\n", format->fill_char);
    149     printf("internal format spec: align %d\n", format->align);
    150     printf("internal format spec: alternate %d\n", format->alternate);
    151     printf("internal format spec: sign %d\n", format->sign);
    152     printf("internal format spec: width %zd\n", format->width);
    153     printf("internal format spec: thousands_separators %d\n",
    154            format->thousands_separators);
    155     printf("internal format spec: precision %zd\n", format->precision);
    156     printf("internal format spec: type %c\n", format->type);
    157     printf("\n");
    158 }
    159 #endif
    160 
    161 
    162 /*
    163   ptr points to the start of the format_spec, end points just past its end.
    164   fills in format with the parsed information.
    165   returns 1 on success, 0 on failure.
    166   if failure, sets the exception
    167 */
    168 static int
    169 parse_internal_render_format_spec(STRINGLIB_CHAR *format_spec,
    170                                   Py_ssize_t format_spec_len,
    171                                   InternalFormatSpec *format,
    172                                   char default_type,
    173                                   char default_align)
    174 {
    175     STRINGLIB_CHAR *ptr = format_spec;
    176     STRINGLIB_CHAR *end = format_spec + format_spec_len;
    177 
    178     /* end-ptr is used throughout this code to specify the length of
    179        the input string */
    180 
    181     Py_ssize_t consumed;
    182     int align_specified = 0;
    183     int fill_char_specified = 0;
    184 
    185     format->fill_char = ' ';
    186     format->align = default_align;
    187     format->alternate = 0;
    188     format->sign = '\0';
    189     format->width = -1;
    190     format->thousands_separators = 0;
    191     format->precision = -1;
    192     format->type = default_type;
    193 
    194     /* If the second char is an alignment token,
    195        then parse the fill char */
    196     if (end-ptr >= 2 && is_alignment_token(ptr[1])) {
    197         format->align = ptr[1];
    198         format->fill_char = ptr[0];
    199         fill_char_specified = 1;
    200         align_specified = 1;
    201         ptr += 2;
    202     }
    203     else if (end-ptr >= 1 && is_alignment_token(ptr[0])) {
    204         format->align = ptr[0];
    205         align_specified = 1;
    206         ++ptr;
    207     }
    208 
    209     /* Parse the various sign options */
    210     if (end-ptr >= 1 && is_sign_element(ptr[0])) {
    211         format->sign = ptr[0];
    212         ++ptr;
    213     }
    214 
    215     /* If the next character is #, we're in alternate mode.  This only
    216        applies to integers. */
    217     if (end-ptr >= 1 && ptr[0] == '#') {
    218         format->alternate = 1;
    219         ++ptr;
    220     }
    221 
    222     /* The special case for 0-padding (backwards compat) */
    223     if (!fill_char_specified && end-ptr >= 1 && ptr[0] == '0') {
    224         format->fill_char = '0';
    225         if (!align_specified) {
    226             format->align = '=';
    227         }
    228         ++ptr;
    229     }
    230 
    231     consumed = get_integer(&ptr, end, &format->width);
    232     if (consumed == -1)
    233         /* Overflow error. Exception already set. */
    234         return 0;
    235 
    236     /* If consumed is 0, we didn't consume any characters for the
    237        width. In that case, reset the width to -1, because
    238        get_integer() will have set it to zero. -1 is how we record
    239        that the width wasn't specified. */
    240     if (consumed == 0)
    241         format->width = -1;
    242 
    243     /* Comma signifies add thousands separators */
    244     if (end-ptr && ptr[0] == ',') {
    245         format->thousands_separators = 1;
    246         ++ptr;
    247     }
    248 
    249     /* Parse field precision */
    250     if (end-ptr && ptr[0] == '.') {
    251         ++ptr;
    252 
    253         consumed = get_integer(&ptr, end, &format->precision);
    254         if (consumed == -1)
    255             /* Overflow error. Exception already set. */
    256             return 0;
    257 
    258         /* Not having a precision after a dot is an error. */
    259         if (consumed == 0) {
    260             PyErr_Format(PyExc_ValueError,
    261                          "Format specifier missing precision");
    262             return 0;
    263         }
    264 
    265     }
    266 
    267     /* Finally, parse the type field. */
    268 
    269     if (end-ptr > 1) {
    270         /* More than one char remain, invalid conversion spec. */
    271         PyErr_Format(PyExc_ValueError, "Invalid conversion specification");
    272         return 0;
    273     }
    274 
    275     if (end-ptr == 1) {
    276         format->type = ptr[0];
    277         ++ptr;
    278     }
    279 
    280     /* Do as much validating as we can, just by looking at the format
    281        specifier.  Do not take into account what type of formatting
    282        we're doing (int, float, string). */
    283 
    284     if (format->thousands_separators) {
    285         switch (format->type) {
    286         case 'd':
    287         case 'e':
    288         case 'f':
    289         case 'g':
    290         case 'E':
    291         case 'G':
    292         case '%':
    293         case 'F':
    294         case '\0':
    295             /* These are allowed. See PEP 378.*/
    296             break;
    297         default:
    298             invalid_comma_type(format->type);
    299             return 0;
    300         }
    301     }
    302 
    303     return 1;
    304 }
    305 
    306 /* Calculate the padding needed. */
    307 static void
    308 calc_padding(Py_ssize_t nchars, Py_ssize_t width, STRINGLIB_CHAR align,
    309              Py_ssize_t *n_lpadding, Py_ssize_t *n_rpadding,
    310              Py_ssize_t *n_total)
    311 {
    312     if (width >= 0) {
    313         if (nchars > width)
    314             *n_total = nchars;
    315         else
    316             *n_total = width;
    317     }
    318     else {
    319         /* not specified, use all of the chars and no more */
    320         *n_total = nchars;
    321     }
    322 
    323     /* Figure out how much leading space we need, based on the
    324        aligning */
    325     if (align == '>')
    326         *n_lpadding = *n_total - nchars;
    327     else if (align == '^')
    328         *n_lpadding = (*n_total - nchars) / 2;
    329     else if (align == '<' || align == '=')
    330         *n_lpadding = 0;
    331     else {
    332         /* We should never have an unspecified alignment. */
    333         *n_lpadding = 0;
    334         assert(0);
    335     }
    336 
    337     *n_rpadding = *n_total - nchars - *n_lpadding;
    338 }
    339 
    340 /* Do the padding, and return a pointer to where the caller-supplied
    341    content goes. */
    342 static STRINGLIB_CHAR *
    343 fill_padding(STRINGLIB_CHAR *p, Py_ssize_t nchars, STRINGLIB_CHAR fill_char,
    344              Py_ssize_t n_lpadding, Py_ssize_t n_rpadding)
    345 {
    346     /* Pad on left. */
    347     if (n_lpadding)
    348         STRINGLIB_FILL(p, fill_char, n_lpadding);
    349 
    350     /* Pad on right. */
    351     if (n_rpadding)
    352         STRINGLIB_FILL(p + nchars + n_lpadding, fill_char, n_rpadding);
    353 
    354     /* Pointer to the user content. */
    355     return p + n_lpadding;
    356 }
    357 
    358 #if defined FORMAT_FLOAT || defined FORMAT_LONG || defined FORMAT_COMPLEX
    359 /************************************************************************/
    360 /*********** common routines for numeric formatting *********************/
    361 /************************************************************************/
    362 
    363 /* Locale type codes. */
    364 #define LT_CURRENT_LOCALE 0
    365 #define LT_DEFAULT_LOCALE 1
    366 #define LT_NO_LOCALE 2
    367 
    368 /* Locale info needed for formatting integers and the part of floats
    369    before and including the decimal. Note that locales only support
    370    8-bit chars, not unicode. */
    371 typedef struct {
    372     char *decimal_point;
    373     char *thousands_sep;
    374     char *grouping;
    375 } LocaleInfo;
    376 
    377 /* describes the layout for an integer, see the comment in
    378    calc_number_widths() for details */
    379 typedef struct {
    380     Py_ssize_t n_lpadding;
    381     Py_ssize_t n_prefix;
    382     Py_ssize_t n_spadding;
    383     Py_ssize_t n_rpadding;
    384     char sign;
    385     Py_ssize_t n_sign;      /* number of digits needed for sign (0/1) */
    386     Py_ssize_t n_grouped_digits; /* Space taken up by the digits, including
    387                                     any grouping chars. */
    388     Py_ssize_t n_decimal;   /* 0 if only an integer */
    389     Py_ssize_t n_remainder; /* Digits in decimal and/or exponent part,
    390                                excluding the decimal itself, if
    391                                present. */
    392 
    393     /* These 2 are not the widths of fields, but are needed by
    394        STRINGLIB_GROUPING. */
    395     Py_ssize_t n_digits;    /* The number of digits before a decimal
    396                                or exponent. */
    397     Py_ssize_t n_min_width; /* The min_width we used when we computed
    398                                the n_grouped_digits width. */
    399 } NumberFieldWidths;
    400 
    401 
    402 /* Given a number of the form:
    403    digits[remainder]
    404    where ptr points to the start and end points to the end, find where
    405     the integer part ends. This could be a decimal, an exponent, both,
    406     or neither.
    407    If a decimal point is present, set *has_decimal and increment
    408     remainder beyond it.
    409    Results are undefined (but shouldn't crash) for improperly
    410     formatted strings.
    411 */
    412 static void
    413 parse_number(STRINGLIB_CHAR *ptr, Py_ssize_t len,
    414              Py_ssize_t *n_remainder, int *has_decimal)
    415 {
    416     STRINGLIB_CHAR *end = ptr + len;
    417     STRINGLIB_CHAR *remainder;
    418 
    419     while (ptr<end && isdigit(*ptr))
    420         ++ptr;
    421     remainder = ptr;
    422 
    423     /* Does remainder start with a decimal point? */
    424     *has_decimal = ptr<end && *remainder == '.';
    425 
    426     /* Skip the decimal point. */
    427     if (*has_decimal)
    428         remainder++;
    429 
    430     *n_remainder = end - remainder;
    431 }
    432 
    433 /* not all fields of format are used.  for example, precision is
    434    unused.  should this take discrete params in order to be more clear
    435    about what it does?  or is passing a single format parameter easier
    436    and more efficient enough to justify a little obfuscation? */
    437 static Py_ssize_t
    438 calc_number_widths(NumberFieldWidths *spec, Py_ssize_t n_prefix,
    439                    STRINGLIB_CHAR sign_char, STRINGLIB_CHAR *number,
    440                    Py_ssize_t n_number, Py_ssize_t n_remainder,
    441                    int has_decimal, const LocaleInfo *locale,
    442                    const InternalFormatSpec *format)
    443 {
    444     Py_ssize_t n_non_digit_non_padding;
    445     Py_ssize_t n_padding;
    446 
    447     spec->n_digits = n_number - n_remainder - (has_decimal?1:0);
    448     spec->n_lpadding = 0;
    449     spec->n_prefix = n_prefix;
    450     spec->n_decimal = has_decimal ? strlen(locale->decimal_point) : 0;
    451     spec->n_remainder = n_remainder;
    452     spec->n_spadding = 0;
    453     spec->n_rpadding = 0;
    454     spec->sign = '\0';
    455     spec->n_sign = 0;
    456 
    457     /* the output will look like:
    458        |                                                                                         |
    459        | <lpadding> <sign> <prefix> <spadding> <grouped_digits> <decimal> <remainder> <rpadding> |
    460        |                                                                                         |
    461 
    462        sign is computed from format->sign and the actual
    463        sign of the number
    464 
    465        prefix is given (it's for the '0x' prefix)
    466 
    467        digits is already known
    468 
    469        the total width is either given, or computed from the
    470        actual digits
    471 
    472        only one of lpadding, spadding, and rpadding can be non-zero,
    473        and it's calculated from the width and other fields
    474     */
    475 
    476     /* compute the various parts we're going to write */
    477     switch (format->sign) {
    478     case '+':
    479         /* always put a + or - */
    480         spec->n_sign = 1;
    481         spec->sign = (sign_char == '-' ? '-' : '+');
    482         break;
    483     case ' ':
    484         spec->n_sign = 1;
    485         spec->sign = (sign_char == '-' ? '-' : ' ');
    486         break;
    487     default:
    488         /* Not specified, or the default (-) */
    489         if (sign_char == '-') {
    490             spec->n_sign = 1;
    491             spec->sign = '-';
    492         }
    493     }
    494 
    495     /* The number of chars used for non-digits and non-padding. */
    496     n_non_digit_non_padding = spec->n_sign + spec->n_prefix + spec->n_decimal +
    497         spec->n_remainder;
    498 
    499     /* min_width can go negative, that's okay. format->width == -1 means
    500        we don't care. */
    501     if (format->fill_char == '0' && format->align == '=')
    502         spec->n_min_width = format->width - n_non_digit_non_padding;
    503     else
    504         spec->n_min_width = 0;
    505 
    506     if (spec->n_digits == 0)
    507         /* This case only occurs when using 'c' formatting, we need
    508            to special case it because the grouping code always wants
    509            to have at least one character. */
    510         spec->n_grouped_digits = 0;
    511     else
    512         spec->n_grouped_digits = STRINGLIB_GROUPING(NULL, 0, NULL,
    513                                                     spec->n_digits,
    514                                                     spec->n_min_width,
    515                                                     locale->grouping,
    516                                                     locale->thousands_sep);
    517 
    518     /* Given the desired width and the total of digit and non-digit
    519        space we consume, see if we need any padding. format->width can
    520        be negative (meaning no padding), but this code still works in
    521        that case. */
    522     n_padding = format->width -
    523                         (n_non_digit_non_padding + spec->n_grouped_digits);
    524     if (n_padding > 0) {
    525         /* Some padding is needed. Determine if it's left, space, or right. */
    526         switch (format->align) {
    527         case '<':
    528             spec->n_rpadding = n_padding;
    529             break;
    530         case '^':
    531             spec->n_lpadding = n_padding / 2;
    532             spec->n_rpadding = n_padding - spec->n_lpadding;
    533             break;
    534         case '=':
    535             spec->n_spadding = n_padding;
    536             break;
    537         case '>':
    538             spec->n_lpadding = n_padding;
    539             break;
    540         default:
    541             /* Shouldn't get here, but treat it as '>' */
    542             spec->n_lpadding = n_padding;
    543             assert(0);
    544             break;
    545         }
    546     }
    547     return spec->n_lpadding + spec->n_sign + spec->n_prefix +
    548         spec->n_spadding + spec->n_grouped_digits + spec->n_decimal +
    549         spec->n_remainder + spec->n_rpadding;
    550 }
    551 
    552 /* Fill in the digit parts of a numbers's string representation,
    553    as determined in calc_number_widths().
    554    No error checking, since we know the buffer is the correct size. */
    555 static void
    556 fill_number(STRINGLIB_CHAR *buf, const NumberFieldWidths *spec,
    557             STRINGLIB_CHAR *digits, Py_ssize_t n_digits,
    558             STRINGLIB_CHAR *prefix, STRINGLIB_CHAR fill_char,
    559             LocaleInfo *locale, int toupper)
    560 {
    561     /* Used to keep track of digits, decimal, and remainder. */
    562     STRINGLIB_CHAR *p = digits;
    563 
    564 #ifndef NDEBUG
    565     Py_ssize_t r;
    566 #endif
    567 
    568     if (spec->n_lpadding) {
    569         STRINGLIB_FILL(buf, fill_char, spec->n_lpadding);
    570         buf += spec->n_lpadding;
    571     }
    572     if (spec->n_sign == 1) {
    573         *buf++ = spec->sign;
    574     }
    575     if (spec->n_prefix) {
    576         memmove(buf,
    577                 prefix,
    578                 spec->n_prefix * sizeof(STRINGLIB_CHAR));
    579         if (toupper) {
    580             Py_ssize_t t;
    581             for (t = 0; t < spec->n_prefix; ++t)
    582                 buf[t] = STRINGLIB_TOUPPER(buf[t]);
    583         }
    584         buf += spec->n_prefix;
    585     }
    586     if (spec->n_spadding) {
    587         STRINGLIB_FILL(buf, fill_char, spec->n_spadding);
    588         buf += spec->n_spadding;
    589     }
    590 
    591     /* Only for type 'c' special case, it has no digits. */
    592     if (spec->n_digits != 0) {
    593         /* Fill the digits with InsertThousandsGrouping. */
    594 #ifndef NDEBUG
    595         r =
    596 #endif
    597             STRINGLIB_GROUPING(buf, spec->n_grouped_digits, digits,
    598                                spec->n_digits, spec->n_min_width,
    599                                locale->grouping, locale->thousands_sep);
    600 #ifndef NDEBUG
    601         assert(r == spec->n_grouped_digits);
    602 #endif
    603         p += spec->n_digits;
    604     }
    605     if (toupper) {
    606         Py_ssize_t t;
    607         for (t = 0; t < spec->n_grouped_digits; ++t)
    608             buf[t] = STRINGLIB_TOUPPER(buf[t]);
    609     }
    610     buf += spec->n_grouped_digits;
    611 
    612     if (spec->n_decimal) {
    613         Py_ssize_t t;
    614         for (t = 0; t < spec->n_decimal; ++t)
    615             buf[t] = locale->decimal_point[t];
    616         buf += spec->n_decimal;
    617         p += 1;
    618     }
    619 
    620     if (spec->n_remainder) {
    621         memcpy(buf, p, spec->n_remainder * sizeof(STRINGLIB_CHAR));
    622         buf += spec->n_remainder;
    623         p += spec->n_remainder;
    624     }
    625 
    626     if (spec->n_rpadding) {
    627         STRINGLIB_FILL(buf, fill_char, spec->n_rpadding);
    628         buf += spec->n_rpadding;
    629     }
    630 }
    631 
    632 static char no_grouping[1] = {CHAR_MAX};
    633 
    634 /* Find the decimal point character(s?), thousands_separator(s?), and
    635    grouping description, either for the current locale if type is
    636    LT_CURRENT_LOCALE, a hard-coded locale if LT_DEFAULT_LOCALE, or
    637    none if LT_NO_LOCALE. */
    638 static void
    639 get_locale_info(int type, LocaleInfo *locale_info)
    640 {
    641     switch (type) {
    642     case LT_CURRENT_LOCALE: {
    643         struct lconv *locale_data = localeconv();
    644         locale_info->decimal_point = locale_data->decimal_point;
    645         locale_info->thousands_sep = locale_data->thousands_sep;
    646         locale_info->grouping = locale_data->grouping;
    647         break;
    648     }
    649     case LT_DEFAULT_LOCALE:
    650         locale_info->decimal_point = ".";
    651         locale_info->thousands_sep = ",";
    652         locale_info->grouping = "\3"; /* Group every 3 characters.  The
    653                                          (implicit) trailing 0 means repeat
    654                                          infinitely. */
    655         break;
    656     case LT_NO_LOCALE:
    657         locale_info->decimal_point = ".";
    658         locale_info->thousands_sep = "";
    659         locale_info->grouping = no_grouping;
    660         break;
    661     default:
    662         assert(0);
    663     }
    664 }
    665 
    666 #endif /* FORMAT_FLOAT || FORMAT_LONG || FORMAT_COMPLEX */
    667 
    668 /************************************************************************/
    669 /*********** string formatting ******************************************/
    670 /************************************************************************/
    671 
    672 static PyObject *
    673 format_string_internal(PyObject *value, const InternalFormatSpec *format)
    674 {
    675     Py_ssize_t lpad;
    676     Py_ssize_t rpad;
    677     Py_ssize_t total;
    678     STRINGLIB_CHAR *p;
    679     Py_ssize_t len = STRINGLIB_LEN(value);
    680     PyObject *result = NULL;
    681 
    682     /* sign is not allowed on strings */
    683     if (format->sign != '\0') {
    684         PyErr_SetString(PyExc_ValueError,
    685                         "Sign not allowed in string format specifier");
    686         goto done;
    687     }
    688 
    689     /* alternate is not allowed on strings */
    690     if (format->alternate) {
    691         PyErr_SetString(PyExc_ValueError,
    692                         "Alternate form (#) not allowed in string format "
    693                         "specifier");
    694         goto done;
    695     }
    696 
    697     /* '=' alignment not allowed on strings */
    698     if (format->align == '=') {
    699         PyErr_SetString(PyExc_ValueError,
    700                         "'=' alignment not allowed "
    701                         "in string format specifier");
    702         goto done;
    703     }
    704 
    705     /* if precision is specified, output no more that format.precision
    706        characters */
    707     if (format->precision >= 0 && len >= format->precision) {
    708         len = format->precision;
    709     }
    710 
    711     calc_padding(len, format->width, format->align, &lpad, &rpad, &total);
    712 
    713     /* allocate the resulting string */
    714     result = STRINGLIB_NEW(NULL, total);
    715     if (result == NULL)
    716         goto done;
    717 
    718     /* Write into that space. First the padding. */
    719     p = fill_padding(STRINGLIB_STR(result), len,
    720                      format->fill_char, lpad, rpad);
    721 
    722     /* Then the source string. */
    723     memcpy(p, STRINGLIB_STR(value), len * sizeof(STRINGLIB_CHAR));
    724 
    725 done:
    726     return result;
    727 }
    728 
    729 
    730 /************************************************************************/
    731 /*********** long formatting ********************************************/
    732 /************************************************************************/
    733 
    734 #if defined FORMAT_LONG || defined FORMAT_INT
    735 typedef PyObject*
    736 (*IntOrLongToString)(PyObject *value, int base);
    737 
    738 static PyObject *
    739 format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format,
    740                             IntOrLongToString tostring)
    741 {
    742     PyObject *result = NULL;
    743     PyObject *tmp = NULL;
    744     STRINGLIB_CHAR *pnumeric_chars;
    745     STRINGLIB_CHAR numeric_char;
    746     STRINGLIB_CHAR sign_char = '\0';
    747     Py_ssize_t n_digits;       /* count of digits need from the computed
    748                                   string */
    749     Py_ssize_t n_remainder = 0; /* Used only for 'c' formatting, which
    750                                    produces non-digits */
    751     Py_ssize_t n_prefix = 0;   /* Count of prefix chars, (e.g., '0x') */
    752     Py_ssize_t n_total;
    753     STRINGLIB_CHAR *prefix = NULL;
    754     NumberFieldWidths spec;
    755     long x;
    756 
    757     /* Locale settings, either from the actual locale or
    758        from a hard-code pseudo-locale */
    759     LocaleInfo locale;
    760 
    761     /* no precision allowed on integers */
    762     if (format->precision != -1) {
    763         PyErr_SetString(PyExc_ValueError,
    764                         "Precision not allowed in integer format specifier");
    765         goto done;
    766     }
    767 
    768     /* special case for character formatting */
    769     if (format->type == 'c') {
    770         /* error to specify a sign */
    771         if (format->sign != '\0') {
    772             PyErr_SetString(PyExc_ValueError,
    773                             "Sign not allowed with integer"
    774                             " format specifier 'c'");
    775             goto done;
    776         }
    777 
    778         /* Error to specify a comma. */
    779         if (format->thousands_separators) {
    780             PyErr_SetString(PyExc_ValueError,
    781                             "Thousands separators not allowed with integer"
    782                             " format specifier 'c'");
    783             goto done;
    784         }
    785 
    786         /* taken from unicodeobject.c formatchar() */
    787         /* Integer input truncated to a character */
    788 /* XXX: won't work for int */
    789         x = PyLong_AsLong(value);
    790         if (x == -1 && PyErr_Occurred())
    791             goto done;
    792 #if STRINGLIB_IS_UNICODE
    793 #ifdef Py_UNICODE_WIDE
    794         if (x < 0 || x > 0x10ffff) {
    795             PyErr_SetString(PyExc_OverflowError,
    796                             "%c arg not in range(0x110000) "
    797                             "(wide Python build)");
    798             goto done;
    799         }
    800 #else
    801         if (x < 0 || x > 0xffff) {
    802             PyErr_SetString(PyExc_OverflowError,
    803                             "%c arg not in range(0x10000) "
    804                             "(narrow Python build)");
    805             goto done;
    806         }
    807 #endif
    808 #else
    809         if (x < 0 || x > 0xff) {
    810             PyErr_SetString(PyExc_OverflowError,
    811                             "%c arg not in range(0x100)");
    812             goto done;
    813         }
    814 #endif
    815         numeric_char = (STRINGLIB_CHAR)x;
    816         pnumeric_chars = &numeric_char;
    817         n_digits = 1;
    818 
    819         /* As a sort-of hack, we tell calc_number_widths that we only
    820            have "remainder" characters. calc_number_widths thinks
    821            these are characters that don't get formatted, only copied
    822            into the output string. We do this for 'c' formatting,
    823            because the characters are likely to be non-digits. */
    824         n_remainder = 1;
    825     }
    826     else {
    827         int base;
    828         int leading_chars_to_skip = 0;  /* Number of characters added by
    829                                            PyNumber_ToBase that we want to
    830                                            skip over. */
    831 
    832         /* Compute the base and how many characters will be added by
    833            PyNumber_ToBase */
    834         switch (format->type) {
    835         case 'b':
    836             base = 2;
    837             leading_chars_to_skip = 2; /* 0b */
    838             break;
    839         case 'o':
    840             base = 8;
    841             leading_chars_to_skip = 2; /* 0o */
    842             break;
    843         case 'x':
    844         case 'X':
    845             base = 16;
    846             leading_chars_to_skip = 2; /* 0x */
    847             break;
    848         default:  /* shouldn't be needed, but stops a compiler warning */
    849         case 'd':
    850         case 'n':
    851             base = 10;
    852             break;
    853         }
    854 
    855         /* The number of prefix chars is the same as the leading
    856            chars to skip */
    857         if (format->alternate)
    858             n_prefix = leading_chars_to_skip;
    859 
    860         /* Do the hard part, converting to a string in a given base */
    861         tmp = tostring(value, base);
    862         if (tmp == NULL)
    863             goto done;
    864 
    865         pnumeric_chars = STRINGLIB_STR(tmp);
    866         n_digits = STRINGLIB_LEN(tmp);
    867 
    868         prefix = pnumeric_chars;
    869 
    870         /* Remember not to modify what pnumeric_chars points to.  it
    871            might be interned.  Only modify it after we copy it into a
    872            newly allocated output buffer. */
    873 
    874         /* Is a sign character present in the output?  If so, remember it
    875            and skip it */
    876         if (pnumeric_chars[0] == '-') {
    877             sign_char = pnumeric_chars[0];
    878             ++prefix;
    879             ++leading_chars_to_skip;
    880         }
    881 
    882         /* Skip over the leading chars (0x, 0b, etc.) */
    883         n_digits -= leading_chars_to_skip;
    884         pnumeric_chars += leading_chars_to_skip;
    885     }
    886 
    887     /* Determine the grouping, separator, and decimal point, if any. */
    888     get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE :
    889                     (format->thousands_separators ?
    890                      LT_DEFAULT_LOCALE :
    891                      LT_NO_LOCALE),
    892                     &locale);
    893 
    894     /* Calculate how much memory we'll need. */
    895     n_total = calc_number_widths(&spec, n_prefix, sign_char, pnumeric_chars,
    896                        n_digits, n_remainder, 0, &locale, format);
    897 
    898     /* Allocate the memory. */
    899     result = STRINGLIB_NEW(NULL, n_total);
    900     if (!result)
    901         goto done;
    902 
    903     /* Populate the memory. */
    904     fill_number(STRINGLIB_STR(result), &spec, pnumeric_chars, n_digits,
    905                 prefix, format->fill_char, &locale, format->type == 'X');
    906 
    907 done:
    908     Py_XDECREF(tmp);
    909     return result;
    910 }
    911 #endif /* defined FORMAT_LONG || defined FORMAT_INT */
    912 
    913 /************************************************************************/
    914 /*********** float formatting *******************************************/
    915 /************************************************************************/
    916 
    917 #ifdef FORMAT_FLOAT
    918 #if STRINGLIB_IS_UNICODE
    919 static void
    920 strtounicode(Py_UNICODE *buffer, const char *charbuffer, Py_ssize_t len)
    921 {
    922     Py_ssize_t i;
    923     for (i = 0; i < len; ++i)
    924         buffer[i] = (Py_UNICODE)charbuffer[i];
    925 }
    926 #endif
    927 
    928 /* much of this is taken from unicodeobject.c */
    929 static PyObject *
    930 format_float_internal(PyObject *value,
    931                       const InternalFormatSpec *format)
    932 {
    933     char *buf = NULL;       /* buffer returned from PyOS_double_to_string */
    934     Py_ssize_t n_digits;
    935     Py_ssize_t n_remainder;
    936     Py_ssize_t n_total;
    937     int has_decimal;
    938     double val;
    939     Py_ssize_t precision;
    940     Py_ssize_t default_precision = 6;
    941     STRINGLIB_CHAR type = format->type;
    942     int add_pct = 0;
    943     STRINGLIB_CHAR *p;
    944     NumberFieldWidths spec;
    945     int flags = 0;
    946     PyObject *result = NULL;
    947     STRINGLIB_CHAR sign_char = '\0';
    948     int float_type; /* Used to see if we have a nan, inf, or regular float. */
    949 
    950 #if STRINGLIB_IS_UNICODE
    951     Py_UNICODE *unicode_tmp = NULL;
    952 #endif
    953 
    954     /* Locale settings, either from the actual locale or
    955        from a hard-code pseudo-locale */
    956     LocaleInfo locale;
    957 
    958     if (format->precision > INT_MAX) {
    959         PyErr_SetString(PyExc_ValueError, "precision too big");
    960         goto done;
    961     }
    962     precision = (int)format->precision;
    963 
    964     /* Alternate is not allowed on floats. */
    965     if (format->alternate) {
    966         PyErr_SetString(PyExc_ValueError,
    967                         "Alternate form (#) not allowed in float format "
    968                         "specifier");
    969         goto done;
    970     }
    971 
    972     if (type == '\0') {
    973         /* Omitted type specifier. This is like 'g' but with at least one
    974            digit after the decimal point, and different default precision.*/
    975         type = 'g';
    976         default_precision = PyFloat_STR_PRECISION;
    977         flags |= Py_DTSF_ADD_DOT_0;
    978     }
    979 
    980     if (type == 'n')
    981         /* 'n' is the same as 'g', except for the locale used to
    982            format the result. We take care of that later. */
    983         type = 'g';
    984 
    985     val = PyFloat_AsDouble(value);
    986     if (val == -1.0 && PyErr_Occurred())
    987         goto done;
    988 
    989     if (type == '%') {
    990         type = 'f';
    991         val *= 100;
    992         add_pct = 1;
    993     }
    994 
    995     if (precision < 0)
    996         precision = default_precision;
    997 
    998     /* Cast "type", because if we're in unicode we need to pass an
    999        8-bit char. This is safe, because we've restricted what "type"
   1000        can be. */
   1001     buf = PyOS_double_to_string(val, (char)type, precision, flags,
   1002                                 &float_type);
   1003     if (buf == NULL)
   1004         goto done;
   1005     n_digits = strlen(buf);
   1006 
   1007     if (add_pct) {
   1008         /* We know that buf has a trailing zero (since we just called
   1009            strlen() on it), and we don't use that fact any more. So we
   1010            can just write over the trailing zero. */
   1011         buf[n_digits] = '%';
   1012         n_digits += 1;
   1013     }
   1014 
   1015     /* Since there is no unicode version of PyOS_double_to_string,
   1016        just use the 8 bit version and then convert to unicode. */
   1017 #if STRINGLIB_IS_UNICODE
   1018     unicode_tmp = (Py_UNICODE*)PyMem_Malloc((n_digits)*sizeof(Py_UNICODE));
   1019     if (unicode_tmp == NULL) {
   1020         PyErr_NoMemory();
   1021         goto done;
   1022     }
   1023     strtounicode(unicode_tmp, buf, n_digits);
   1024     p = unicode_tmp;
   1025 #else
   1026     p = buf;
   1027 #endif
   1028 
   1029     /* Is a sign character present in the output?  If so, remember it
   1030        and skip it */
   1031     if (*p == '-') {
   1032         sign_char = *p;
   1033         ++p;
   1034         --n_digits;
   1035     }
   1036 
   1037     /* Determine if we have any "remainder" (after the digits, might include
   1038        decimal or exponent or both (or neither)) */
   1039     parse_number(p, n_digits, &n_remainder, &has_decimal);
   1040 
   1041     /* Determine the grouping, separator, and decimal point, if any. */
   1042     get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE :
   1043                     (format->thousands_separators ?
   1044                      LT_DEFAULT_LOCALE :
   1045                      LT_NO_LOCALE),
   1046                     &locale);
   1047 
   1048     /* Calculate how much memory we'll need. */
   1049     n_total = calc_number_widths(&spec, 0, sign_char, p, n_digits,
   1050                                  n_remainder, has_decimal, &locale, format);
   1051 
   1052     /* Allocate the memory. */
   1053     result = STRINGLIB_NEW(NULL, n_total);
   1054     if (result == NULL)
   1055         goto done;
   1056 
   1057     /* Populate the memory. */
   1058     fill_number(STRINGLIB_STR(result), &spec, p, n_digits, NULL,
   1059                 format->fill_char, &locale, 0);
   1060 
   1061 done:
   1062     PyMem_Free(buf);
   1063 #if STRINGLIB_IS_UNICODE
   1064     PyMem_Free(unicode_tmp);
   1065 #endif
   1066     return result;
   1067 }
   1068 #endif /* FORMAT_FLOAT */
   1069 
   1070 /************************************************************************/
   1071 /*********** complex formatting *****************************************/
   1072 /************************************************************************/
   1073 
   1074 #ifdef FORMAT_COMPLEX
   1075 
   1076 static PyObject *
   1077 format_complex_internal(PyObject *value,
   1078                         const InternalFormatSpec *format)
   1079 {
   1080     double re;
   1081     double im;
   1082     char *re_buf = NULL;       /* buffer returned from PyOS_double_to_string */
   1083     char *im_buf = NULL;       /* buffer returned from PyOS_double_to_string */
   1084 
   1085     InternalFormatSpec tmp_format = *format;
   1086     Py_ssize_t n_re_digits;
   1087     Py_ssize_t n_im_digits;
   1088     Py_ssize_t n_re_remainder;
   1089     Py_ssize_t n_im_remainder;
   1090     Py_ssize_t n_re_total;
   1091     Py_ssize_t n_im_total;
   1092     int re_has_decimal;
   1093     int im_has_decimal;
   1094     Py_ssize_t precision;
   1095     Py_ssize_t default_precision = 6;
   1096     STRINGLIB_CHAR type = format->type;
   1097     STRINGLIB_CHAR *p_re;
   1098     STRINGLIB_CHAR *p_im;
   1099     NumberFieldWidths re_spec;
   1100     NumberFieldWidths im_spec;
   1101     int flags = 0;
   1102     PyObject *result = NULL;
   1103     STRINGLIB_CHAR *p;
   1104     STRINGLIB_CHAR re_sign_char = '\0';
   1105     STRINGLIB_CHAR im_sign_char = '\0';
   1106     int re_float_type; /* Used to see if we have a nan, inf, or regular float. */
   1107     int im_float_type;
   1108     int add_parens = 0;
   1109     int skip_re = 0;
   1110     Py_ssize_t lpad;
   1111     Py_ssize_t rpad;
   1112     Py_ssize_t total;
   1113 
   1114 #if STRINGLIB_IS_UNICODE
   1115     Py_UNICODE *re_unicode_tmp = NULL;
   1116     Py_UNICODE *im_unicode_tmp = NULL;
   1117 #endif
   1118 
   1119     /* Locale settings, either from the actual locale or
   1120        from a hard-code pseudo-locale */
   1121     LocaleInfo locale;
   1122 
   1123     if (format->precision > INT_MAX) {
   1124         PyErr_SetString(PyExc_ValueError, "precision too big");
   1125         goto done;
   1126     }
   1127     precision = (int)format->precision;
   1128 
   1129     /* Alternate is not allowed on complex. */
   1130     if (format->alternate) {
   1131         PyErr_SetString(PyExc_ValueError,
   1132                         "Alternate form (#) not allowed in complex format "
   1133                         "specifier");
   1134         goto done;
   1135     }
   1136 
   1137     /* Neither is zero pading. */
   1138     if (format->fill_char == '0') {
   1139         PyErr_SetString(PyExc_ValueError,
   1140                         "Zero padding is not allowed in complex format "
   1141                         "specifier");
   1142         goto done;
   1143     }
   1144 
   1145     /* Neither is '=' alignment . */
   1146     if (format->align == '=') {
   1147         PyErr_SetString(PyExc_ValueError,
   1148                         "'=' alignment flag is not allowed in complex format "
   1149                         "specifier");
   1150         goto done;
   1151     }
   1152 
   1153     re = PyComplex_RealAsDouble(value);
   1154     if (re == -1.0 && PyErr_Occurred())
   1155         goto done;
   1156     im = PyComplex_ImagAsDouble(value);
   1157     if (im == -1.0 && PyErr_Occurred())
   1158         goto done;
   1159 
   1160     if (type == '\0') {
   1161         /* Omitted type specifier. Should be like str(self). */
   1162         type = 'g';
   1163         default_precision = PyFloat_STR_PRECISION;
   1164         if (re == 0.0 && copysign(1.0, re) == 1.0)
   1165             skip_re = 1;
   1166         else
   1167             add_parens = 1;
   1168     }
   1169 
   1170     if (type == 'n')
   1171         /* 'n' is the same as 'g', except for the locale used to
   1172            format the result. We take care of that later. */
   1173         type = 'g';
   1174 
   1175     if (precision < 0)
   1176         precision = default_precision;
   1177 
   1178     /* Cast "type", because if we're in unicode we need to pass an
   1179        8-bit char. This is safe, because we've restricted what "type"
   1180        can be. */
   1181     re_buf = PyOS_double_to_string(re, (char)type, precision, flags,
   1182                                    &re_float_type);
   1183     if (re_buf == NULL)
   1184         goto done;
   1185     im_buf = PyOS_double_to_string(im, (char)type, precision, flags,
   1186                                    &im_float_type);
   1187     if (im_buf == NULL)
   1188         goto done;
   1189 
   1190     n_re_digits = strlen(re_buf);
   1191     n_im_digits = strlen(im_buf);
   1192 
   1193     /* Since there is no unicode version of PyOS_double_to_string,
   1194        just use the 8 bit version and then convert to unicode. */
   1195 #if STRINGLIB_IS_UNICODE
   1196     re_unicode_tmp = (Py_UNICODE*)PyMem_Malloc((n_re_digits)*sizeof(Py_UNICODE));
   1197     if (re_unicode_tmp == NULL) {
   1198         PyErr_NoMemory();
   1199         goto done;
   1200     }
   1201     strtounicode(re_unicode_tmp, re_buf, n_re_digits);
   1202     p_re = re_unicode_tmp;
   1203 
   1204     im_unicode_tmp = (Py_UNICODE*)PyMem_Malloc((n_im_digits)*sizeof(Py_UNICODE));
   1205     if (im_unicode_tmp == NULL) {
   1206         PyErr_NoMemory();
   1207         goto done;
   1208     }
   1209     strtounicode(im_unicode_tmp, im_buf, n_im_digits);
   1210     p_im = im_unicode_tmp;
   1211 #else
   1212     p_re = re_buf;
   1213     p_im = im_buf;
   1214 #endif
   1215 
   1216     /* Is a sign character present in the output?  If so, remember it
   1217        and skip it */
   1218     if (*p_re == '-') {
   1219         re_sign_char = *p_re;
   1220         ++p_re;
   1221         --n_re_digits;
   1222     }
   1223     if (*p_im == '-') {
   1224         im_sign_char = *p_im;
   1225         ++p_im;
   1226         --n_im_digits;
   1227     }
   1228 
   1229     /* Determine if we have any "remainder" (after the digits, might include
   1230        decimal or exponent or both (or neither)) */
   1231     parse_number(p_re, n_re_digits, &n_re_remainder, &re_has_decimal);
   1232     parse_number(p_im, n_im_digits, &n_im_remainder, &im_has_decimal);
   1233 
   1234     /* Determine the grouping, separator, and decimal point, if any. */
   1235     get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE :
   1236                     (format->thousands_separators ?
   1237                      LT_DEFAULT_LOCALE :
   1238                      LT_NO_LOCALE),
   1239                     &locale);
   1240 
   1241     /* Turn off any padding. We'll do it later after we've composed
   1242        the numbers without padding. */
   1243     tmp_format.fill_char = '\0';
   1244     tmp_format.align = '<';
   1245     tmp_format.width = -1;
   1246 
   1247     /* Calculate how much memory we'll need. */
   1248     n_re_total = calc_number_widths(&re_spec, 0, re_sign_char, p_re,
   1249                                     n_re_digits, n_re_remainder,
   1250                                     re_has_decimal, &locale, &tmp_format);
   1251 
   1252     /* Same formatting, but always include a sign, unless the real part is
   1253      * going to be omitted, in which case we use whatever sign convention was
   1254      * requested by the original format. */
   1255     if (!skip_re)
   1256         tmp_format.sign = '+';
   1257     n_im_total = calc_number_widths(&im_spec, 0, im_sign_char, p_im,
   1258                                     n_im_digits, n_im_remainder,
   1259                                     im_has_decimal, &locale, &tmp_format);
   1260 
   1261     if (skip_re)
   1262         n_re_total = 0;
   1263 
   1264     /* Add 1 for the 'j', and optionally 2 for parens. */
   1265     calc_padding(n_re_total + n_im_total + 1 + add_parens * 2,
   1266                  format->width, format->align, &lpad, &rpad, &total);
   1267 
   1268     result = STRINGLIB_NEW(NULL, total);
   1269     if (result == NULL)
   1270         goto done;
   1271 
   1272     /* Populate the memory. First, the padding. */
   1273     p = fill_padding(STRINGLIB_STR(result),
   1274                      n_re_total + n_im_total + 1 + add_parens * 2,
   1275                      format->fill_char, lpad, rpad);
   1276 
   1277     if (add_parens)
   1278         *p++ = '(';
   1279 
   1280     if (!skip_re) {
   1281         fill_number(p, &re_spec, p_re, n_re_digits, NULL, 0, &locale, 0);
   1282         p += n_re_total;
   1283     }
   1284     fill_number(p, &im_spec, p_im, n_im_digits, NULL, 0, &locale, 0);
   1285     p += n_im_total;
   1286     *p++ = 'j';
   1287 
   1288     if (add_parens)
   1289         *p++ = ')';
   1290 
   1291 done:
   1292     PyMem_Free(re_buf);
   1293     PyMem_Free(im_buf);
   1294 #if STRINGLIB_IS_UNICODE
   1295     PyMem_Free(re_unicode_tmp);
   1296     PyMem_Free(im_unicode_tmp);
   1297 #endif
   1298     return result;
   1299 }
   1300 #endif /* FORMAT_COMPLEX */
   1301 
   1302 /************************************************************************/
   1303 /*********** built in formatters ****************************************/
   1304 /************************************************************************/
   1305 PyObject *
   1306 FORMAT_STRING(PyObject *obj,
   1307               STRINGLIB_CHAR *format_spec,
   1308               Py_ssize_t format_spec_len)
   1309 {
   1310     InternalFormatSpec format;
   1311     PyObject *result = NULL;
   1312 
   1313     /* check for the special case of zero length format spec, make
   1314        it equivalent to str(obj) */
   1315     if (format_spec_len == 0) {
   1316         result = STRINGLIB_TOSTR(obj);
   1317         goto done;
   1318     }
   1319 
   1320     /* parse the format_spec */
   1321     if (!parse_internal_render_format_spec(format_spec, format_spec_len,
   1322                                            &format, 's', '<'))
   1323         goto done;
   1324 
   1325     /* type conversion? */
   1326     switch (format.type) {
   1327     case 's':
   1328         /* no type conversion needed, already a string.  do the formatting */
   1329         result = format_string_internal(obj, &format);
   1330         break;
   1331     default:
   1332         /* unknown */
   1333         unknown_presentation_type(format.type, obj->ob_type->tp_name);
   1334         goto done;
   1335     }
   1336 
   1337 done:
   1338     return result;
   1339 }
   1340 
   1341 #if defined FORMAT_LONG || defined FORMAT_INT
   1342 static PyObject*
   1343 format_int_or_long(PyObject* obj,
   1344                    STRINGLIB_CHAR *format_spec,
   1345                    Py_ssize_t format_spec_len,
   1346                    IntOrLongToString tostring)
   1347 {
   1348     PyObject *result = NULL;
   1349     PyObject *tmp = NULL;
   1350     InternalFormatSpec format;
   1351 
   1352     /* check for the special case of zero length format spec, make
   1353        it equivalent to str(obj) */
   1354     if (format_spec_len == 0) {
   1355         result = STRINGLIB_TOSTR(obj);
   1356         goto done;
   1357     }
   1358 
   1359     /* parse the format_spec */
   1360     if (!parse_internal_render_format_spec(format_spec,
   1361                                            format_spec_len,
   1362                                            &format, 'd', '>'))
   1363         goto done;
   1364 
   1365     /* type conversion? */
   1366     switch (format.type) {
   1367     case 'b':
   1368     case 'c':
   1369     case 'd':
   1370     case 'o':
   1371     case 'x':
   1372     case 'X':
   1373     case 'n':
   1374         /* no type conversion needed, already an int (or long).  do
   1375            the formatting */
   1376             result = format_int_or_long_internal(obj, &format, tostring);
   1377         break;
   1378 
   1379     case 'e':
   1380     case 'E':
   1381     case 'f':
   1382     case 'F':
   1383     case 'g':
   1384     case 'G':
   1385     case '%':
   1386         /* convert to float */
   1387         tmp = PyNumber_Float(obj);
   1388         if (tmp == NULL)
   1389             goto done;
   1390         result = format_float_internal(tmp, &format);
   1391         break;
   1392 
   1393     default:
   1394         /* unknown */
   1395         unknown_presentation_type(format.type, obj->ob_type->tp_name);
   1396         goto done;
   1397     }
   1398 
   1399 done:
   1400     Py_XDECREF(tmp);
   1401     return result;
   1402 }
   1403 #endif /* FORMAT_LONG || defined FORMAT_INT */
   1404 
   1405 #ifdef FORMAT_LONG
   1406 /* Need to define long_format as a function that will convert a long
   1407    to a string.  In 3.0, _PyLong_Format has the correct signature.  In
   1408    2.x, we need to fudge a few parameters */
   1409 #if PY_VERSION_HEX >= 0x03000000
   1410 #define long_format _PyLong_Format
   1411 #else
   1412 static PyObject*
   1413 long_format(PyObject* value, int base)
   1414 {
   1415     /* Convert to base, don't add trailing 'L', and use the new octal
   1416        format. We already know this is a long object */
   1417     assert(PyLong_Check(value));
   1418     /* convert to base, don't add 'L', and use the new octal format */
   1419     return _PyLong_Format(value, base, 0, 1);
   1420 }
   1421 #endif
   1422 
   1423 PyObject *
   1424 FORMAT_LONG(PyObject *obj,
   1425             STRINGLIB_CHAR *format_spec,
   1426             Py_ssize_t format_spec_len)
   1427 {
   1428     return format_int_or_long(obj, format_spec, format_spec_len,
   1429                               long_format);
   1430 }
   1431 #endif /* FORMAT_LONG */
   1432 
   1433 #ifdef FORMAT_INT
   1434 /* this is only used for 2.x, not 3.0 */
   1435 static PyObject*
   1436 int_format(PyObject* value, int base)
   1437 {
   1438     /* Convert to base, and use the new octal format. We already
   1439        know this is an int object */
   1440     assert(PyInt_Check(value));
   1441     return _PyInt_Format((PyIntObject*)value, base, 1);
   1442 }
   1443 
   1444 PyObject *
   1445 FORMAT_INT(PyObject *obj,
   1446            STRINGLIB_CHAR *format_spec,
   1447            Py_ssize_t format_spec_len)
   1448 {
   1449     return format_int_or_long(obj, format_spec, format_spec_len,
   1450                               int_format);
   1451 }
   1452 #endif /* FORMAT_INT */
   1453 
   1454 #ifdef FORMAT_FLOAT
   1455 PyObject *
   1456 FORMAT_FLOAT(PyObject *obj,
   1457              STRINGLIB_CHAR *format_spec,
   1458              Py_ssize_t format_spec_len)
   1459 {
   1460     PyObject *result = NULL;
   1461     InternalFormatSpec format;
   1462 
   1463     /* check for the special case of zero length format spec, make
   1464        it equivalent to str(obj) */
   1465     if (format_spec_len == 0) {
   1466         result = STRINGLIB_TOSTR(obj);
   1467         goto done;
   1468     }
   1469 
   1470     /* parse the format_spec */
   1471     if (!parse_internal_render_format_spec(format_spec,
   1472                                            format_spec_len,
   1473                                            &format, '\0', '>'))
   1474         goto done;
   1475 
   1476     /* type conversion? */
   1477     switch (format.type) {
   1478     case '\0': /* No format code: like 'g', but with at least one decimal. */
   1479     case 'e':
   1480     case 'E':
   1481     case 'f':
   1482     case 'F':
   1483     case 'g':
   1484     case 'G':
   1485     case 'n':
   1486     case '%':
   1487         /* no conversion, already a float.  do the formatting */
   1488         result = format_float_internal(obj, &format);
   1489         break;
   1490 
   1491     default:
   1492         /* unknown */
   1493         unknown_presentation_type(format.type, obj->ob_type->tp_name);
   1494         goto done;
   1495     }
   1496 
   1497 done:
   1498     return result;
   1499 }
   1500 #endif /* FORMAT_FLOAT */
   1501 
   1502 #ifdef FORMAT_COMPLEX
   1503 PyObject *
   1504 FORMAT_COMPLEX(PyObject *obj,
   1505                STRINGLIB_CHAR *format_spec,
   1506                Py_ssize_t format_spec_len)
   1507 {
   1508     PyObject *result = NULL;
   1509     InternalFormatSpec format;
   1510 
   1511     /* check for the special case of zero length format spec, make
   1512        it equivalent to str(obj) */
   1513     if (format_spec_len == 0) {
   1514         result = STRINGLIB_TOSTR(obj);
   1515         goto done;
   1516     }
   1517 
   1518     /* parse the format_spec */
   1519     if (!parse_internal_render_format_spec(format_spec,
   1520                                            format_spec_len,
   1521                                            &format, '\0', '>'))
   1522         goto done;
   1523 
   1524     /* type conversion? */
   1525     switch (format.type) {
   1526     case '\0': /* No format code: like 'g', but with at least one decimal. */
   1527     case 'e':
   1528     case 'E':
   1529     case 'f':
   1530     case 'F':
   1531     case 'g':
   1532     case 'G':
   1533     case 'n':
   1534         /* no conversion, already a complex.  do the formatting */
   1535         result = format_complex_internal(obj, &format);
   1536         break;
   1537 
   1538     default:
   1539         /* unknown */
   1540         unknown_presentation_type(format.type, obj->ob_type->tp_name);
   1541         goto done;
   1542     }
   1543 
   1544 done:
   1545     return result;
   1546 }
   1547 #endif /* FORMAT_COMPLEX */
   1548