From d1d013c01c268d869597b35cbcd8b5d7c5baf2ae Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Martin=20v=2E=20L=C3=B6wis?= Date: Wed, 28 Sep 2011 07:41:54 +0200 Subject: Implement PEP 393. --- Python/formatter_unicode.c | 1445 +++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 1439 insertions(+), 6 deletions(-) (limited to 'Python/formatter_unicode.c') diff --git a/Python/formatter_unicode.c b/Python/formatter_unicode.c index c350907da1..9fb840b160 100644 --- a/Python/formatter_unicode.c +++ b/Python/formatter_unicode.c @@ -3,12 +3,1445 @@ of int.__float__, etc., that take and return unicode objects */ #include "Python.h" -#include "../Objects/stringlib/unicodedefs.h" +#include +/* Raises an exception about an unknown presentation type for this + * type. */ -#define FORMAT_STRING _PyUnicode_FormatAdvanced -#define FORMAT_LONG _PyLong_FormatAdvanced -#define FORMAT_FLOAT _PyFloat_FormatAdvanced -#define FORMAT_COMPLEX _PyComplex_FormatAdvanced +static void +unknown_presentation_type(Py_UCS4 presentation_type, + const char* type_name) +{ + /* %c might be out-of-range, hence the two cases. */ + if (presentation_type > 32 && presentation_type < 128) + PyErr_Format(PyExc_ValueError, + "Unknown format code '%c' " + "for object of type '%.200s'", + (char)presentation_type, + type_name); + else + PyErr_Format(PyExc_ValueError, + "Unknown format code '\\x%x' " + "for object of type '%.200s'", + (unsigned int)presentation_type, + type_name); +} -#include "../Objects/stringlib/formatter.h" +static void +invalid_comma_type(Py_UCS4 presentation_type) +{ + if (presentation_type > 32 && presentation_type < 128) + PyErr_Format(PyExc_ValueError, + "Cannot specify ',' with '%c'.", + (char)presentation_type); + else + PyErr_Format(PyExc_ValueError, + "Cannot specify ',' with '\\x%x'.", + (unsigned int)presentation_type); +} + +/* + get_integer consumes 0 or more decimal digit characters from an + input string, updates *result with the corresponding positive + integer, and returns the number of digits consumed. + + returns -1 on error. +*/ +static int +get_integer(PyObject *str, Py_ssize_t *pos, Py_ssize_t end, + Py_ssize_t *result) +{ + Py_ssize_t accumulator, digitval, oldaccumulator; + int numdigits; + accumulator = numdigits = 0; + for (;;(*pos)++, numdigits++) { + if (*pos >= end) + break; + digitval = Py_UNICODE_TODECIMAL(PyUnicode_READ_CHAR(str, *pos)); + if (digitval < 0) + break; + /* + This trick was copied from old Unicode format code. It's cute, + but would really suck on an old machine with a slow divide + implementation. Fortunately, in the normal case we do not + expect too many digits. + */ + oldaccumulator = accumulator; + accumulator *= 10; + if ((accumulator+10)/10 != oldaccumulator+1) { + PyErr_Format(PyExc_ValueError, + "Too many decimal digits in format string"); + return -1; + } + accumulator += digitval; + } + *result = accumulator; + return numdigits; +} + +/************************************************************************/ +/*********** standard format specifier parsing **************************/ +/************************************************************************/ + +/* returns true if this character is a specifier alignment token */ +Py_LOCAL_INLINE(int) +is_alignment_token(Py_UCS4 c) +{ + switch (c) { + case '<': case '>': case '=': case '^': + return 1; + default: + return 0; + } +} + +/* returns true if this character is a sign element */ +Py_LOCAL_INLINE(int) +is_sign_element(Py_UCS4 c) +{ + switch (c) { + case ' ': case '+': case '-': + return 1; + default: + return 0; + } +} + + +typedef struct { + Py_UCS4 fill_char; + Py_UCS4 align; + int alternate; + Py_UCS4 sign; + Py_ssize_t width; + int thousands_separators; + Py_ssize_t precision; + Py_UCS4 type; +} InternalFormatSpec; + +#if 0 +/* Occassionally useful for debugging. Should normally be commented out. */ +static void +DEBUG_PRINT_FORMAT_SPEC(InternalFormatSpec *format) +{ + printf("internal format spec: fill_char %d\n", format->fill_char); + printf("internal format spec: align %d\n", format->align); + printf("internal format spec: alternate %d\n", format->alternate); + printf("internal format spec: sign %d\n", format->sign); + printf("internal format spec: width %zd\n", format->width); + printf("internal format spec: thousands_separators %d\n", + format->thousands_separators); + printf("internal format spec: precision %zd\n", format->precision); + printf("internal format spec: type %c\n", format->type); + printf("\n"); +} +#endif + + +/* + ptr points to the start of the format_spec, end points just past its end. + fills in format with the parsed information. + returns 1 on success, 0 on failure. + if failure, sets the exception +*/ +static int +parse_internal_render_format_spec(PyObject *format_spec, + Py_ssize_t start, Py_ssize_t end, + InternalFormatSpec *format, + char default_type, + char default_align) +{ + Py_ssize_t pos = start; + /* end-pos is used throughout this code to specify the length of + the input string */ +#define READ_spec(index) PyUnicode_READ_CHAR(format_spec, index) + + Py_ssize_t consumed; + int align_specified = 0; + + format->fill_char = '\0'; + format->align = default_align; + format->alternate = 0; + format->sign = '\0'; + format->width = -1; + format->thousands_separators = 0; + format->precision = -1; + format->type = default_type; + + /* If the second char is an alignment token, + then parse the fill char */ + if (end-pos >= 2 && is_alignment_token(READ_spec(pos+1))) { + format->align = READ_spec(pos+1); + format->fill_char = READ_spec(pos); + align_specified = 1; + pos += 2; + } + else if (end-pos >= 1 && is_alignment_token(READ_spec(pos))) { + format->align = READ_spec(pos); + align_specified = 1; + ++pos; + } + + /* Parse the various sign options */ + if (end-pos >= 1 && is_sign_element(READ_spec(pos))) { + format->sign = READ_spec(pos); + ++pos; + } + + /* If the next character is #, we're in alternate mode. This only + applies to integers. */ + if (end-pos >= 1 && READ_spec(pos) == '#') { + format->alternate = 1; + ++pos; + } + + /* The special case for 0-padding (backwards compat) */ + if (format->fill_char == '\0' && end-pos >= 1 && READ_spec(pos) == '0') { + format->fill_char = '0'; + if (!align_specified) { + format->align = '='; + } + ++pos; + } + + consumed = get_integer(format_spec, &pos, end, &format->width); + if (consumed == -1) + /* Overflow error. Exception already set. */ + return 0; + + /* If consumed is 0, we didn't consume any characters for the + width. In that case, reset the width to -1, because + get_integer() will have set it to zero. -1 is how we record + that the width wasn't specified. */ + if (consumed == 0) + format->width = -1; + + /* Comma signifies add thousands separators */ + if (end-pos && READ_spec(pos) == ',') { + format->thousands_separators = 1; + ++pos; + } + + /* Parse field precision */ + if (end-pos && READ_spec(pos) == '.') { + ++pos; + + consumed = get_integer(format_spec, &pos, end, &format->precision); + if (consumed == -1) + /* Overflow error. Exception already set. */ + return 0; + + /* Not having a precision after a dot is an error. */ + if (consumed == 0) { + PyErr_Format(PyExc_ValueError, + "Format specifier missing precision"); + return 0; + } + + } + + /* Finally, parse the type field. */ + + if (end-pos > 1) { + /* More than one char remain, invalid conversion spec. */ + PyErr_Format(PyExc_ValueError, "Invalid conversion specification"); + return 0; + } + + if (end-pos == 1) { + format->type = READ_spec(pos); + ++pos; + } + + /* Do as much validating as we can, just by looking at the format + specifier. Do not take into account what type of formatting + we're doing (int, float, string). */ + + if (format->thousands_separators) { + switch (format->type) { + case 'd': + case 'e': + case 'f': + case 'g': + case 'E': + case 'G': + case '%': + case 'F': + case '\0': + /* These are allowed. See PEP 378.*/ + break; + default: + invalid_comma_type(format->type); + return 0; + } + } + + if (format->fill_char > 127 || format->align > 127 || + format->sign > 127) { + PyErr_SetString(PyExc_ValueError, "fill character too large"); + return 0; + } + + return 1; +} + +/* Calculate the padding needed. */ +static void +calc_padding(Py_ssize_t nchars, Py_ssize_t width, Py_UCS4 align, + Py_ssize_t *n_lpadding, Py_ssize_t *n_rpadding, + Py_ssize_t *n_total) +{ + if (width >= 0) { + if (nchars > width) + *n_total = nchars; + else + *n_total = width; + } + else { + /* not specified, use all of the chars and no more */ + *n_total = nchars; + } + + /* Figure out how much leading space we need, based on the + aligning */ + if (align == '>') + *n_lpadding = *n_total - nchars; + else if (align == '^') + *n_lpadding = (*n_total - nchars) / 2; + else if (align == '<' || align == '=') + *n_lpadding = 0; + else { + /* We should never have an unspecified alignment. */ + *n_lpadding = 0; + assert(0); + } + + *n_rpadding = *n_total - nchars - *n_lpadding; +} + +static void +unicode_fill(PyObject *str, Py_ssize_t start, Py_ssize_t end, Py_UCS4 ch) +{ + int kind = PyUnicode_KIND(str); + void *data = PyUnicode_DATA(str); + while (start < end) + PyUnicode_WRITE(kind, data, start++, ch); +} + +/* Do the padding, and return a pointer to where the caller-supplied + content goes. */ +static Py_ssize_t +fill_padding(PyObject *s, Py_ssize_t start, Py_ssize_t nchars, + Py_UCS4 fill_char, Py_ssize_t n_lpadding, + Py_ssize_t n_rpadding) +{ + /* Pad on left. */ + if (n_lpadding) + unicode_fill(s, start, start + n_lpadding, fill_char); + + /* Pad on right. */ + if (n_rpadding) + unicode_fill(s, start + nchars + n_lpadding, + start + nchars + n_lpadding + n_rpadding, fill_char); + + /* Pointer to the user content. */ + return start + n_lpadding; +} + +/************************************************************************/ +/*********** common routines for numeric formatting *********************/ +/************************************************************************/ + +/* Locale type codes. */ +#define LT_CURRENT_LOCALE 0 +#define LT_DEFAULT_LOCALE 1 +#define LT_NO_LOCALE 2 + +/* Locale info needed for formatting integers and the part of floats + before and including the decimal. Note that locales only support + 8-bit chars, not unicode. */ +typedef struct { + char *decimal_point; + char *thousands_sep; + char *grouping; +} LocaleInfo; + +/* describes the layout for an integer, see the comment in + calc_number_widths() for details */ +typedef struct { + Py_ssize_t n_lpadding; + Py_ssize_t n_prefix; + Py_ssize_t n_spadding; + Py_ssize_t n_rpadding; + char sign; + Py_ssize_t n_sign; /* number of digits needed for sign (0/1) */ + Py_ssize_t n_grouped_digits; /* Space taken up by the digits, including + any grouping chars. */ + Py_ssize_t n_decimal; /* 0 if only an integer */ + Py_ssize_t n_remainder; /* Digits in decimal and/or exponent part, + excluding the decimal itself, if + present. */ + + /* These 2 are not the widths of fields, but are needed by + STRINGLIB_GROUPING. */ + Py_ssize_t n_digits; /* The number of digits before a decimal + or exponent. */ + Py_ssize_t n_min_width; /* The min_width we used when we computed + the n_grouped_digits width. */ +} NumberFieldWidths; + + +/* Given a number of the form: + digits[remainder] + where ptr points to the start and end points to the end, find where + the integer part ends. This could be a decimal, an exponent, both, + or neither. + If a decimal point is present, set *has_decimal and increment + remainder beyond it. + Results are undefined (but shouldn't crash) for improperly + formatted strings. +*/ +static void +parse_number(PyObject *s, Py_ssize_t pos, Py_ssize_t end, + Py_ssize_t *n_remainder, int *has_decimal) +{ + Py_ssize_t remainder; + + while (posn_digits = n_end - n_start - n_remainder - (has_decimal?1:0); + spec->n_lpadding = 0; + spec->n_prefix = n_prefix; + spec->n_decimal = has_decimal ? strlen(locale->decimal_point) : 0; + spec->n_remainder = n_remainder; + spec->n_spadding = 0; + spec->n_rpadding = 0; + spec->sign = '\0'; + spec->n_sign = 0; + + /* the output will look like: + | | + | | + | | + + sign is computed from format->sign and the actual + sign of the number + + prefix is given (it's for the '0x' prefix) + + digits is already known + + the total width is either given, or computed from the + actual digits + + only one of lpadding, spadding, and rpadding can be non-zero, + and it's calculated from the width and other fields + */ + + /* compute the various parts we're going to write */ + switch (format->sign) { + case '+': + /* always put a + or - */ + spec->n_sign = 1; + spec->sign = (sign_char == '-' ? '-' : '+'); + break; + case ' ': + spec->n_sign = 1; + spec->sign = (sign_char == '-' ? '-' : ' '); + break; + default: + /* Not specified, or the default (-) */ + if (sign_char == '-') { + spec->n_sign = 1; + spec->sign = '-'; + } + } + + /* The number of chars used for non-digits and non-padding. */ + n_non_digit_non_padding = spec->n_sign + spec->n_prefix + spec->n_decimal + + spec->n_remainder; + + /* min_width can go negative, that's okay. format->width == -1 means + we don't care. */ + if (format->fill_char == '0' && format->align == '=') + spec->n_min_width = format->width - n_non_digit_non_padding; + else + spec->n_min_width = 0; + + if (spec->n_digits == 0) + /* This case only occurs when using 'c' formatting, we need + to special case it because the grouping code always wants + to have at least one character. */ + spec->n_grouped_digits = 0; + else + spec->n_grouped_digits = _PyUnicode_InsertThousandsGrouping( + PyUnicode_1BYTE_KIND, NULL, 0, NULL, + spec->n_digits, spec->n_min_width, + locale->grouping, locale->thousands_sep); + + /* Given the desired width and the total of digit and non-digit + space we consume, see if we need any padding. format->width can + be negative (meaning no padding), but this code still works in + that case. */ + n_padding = format->width - + (n_non_digit_non_padding + spec->n_grouped_digits); + if (n_padding > 0) { + /* Some padding is needed. Determine if it's left, space, or right. */ + switch (format->align) { + case '<': + spec->n_rpadding = n_padding; + break; + case '^': + spec->n_lpadding = n_padding / 2; + spec->n_rpadding = n_padding - spec->n_lpadding; + break; + case '=': + spec->n_spadding = n_padding; + break; + case '>': + spec->n_lpadding = n_padding; + break; + default: + /* Shouldn't get here, but treat it as '>' */ + spec->n_lpadding = n_padding; + assert(0); + break; + } + } + return spec->n_lpadding + spec->n_sign + spec->n_prefix + + spec->n_spadding + spec->n_grouped_digits + spec->n_decimal + + spec->n_remainder + spec->n_rpadding; +} + +/* Fill in the digit parts of a numbers's string representation, + as determined in calc_number_widths(). + No error checking, since we know the buffer is the correct size. */ +static void +fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, + PyObject *digits, Py_ssize_t d_start, Py_ssize_t d_end, + PyObject *prefix, Py_ssize_t p_start, Py_UCS4 fill_char, + LocaleInfo *locale, int toupper) +{ + /* Used to keep track of digits, decimal, and remainder. */ + Py_ssize_t d_pos = d_start; + unsigned int kind = PyUnicode_KIND(out); + void *data = PyUnicode_DATA(out); + +#ifndef NDEBUG + Py_ssize_t r; +#endif + + if (spec->n_lpadding) { + unicode_fill(out, pos, pos + spec->n_lpadding, fill_char); + pos += spec->n_lpadding; + } + if (spec->n_sign == 1) { + PyUnicode_WRITE(kind, data, pos++, spec->sign); + } + if (spec->n_prefix) { + PyUnicode_CopyCharacters(out, pos, prefix, p_start, spec->n_prefix); + if (toupper) { + Py_ssize_t t; + /* XXX if the upper-case prefix is wider than the target + buffer, the caller should have allocated a wider string, + but currently doesn't. */ + for (t = 0; t < spec->n_prefix; ++t) + PyUnicode_WRITE(kind, data, pos + t, + Py_UNICODE_TOUPPER( + PyUnicode_READ(kind, data, pos + t))); + } + pos += spec->n_prefix; + } + if (spec->n_spadding) { + unicode_fill(out, pos, pos + spec->n_spadding, fill_char); + pos += spec->n_spadding; + } + + /* Only for type 'c' special case, it has no digits. */ + if (spec->n_digits != 0) { + /* Fill the digits with InsertThousandsGrouping. */ + char *pdigits = PyUnicode_DATA(digits); + if (PyUnicode_KIND(digits) < kind) { + pdigits = _PyUnicode_AsKind(digits, kind); + if (pdigits == NULL) { + /* XXX report exception */ + Py_FatalError("out of memory"); + return; + } + } +#ifndef NDEBUG + r = +#endif + _PyUnicode_InsertThousandsGrouping( + kind, + (char*)data + PyUnicode_KIND_SIZE(kind, pos), + spec->n_grouped_digits, + pdigits + PyUnicode_KIND_SIZE(kind, d_pos), + spec->n_digits, spec->n_min_width, + locale->grouping, locale->thousands_sep); +#ifndef NDEBUG + assert(r == spec->n_grouped_digits); +#endif + if (PyUnicode_KIND(digits) < kind) + PyMem_Free(pdigits); + d_pos += spec->n_digits; + } + if (toupper) { + Py_ssize_t t; + for (t = 0; t < spec->n_grouped_digits; ++t) + PyUnicode_WRITE(kind, data, pos + t, + Py_UNICODE_TOUPPER( + PyUnicode_READ(kind, data, pos + t))); + } + pos += spec->n_grouped_digits; + + if (spec->n_decimal) { + Py_ssize_t t; + for (t = 0; t < spec->n_decimal; ++t) + PyUnicode_WRITE(kind, data, pos + t, + locale->decimal_point[t]); + pos += spec->n_decimal; + d_pos += 1; + } + + if (spec->n_remainder) { + PyUnicode_CopyCharacters(out, pos, digits, d_pos, spec->n_remainder); + pos += spec->n_remainder; + d_pos += spec->n_remainder; + } + + if (spec->n_rpadding) { + unicode_fill(out, pos, pos + spec->n_rpadding, fill_char); + pos += spec->n_rpadding; + } +} + +static char no_grouping[1] = {CHAR_MAX}; + +/* Find the decimal point character(s?), thousands_separator(s?), and + grouping description, either for the current locale if type is + LT_CURRENT_LOCALE, a hard-coded locale if LT_DEFAULT_LOCALE, or + none if LT_NO_LOCALE. */ +static void +get_locale_info(int type, LocaleInfo *locale_info) +{ + switch (type) { + case LT_CURRENT_LOCALE: { + struct lconv *locale_data = localeconv(); + locale_info->decimal_point = locale_data->decimal_point; + locale_info->thousands_sep = locale_data->thousands_sep; + locale_info->grouping = locale_data->grouping; + break; + } + case LT_DEFAULT_LOCALE: + locale_info->decimal_point = "."; + locale_info->thousands_sep = ","; + locale_info->grouping = "\3"; /* Group every 3 characters. The + (implicit) trailing 0 means repeat + infinitely. */ + break; + case LT_NO_LOCALE: + locale_info->decimal_point = "."; + locale_info->thousands_sep = ""; + locale_info->grouping = no_grouping; + break; + default: + assert(0); + } +} + +/************************************************************************/ +/*********** string formatting ******************************************/ +/************************************************************************/ + +static PyObject * +format_string_internal(PyObject *value, const InternalFormatSpec *format) +{ + Py_ssize_t lpad; + Py_ssize_t rpad; + Py_ssize_t total; + Py_ssize_t pos; + Py_ssize_t len = PyUnicode_GET_SIZE(value); + PyObject *result = NULL; + int maxchar = 127; + + /* sign is not allowed on strings */ + if (format->sign != '\0') { + PyErr_SetString(PyExc_ValueError, + "Sign not allowed in string format specifier"); + goto done; + } + + /* alternate is not allowed on strings */ + if (format->alternate) { + PyErr_SetString(PyExc_ValueError, + "Alternate form (#) not allowed in string format " + "specifier"); + goto done; + } + + /* '=' alignment not allowed on strings */ + if (format->align == '=') { + PyErr_SetString(PyExc_ValueError, + "'=' alignment not allowed " + "in string format specifier"); + goto done; + } + + /* if precision is specified, output no more that format.precision + characters */ + if (format->precision >= 0 && len >= format->precision) { + len = format->precision; + } + + calc_padding(len, format->width, format->align, &lpad, &rpad, &total); + + /* allocate the resulting string */ + result = PyUnicode_New(total, maxchar); + if (result == NULL) + goto done; + + /* Write into that space. First the padding. */ + pos = fill_padding(result, 0, len, + format->fill_char=='\0'?' ':format->fill_char, + lpad, rpad); + + /* Then the source string. */ + PyUnicode_CopyCharacters(result, pos, value, 0, len); + +done: + return result; +} + + +/************************************************************************/ +/*********** long formatting ********************************************/ +/************************************************************************/ + +typedef PyObject* +(*IntOrLongToString)(PyObject *value, int base); + +static PyObject * +format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format, + IntOrLongToString tostring) +{ + PyObject *result = NULL; + int maxchar = 127; + PyObject *tmp = NULL; + Py_ssize_t inumeric_chars; + Py_UCS4 sign_char = '\0'; + Py_ssize_t n_digits; /* count of digits need from the computed + string */ + Py_ssize_t n_remainder = 0; /* Used only for 'c' formatting, which + produces non-digits */ + Py_ssize_t n_prefix = 0; /* Count of prefix chars, (e.g., '0x') */ + Py_ssize_t n_total; + Py_ssize_t prefix; + NumberFieldWidths spec; + long x; + + /* Locale settings, either from the actual locale or + from a hard-code pseudo-locale */ + LocaleInfo locale; + + /* no precision allowed on integers */ + if (format->precision != -1) { + PyErr_SetString(PyExc_ValueError, + "Precision not allowed in integer format specifier"); + goto done; + } + + /* special case for character formatting */ + if (format->type == 'c') { + /* error to specify a sign */ + if (format->sign != '\0') { + PyErr_SetString(PyExc_ValueError, + "Sign not allowed with integer" + " format specifier 'c'"); + goto done; + } + + /* taken from unicodeobject.c formatchar() */ + /* Integer input truncated to a character */ +/* XXX: won't work for int */ + x = PyLong_AsLong(value); + if (x == -1 && PyErr_Occurred()) + goto done; + if (x < 0 || x > 0x10ffff) { + PyErr_SetString(PyExc_OverflowError, + "%c arg not in range(0x110000) " + "(wide Python build)"); + goto done; + } + tmp = PyUnicode_FromOrdinal(x); + inumeric_chars = 0; + n_digits = 1; + if (x > maxchar) + maxchar = x; + + /* As a sort-of hack, we tell calc_number_widths that we only + have "remainder" characters. calc_number_widths thinks + these are characters that don't get formatted, only copied + into the output string. We do this for 'c' formatting, + because the characters are likely to be non-digits. */ + n_remainder = 1; + } + else { + int base; + int leading_chars_to_skip = 0; /* Number of characters added by + PyNumber_ToBase that we want to + skip over. */ + + /* Compute the base and how many characters will be added by + PyNumber_ToBase */ + switch (format->type) { + case 'b': + base = 2; + leading_chars_to_skip = 2; /* 0b */ + break; + case 'o': + base = 8; + leading_chars_to_skip = 2; /* 0o */ + break; + case 'x': + case 'X': + base = 16; + leading_chars_to_skip = 2; /* 0x */ + break; + default: /* shouldn't be needed, but stops a compiler warning */ + case 'd': + case 'n': + base = 10; + break; + } + + /* The number of prefix chars is the same as the leading + chars to skip */ + if (format->alternate) + n_prefix = leading_chars_to_skip; + + /* Do the hard part, converting to a string in a given base */ + tmp = tostring(value, base); + if (tmp == NULL || PyUnicode_READY(tmp) == -1) + goto done; + + inumeric_chars = 0; + n_digits = PyUnicode_GET_LENGTH(tmp); + + prefix = inumeric_chars; + + /* Is a sign character present in the output? If so, remember it + and skip it */ + if (PyUnicode_READ_CHAR(tmp, inumeric_chars) == '-') { + sign_char = '-'; + ++prefix; + ++leading_chars_to_skip; + } + + /* Skip over the leading chars (0x, 0b, etc.) */ + n_digits -= leading_chars_to_skip; + inumeric_chars += leading_chars_to_skip; + } + + /* Determine the grouping, separator, and decimal point, if any. */ + get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : + (format->thousands_separators ? + LT_DEFAULT_LOCALE : + LT_NO_LOCALE), + &locale); + + /* Calculate how much memory we'll need. */ + n_total = calc_number_widths(&spec, n_prefix, sign_char, tmp, inumeric_chars, + inumeric_chars + n_digits, n_remainder, 0, &locale, format); + + /* Allocate the memory. */ + result = PyUnicode_New(n_total, maxchar); + if (!result) + goto done; + + /* Populate the memory. */ + fill_number(result, 0, &spec, tmp, inumeric_chars, inumeric_chars + n_digits, + tmp, prefix, + format->fill_char == '\0' ? ' ' : format->fill_char, + &locale, format->type == 'X'); + +done: + Py_XDECREF(tmp); + return result; +} + +/************************************************************************/ +/*********** float formatting *******************************************/ +/************************************************************************/ + +static PyObject* +strtounicode(char *charbuffer, Py_ssize_t len) +{ + return PyUnicode_FromKindAndData(PyUnicode_1BYTE_KIND, charbuffer, len); +} + +/* much of this is taken from unicodeobject.c */ +static PyObject * +format_float_internal(PyObject *value, + const InternalFormatSpec *format) +{ + char *buf = NULL; /* buffer returned from PyOS_double_to_string */ + Py_ssize_t n_digits; + Py_ssize_t n_remainder; + Py_ssize_t n_total; + int has_decimal; + double val; + Py_ssize_t precision = format->precision; + Py_ssize_t default_precision = 6; + Py_UCS4 type = format->type; + int add_pct = 0; + Py_ssize_t index; + NumberFieldWidths spec; + int flags = 0; + PyObject *result = NULL; + int maxchar = 127; + Py_UCS4 sign_char = '\0'; + int float_type; /* Used to see if we have a nan, inf, or regular float. */ + PyObject *unicode_tmp = NULL; + + /* Locale settings, either from the actual locale or + from a hard-code pseudo-locale */ + LocaleInfo locale; + + if (format->alternate) + flags |= Py_DTSF_ALT; + + if (type == '\0') { + /* Omitted type specifier. Behaves in the same way as repr(x) + and str(x) if no precision is given, else like 'g', but with + at least one digit after the decimal point. */ + flags |= Py_DTSF_ADD_DOT_0; + type = 'r'; + default_precision = 0; + } + + if (type == 'n') + /* 'n' is the same as 'g', except for the locale used to + format the result. We take care of that later. */ + type = 'g'; + + val = PyFloat_AsDouble(value); + if (val == -1.0 && PyErr_Occurred()) + goto done; + + if (type == '%') { + type = 'f'; + val *= 100; + add_pct = 1; + } + + if (precision < 0) + precision = default_precision; + else if (type == 'r') + type = 'g'; + + /* Cast "type", because if we're in unicode we need to pass a + 8-bit char. This is safe, because we've restricted what "type" + can be. */ + buf = PyOS_double_to_string(val, (char)type, precision, flags, + &float_type); + if (buf == NULL) + goto done; + n_digits = strlen(buf); + + if (add_pct) { + /* We know that buf has a trailing zero (since we just called + strlen() on it), and we don't use that fact any more. So we + can just write over the trailing zero. */ + buf[n_digits] = '%'; + n_digits += 1; + } + + /* Since there is no unicode version of PyOS_double_to_string, + just use the 8 bit version and then convert to unicode. */ + unicode_tmp = strtounicode(buf, n_digits); + if (unicode_tmp == NULL) + goto done; + index = 0; + + /* Is a sign character present in the output? If so, remember it + and skip it */ + if (PyUnicode_READ_CHAR(unicode_tmp, index) == '-') { + sign_char = '-'; + ++index; + --n_digits; + } + + /* Determine if we have any "remainder" (after the digits, might include + decimal or exponent or both (or neither)) */ + parse_number(unicode_tmp, index, index + n_digits, &n_remainder, &has_decimal); + + /* Determine the grouping, separator, and decimal point, if any. */ + get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : + (format->thousands_separators ? + LT_DEFAULT_LOCALE : + LT_NO_LOCALE), + &locale); + + /* Calculate how much memory we'll need. */ + n_total = calc_number_widths(&spec, 0, sign_char, unicode_tmp, index, + index + n_digits, n_remainder, has_decimal, + &locale, format); + + /* Allocate the memory. */ + result = PyUnicode_New(n_total, maxchar); + if (result == NULL) + goto done; + + /* Populate the memory. */ + fill_number(result, 0, &spec, unicode_tmp, index, index + n_digits, + NULL, 0, + format->fill_char == '\0' ? ' ' : format->fill_char, &locale, + 0); + +done: + PyMem_Free(buf); + Py_DECREF(unicode_tmp); + return result; +} + +/************************************************************************/ +/*********** complex formatting *****************************************/ +/************************************************************************/ + +static PyObject * +format_complex_internal(PyObject *value, + const InternalFormatSpec *format) +{ + double re; + double im; + char *re_buf = NULL; /* buffer returned from PyOS_double_to_string */ + char *im_buf = NULL; /* buffer returned from PyOS_double_to_string */ + + InternalFormatSpec tmp_format = *format; + Py_ssize_t n_re_digits; + Py_ssize_t n_im_digits; + Py_ssize_t n_re_remainder; + Py_ssize_t n_im_remainder; + Py_ssize_t n_re_total; + Py_ssize_t n_im_total; + int re_has_decimal; + int im_has_decimal; + Py_ssize_t precision = format->precision; + Py_ssize_t default_precision = 6; + Py_UCS4 type = format->type; + Py_ssize_t i_re; + Py_ssize_t i_im; + NumberFieldWidths re_spec; + NumberFieldWidths im_spec; + int flags = 0; + PyObject *result = NULL; + int maxchar = 127; + int rkind; + void *rdata; + Py_ssize_t index; + Py_UCS4 re_sign_char = '\0'; + Py_UCS4 im_sign_char = '\0'; + int re_float_type; /* Used to see if we have a nan, inf, or regular float. */ + int im_float_type; + int add_parens = 0; + int skip_re = 0; + Py_ssize_t lpad; + Py_ssize_t rpad; + Py_ssize_t total; + PyObject *re_unicode_tmp = NULL; + PyObject *im_unicode_tmp = NULL; + + /* Locale settings, either from the actual locale or + from a hard-code pseudo-locale */ + LocaleInfo locale; + + /* Zero padding is not allowed. */ + if (format->fill_char == '0') { + PyErr_SetString(PyExc_ValueError, + "Zero padding is not allowed in complex format " + "specifier"); + goto done; + } + + /* Neither is '=' alignment . */ + if (format->align == '=') { + PyErr_SetString(PyExc_ValueError, + "'=' alignment flag is not allowed in complex format " + "specifier"); + goto done; + } + + re = PyComplex_RealAsDouble(value); + if (re == -1.0 && PyErr_Occurred()) + goto done; + im = PyComplex_ImagAsDouble(value); + if (im == -1.0 && PyErr_Occurred()) + goto done; + + if (format->alternate) + flags |= Py_DTSF_ALT; + + if (type == '\0') { + /* Omitted type specifier. Should be like str(self). */ + type = 'r'; + default_precision = 0; + if (re == 0.0 && copysign(1.0, re) == 1.0) + skip_re = 1; + else + add_parens = 1; + } + + if (type == 'n') + /* 'n' is the same as 'g', except for the locale used to + format the result. We take care of that later. */ + type = 'g'; + + if (precision < 0) + precision = default_precision; + else if (type == 'r') + type = 'g'; + + /* Cast "type", because if we're in unicode we need to pass a + 8-bit char. This is safe, because we've restricted what "type" + can be. */ + re_buf = PyOS_double_to_string(re, (char)type, precision, flags, + &re_float_type); + if (re_buf == NULL) + goto done; + im_buf = PyOS_double_to_string(im, (char)type, precision, flags, + &im_float_type); + if (im_buf == NULL) + goto done; + + n_re_digits = strlen(re_buf); + n_im_digits = strlen(im_buf); + + /* Since there is no unicode version of PyOS_double_to_string, + just use the 8 bit version and then convert to unicode. */ + re_unicode_tmp = strtounicode(re_buf, n_re_digits); + if (re_unicode_tmp == NULL) + goto done; + i_re = 0; + + im_unicode_tmp = strtounicode(im_buf, n_im_digits); + if (im_unicode_tmp == NULL) + goto done; + i_im = 0; + + /* Is a sign character present in the output? If so, remember it + and skip it */ + if (PyUnicode_READ_CHAR(re_unicode_tmp, i_re) == '-') { + re_sign_char = '-'; + ++i_re; + --n_re_digits; + } + if (PyUnicode_READ_CHAR(im_unicode_tmp, i_im) == '-') { + im_sign_char = '-'; + ++i_im; + --n_im_digits; + } + + /* Determine if we have any "remainder" (after the digits, might include + decimal or exponent or both (or neither)) */ + parse_number(re_unicode_tmp, i_re, i_re + n_re_digits, + &n_re_remainder, &re_has_decimal); + parse_number(im_unicode_tmp, i_im, i_im + n_im_digits, + &n_im_remainder, &im_has_decimal); + + /* Determine the grouping, separator, and decimal point, if any. */ + get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : + (format->thousands_separators ? + LT_DEFAULT_LOCALE : + LT_NO_LOCALE), + &locale); + + /* Turn off any padding. We'll do it later after we've composed + the numbers without padding. */ + tmp_format.fill_char = '\0'; + tmp_format.align = '<'; + tmp_format.width = -1; + + /* Calculate how much memory we'll need. */ + n_re_total = calc_number_widths(&re_spec, 0, re_sign_char, re_unicode_tmp, + i_re, i_re + n_re_digits, n_re_remainder, + re_has_decimal, &locale, &tmp_format); + + /* Same formatting, but always include a sign, unless the real part is + * going to be omitted, in which case we use whatever sign convention was + * requested by the original format. */ + if (!skip_re) + tmp_format.sign = '+'; + n_im_total = calc_number_widths(&im_spec, 0, im_sign_char, im_unicode_tmp, + i_im, i_im + n_im_digits, n_im_remainder, + im_has_decimal, &locale, &tmp_format); + + if (skip_re) + n_re_total = 0; + + /* Add 1 for the 'j', and optionally 2 for parens. */ + calc_padding(n_re_total + n_im_total + 1 + add_parens * 2, + format->width, format->align, &lpad, &rpad, &total); + + result = PyUnicode_New(total, maxchar); + if (result == NULL) + goto done; + rkind = PyUnicode_KIND(result); + rdata = PyUnicode_DATA(result); + + /* Populate the memory. First, the padding. */ + index = fill_padding(result, 0, + n_re_total + n_im_total + 1 + add_parens * 2, + format->fill_char=='\0' ? ' ' : format->fill_char, + lpad, rpad); + + if (add_parens) + PyUnicode_WRITE(rkind, rdata, index++, '('); + + if (!skip_re) { + fill_number(result, index, &re_spec, re_unicode_tmp, + i_re, i_re + n_re_digits, NULL, 0, 0, &locale, 0); + index += n_re_total; + } + fill_number(result, index, &im_spec, im_unicode_tmp, + i_im, i_im + n_im_digits, NULL, 0, 0, &locale, 0); + index += n_im_total; + PyUnicode_WRITE(rkind, rdata, index++, 'j'); + + if (add_parens) + PyUnicode_WRITE(rkind, rdata, index++, ')'); + +done: + PyMem_Free(re_buf); + PyMem_Free(im_buf); + Py_XDECREF(re_unicode_tmp); + Py_XDECREF(im_unicode_tmp); + return result; +} + +/************************************************************************/ +/*********** built in formatters ****************************************/ +/************************************************************************/ +PyObject * +_PyUnicode_FormatAdvanced(PyObject *obj, + PyObject *format_spec, + Py_ssize_t start, Py_ssize_t end) +{ + InternalFormatSpec format; + PyObject *result = NULL; + + /* check for the special case of zero length format spec, make + it equivalent to str(obj) */ + if (start == end) { + result = PyObject_Str(obj); + goto done; + } + + /* parse the format_spec */ + if (!parse_internal_render_format_spec(format_spec, start, end, + &format, 's', '<')) + goto done; + + /* type conversion? */ + switch (format.type) { + case 's': + /* no type conversion needed, already a string. do the formatting */ + result = format_string_internal(obj, &format); + break; + default: + /* unknown */ + unknown_presentation_type(format.type, obj->ob_type->tp_name); + goto done; + } + +done: + return result; +} + +static PyObject* +format_int_or_long(PyObject* obj, PyObject* format_spec, + Py_ssize_t start, Py_ssize_t end, + IntOrLongToString tostring) +{ + PyObject *result = NULL; + PyObject *tmp = NULL; + InternalFormatSpec format; + + /* check for the special case of zero length format spec, make + it equivalent to str(obj) */ + if (start == end) { + result = PyObject_Str(obj); + goto done; + } + + /* parse the format_spec */ + if (!parse_internal_render_format_spec(format_spec, start, end, + &format, 'd', '>')) + goto done; + + /* type conversion? */ + switch (format.type) { + case 'b': + case 'c': + case 'd': + case 'o': + case 'x': + case 'X': + case 'n': + /* no type conversion needed, already an int (or long). do + the formatting */ + result = format_int_or_long_internal(obj, &format, tostring); + break; + + case 'e': + case 'E': + case 'f': + case 'F': + case 'g': + case 'G': + case '%': + /* convert to float */ + tmp = PyNumber_Float(obj); + if (tmp == NULL) + goto done; + result = format_float_internal(tmp, &format); + break; + + default: + /* unknown */ + unknown_presentation_type(format.type, obj->ob_type->tp_name); + goto done; + } + +done: + Py_XDECREF(tmp); + return result; +} + +/* Need to define long_format as a function that will convert a long + to a string. In 3.0, _PyLong_Format has the correct signature. */ +#define long_format _PyLong_Format + +PyObject * +_PyLong_FormatAdvanced(PyObject *obj, + PyObject *format_spec, + Py_ssize_t start, Py_ssize_t end) +{ + return format_int_or_long(obj, format_spec, start, end, + long_format); +} + +PyObject * +_PyFloat_FormatAdvanced(PyObject *obj, + PyObject *format_spec, + Py_ssize_t start, Py_ssize_t end) +{ + PyObject *result = NULL; + InternalFormatSpec format; + + /* check for the special case of zero length format spec, make + it equivalent to str(obj) */ + if (start == end) { + result = PyObject_Str(obj); + goto done; + } + + /* parse the format_spec */ + if (!parse_internal_render_format_spec(format_spec, start, end, + &format, '\0', '>')) + goto done; + + /* type conversion? */ + switch (format.type) { + case '\0': /* No format code: like 'g', but with at least one decimal. */ + case 'e': + case 'E': + case 'f': + case 'F': + case 'g': + case 'G': + case 'n': + case '%': + /* no conversion, already a float. do the formatting */ + result = format_float_internal(obj, &format); + break; + + default: + /* unknown */ + unknown_presentation_type(format.type, obj->ob_type->tp_name); + goto done; + } + +done: + return result; +} + +PyObject * +_PyComplex_FormatAdvanced(PyObject *obj, + PyObject *format_spec, + Py_ssize_t start, Py_ssize_t end) +{ + PyObject *result = NULL; + InternalFormatSpec format; + + /* check for the special case of zero length format spec, make + it equivalent to str(obj) */ + if (start == end) { + result = PyObject_Str(obj); + goto done; + } + + /* parse the format_spec */ + if (!parse_internal_render_format_spec(format_spec, start, end, + &format, '\0', '>')) + goto done; + + /* type conversion? */ + switch (format.type) { + case '\0': /* No format code: like 'g', but with at least one decimal. */ + case 'e': + case 'E': + case 'f': + case 'F': + case 'g': + case 'G': + case 'n': + /* no conversion, already a complex. do the formatting */ + result = format_complex_internal(obj, &format); + break; + + default: + /* unknown */ + unknown_presentation_type(format.type, obj->ob_type->tp_name); + goto done; + } + +done: + return result; +} -- cgit v1.2.1 From 0511538a040f2b8c33fdafcfeabf4ad8bf8ff754 Mon Sep 17 00:00:00 2001 From: Victor Stinner Date: Wed, 28 Sep 2011 21:50:16 +0200 Subject: fill_char() can now propagate an error --- Python/formatter_unicode.c | 72 ++++++++++++++++++++++++++++++---------------- 1 file changed, 47 insertions(+), 25 deletions(-) (limited to 'Python/formatter_unicode.c') diff --git a/Python/formatter_unicode.c b/Python/formatter_unicode.c index 9fb840b160..86385f6f7f 100644 --- a/Python/formatter_unicode.c +++ b/Python/formatter_unicode.c @@ -341,7 +341,7 @@ fill_padding(PyObject *s, Py_ssize_t start, Py_ssize_t nchars, /* Pad on right. */ if (n_rpadding) - unicode_fill(s, start + nchars + n_lpadding, + unicode_fill(s, start + nchars + n_lpadding, start + nchars + n_lpadding + n_rpadding, fill_char); /* Pointer to the user content. */ @@ -501,7 +501,7 @@ calc_number_widths(NumberFieldWidths *spec, Py_ssize_t n_prefix, spec->n_grouped_digits = 0; else spec->n_grouped_digits = _PyUnicode_InsertThousandsGrouping( - PyUnicode_1BYTE_KIND, NULL, 0, NULL, + PyUnicode_1BYTE_KIND, NULL, 0, NULL, spec->n_digits, spec->n_min_width, locale->grouping, locale->thousands_sep); @@ -541,11 +541,12 @@ calc_number_widths(NumberFieldWidths *spec, Py_ssize_t n_prefix, /* Fill in the digit parts of a numbers's string representation, as determined in calc_number_widths(). - No error checking, since we know the buffer is the correct size. */ -static void + Return -1 on error, or 0 on success. */ +static int fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, PyObject *digits, Py_ssize_t d_start, Py_ssize_t d_end, - PyObject *prefix, Py_ssize_t p_start, Py_UCS4 fill_char, + PyObject *prefix, Py_ssize_t p_start, + Py_UCS4 fill_char, LocaleInfo *locale, int toupper) { /* Used to keep track of digits, decimal, and remainder. */ @@ -589,11 +590,8 @@ fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, char *pdigits = PyUnicode_DATA(digits); if (PyUnicode_KIND(digits) < kind) { pdigits = _PyUnicode_AsKind(digits, kind); - if (pdigits == NULL) { - /* XXX report exception */ - Py_FatalError("out of memory"); - return; - } + if (pdigits == NULL) + return -1; } #ifndef NDEBUG r = @@ -640,6 +638,7 @@ fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, unicode_fill(out, pos, pos + spec->n_rpadding, fill_char); pos += spec->n_rpadding; } + return 0; } static char no_grouping[1] = {CHAR_MAX}; @@ -765,6 +764,7 @@ format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format, Py_ssize_t prefix; NumberFieldWidths spec; long x; + int err; /* Locale settings, either from the actual locale or from a hard-code pseudo-locale */ @@ -886,10 +886,13 @@ format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format, goto done; /* Populate the memory. */ - fill_number(result, 0, &spec, tmp, inumeric_chars, inumeric_chars + n_digits, - tmp, prefix, - format->fill_char == '\0' ? ' ' : format->fill_char, - &locale, format->type == 'X'); + err = fill_number(result, 0, &spec, + tmp, inumeric_chars, inumeric_chars + n_digits, + tmp, prefix, + format->fill_char == '\0' ? ' ' : format->fill_char, + &locale, format->type == 'X'); + if (err) + Py_CLEAR(result); done: Py_XDECREF(tmp); @@ -929,6 +932,7 @@ format_float_internal(PyObject *value, Py_UCS4 sign_char = '\0'; int float_type; /* Used to see if we have a nan, inf, or regular float. */ PyObject *unicode_tmp = NULL; + int err; /* Locale settings, either from the actual locale or from a hard-code pseudo-locale */ @@ -1010,7 +1014,7 @@ format_float_internal(PyObject *value, &locale); /* Calculate how much memory we'll need. */ - n_total = calc_number_widths(&spec, 0, sign_char, unicode_tmp, index, + n_total = calc_number_widths(&spec, 0, sign_char, unicode_tmp, index, index + n_digits, n_remainder, has_decimal, &locale, format); @@ -1020,10 +1024,13 @@ format_float_internal(PyObject *value, goto done; /* Populate the memory. */ - fill_number(result, 0, &spec, unicode_tmp, index, index + n_digits, - NULL, 0, - format->fill_char == '\0' ? ' ' : format->fill_char, &locale, - 0); + err = fill_number(result, 0, &spec, + unicode_tmp, index, index + n_digits, + NULL, 0, + format->fill_char == '\0' ? ' ' : format->fill_char, + &locale, 0); + if (err) + Py_CLEAR(result); done: PyMem_Free(buf); @@ -1077,6 +1084,7 @@ format_complex_internal(PyObject *value, Py_ssize_t total; PyObject *re_unicode_tmp = NULL; PyObject *im_unicode_tmp = NULL; + int err; /* Locale settings, either from the actual locale or from a hard-code pseudo-locale */ @@ -1170,9 +1178,9 @@ format_complex_internal(PyObject *value, /* Determine if we have any "remainder" (after the digits, might include decimal or exponent or both (or neither)) */ - parse_number(re_unicode_tmp, i_re, i_re + n_re_digits, + parse_number(re_unicode_tmp, i_re, i_re + n_re_digits, &n_re_remainder, &re_has_decimal); - parse_number(im_unicode_tmp, i_im, i_im + n_im_digits, + parse_number(im_unicode_tmp, i_im, i_im + n_im_digits, &n_im_remainder, &im_has_decimal); /* Determine the grouping, separator, and decimal point, if any. */ @@ -1225,12 +1233,26 @@ format_complex_internal(PyObject *value, PyUnicode_WRITE(rkind, rdata, index++, '('); if (!skip_re) { - fill_number(result, index, &re_spec, re_unicode_tmp, - i_re, i_re + n_re_digits, NULL, 0, 0, &locale, 0); + err = fill_number(result, index, &re_spec, + re_unicode_tmp, i_re, i_re + n_re_digits, + NULL, 0, + 0, + &locale, 0); + if (err) { + Py_CLEAR(result); + goto done; + } index += n_re_total; } - fill_number(result, index, &im_spec, im_unicode_tmp, - i_im, i_im + n_im_digits, NULL, 0, 0, &locale, 0); + err = fill_number(result, index, &im_spec, + im_unicode_tmp, i_im, i_im + n_im_digits, + NULL, 0, + 0, + &locale, 0); + if (err) { + Py_CLEAR(result); + goto done; + } index += n_im_total; PyUnicode_WRITE(rkind, rdata, index++, 'j'); -- cgit v1.2.1 From 895a5d4f7037dfc0563aa482d07b5d32b5dca481 Mon Sep 17 00:00:00 2001 From: Victor Stinner Date: Wed, 28 Sep 2011 21:50:42 +0200 Subject: fill_number() ensures that the 'digits' string is ready --- Python/formatter_unicode.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'Python/formatter_unicode.c') diff --git a/Python/formatter_unicode.c b/Python/formatter_unicode.c index 86385f6f7f..52ccafa8d0 100644 --- a/Python/formatter_unicode.c +++ b/Python/formatter_unicode.c @@ -587,7 +587,10 @@ fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, /* Only for type 'c' special case, it has no digits. */ if (spec->n_digits != 0) { /* Fill the digits with InsertThousandsGrouping. */ - char *pdigits = PyUnicode_DATA(digits); + char *pdigits; + if (PyUnicode_READY(digits)) + return -1; + pdigits = PyUnicode_DATA(digits); if (PyUnicode_KIND(digits) < kind) { pdigits = _PyUnicode_AsKind(digits, kind); if (pdigits == NULL) -- cgit v1.2.1 From 5afe8babef534ce040e31ce8d000ce622a3366c7 Mon Sep 17 00:00:00 2001 From: Victor Stinner Date: Wed, 28 Sep 2011 21:53:49 +0200 Subject: fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure --- Python/formatter_unicode.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) (limited to 'Python/formatter_unicode.c') diff --git a/Python/formatter_unicode.c b/Python/formatter_unicode.c index 52ccafa8d0..609df648b0 100644 --- a/Python/formatter_unicode.c +++ b/Python/formatter_unicode.c @@ -566,7 +566,10 @@ fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, PyUnicode_WRITE(kind, data, pos++, spec->sign); } if (spec->n_prefix) { - PyUnicode_CopyCharacters(out, pos, prefix, p_start, spec->n_prefix); + if (PyUnicode_CopyCharacters(out, pos, + prefix, p_start, + spec->n_prefix) < 0) + return -1; if (toupper) { Py_ssize_t t; /* XXX if the upper-case prefix is wider than the target @@ -632,7 +635,8 @@ fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, } if (spec->n_remainder) { - PyUnicode_CopyCharacters(out, pos, digits, d_pos, spec->n_remainder); + if (PyUnicode_CopyCharacters(out, pos, digits, d_pos, spec->n_remainder) < 0) + return -1; pos += spec->n_remainder; d_pos += spec->n_remainder; } @@ -735,7 +739,8 @@ format_string_internal(PyObject *value, const InternalFormatSpec *format) lpad, rpad); /* Then the source string. */ - PyUnicode_CopyCharacters(result, pos, value, 0, len); + if (PyUnicode_CopyCharacters(result, pos, value, 0, len) < 0) + Py_CLEAR(result); done: return result; -- cgit v1.2.1 From 4cea3a5c83fca73972f3cf75b9688542b85c4042 Mon Sep 17 00:00:00 2001 From: Victor Stinner Date: Wed, 5 Oct 2011 21:24:08 +0200 Subject: Add asciilib: similar to ucs1, ucs2 and ucs4 library, but specialized to ASCII ucs1, ucs2 and ucs4 libraries have to scan created substring to find the maximum character, whereas it is not need to ASCII strings. Because ASCII strings are common, it is useful to optimize ASCII. --- Python/formatter_unicode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'Python/formatter_unicode.c') diff --git a/Python/formatter_unicode.c b/Python/formatter_unicode.c index 609df648b0..c989d83654 100644 --- a/Python/formatter_unicode.c +++ b/Python/formatter_unicode.c @@ -501,7 +501,7 @@ calc_number_widths(NumberFieldWidths *spec, Py_ssize_t n_prefix, spec->n_grouped_digits = 0; else spec->n_grouped_digits = _PyUnicode_InsertThousandsGrouping( - PyUnicode_1BYTE_KIND, NULL, 0, NULL, + NULL, PyUnicode_1BYTE_KIND, NULL, 0, NULL, spec->n_digits, spec->n_min_width, locale->grouping, locale->thousands_sep); @@ -603,7 +603,7 @@ fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, r = #endif _PyUnicode_InsertThousandsGrouping( - kind, + out, kind, (char*)data + PyUnicode_KIND_SIZE(kind, pos), spec->n_grouped_digits, pdigits + PyUnicode_KIND_SIZE(kind, d_pos), -- cgit v1.2.1 From 7cba64b1d6dd00da0c98fceaac11e3fc7086c852 Mon Sep 17 00:00:00 2001 From: Victor Stinner Date: Thu, 6 Oct 2011 01:45:57 +0200 Subject: Don't check for the maximum character when copying from unicodeobject.c * Create copy_characters() function which doesn't check for the maximum character in release mode * _PyUnicode_CheckConsistency() is no more static to be able to use it in _PyUnicode_FormatAdvanced() (in formatter_unicode.c) * _PyUnicode_CheckConsistency() checks the string hash --- Python/formatter_unicode.c | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-) (limited to 'Python/formatter_unicode.c') diff --git a/Python/formatter_unicode.c b/Python/formatter_unicode.c index c989d83654..a389734390 100644 --- a/Python/formatter_unicode.c +++ b/Python/formatter_unicode.c @@ -1284,33 +1284,31 @@ _PyUnicode_FormatAdvanced(PyObject *obj, Py_ssize_t start, Py_ssize_t end) { InternalFormatSpec format; - PyObject *result = NULL; + PyObject *result; /* check for the special case of zero length format spec, make it equivalent to str(obj) */ - if (start == end) { - result = PyObject_Str(obj); - goto done; - } + if (start == end) + return PyObject_Str(obj); /* parse the format_spec */ if (!parse_internal_render_format_spec(format_spec, start, end, &format, 's', '<')) - goto done; + return NULL; /* type conversion? */ switch (format.type) { case 's': /* no type conversion needed, already a string. do the formatting */ result = format_string_internal(obj, &format); + if (result != NULL) + assert(_PyUnicode_CheckConsistency(result, 1)); break; default: /* unknown */ unknown_presentation_type(format.type, obj->ob_type->tp_name); - goto done; + result = NULL; } - -done: return result; } -- cgit v1.2.1 From 631b94c03720b2c23e0098dfae52fed6526c6782 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Martin=20v=2E=20L=C3=B6wis?= Date: Fri, 7 Oct 2011 20:55:35 +0200 Subject: Change PyUnicode_KIND to 1,2,4. Drop _KIND_SIZE and _CHARACTER_SIZE. --- Python/formatter_unicode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'Python/formatter_unicode.c') diff --git a/Python/formatter_unicode.c b/Python/formatter_unicode.c index a389734390..037880068a 100644 --- a/Python/formatter_unicode.c +++ b/Python/formatter_unicode.c @@ -604,9 +604,9 @@ fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, #endif _PyUnicode_InsertThousandsGrouping( out, kind, - (char*)data + PyUnicode_KIND_SIZE(kind, pos), + (char*)data + kind * pos, spec->n_grouped_digits, - pdigits + PyUnicode_KIND_SIZE(kind, d_pos), + pdigits + kind * d_pos, spec->n_digits, spec->n_min_width, locale->grouping, locale->thousands_sep); #ifndef NDEBUG -- cgit v1.2.1 From a9a394e1adfc52461ed6e062282256b6f59e809b Mon Sep 17 00:00:00 2001 From: Victor Stinner Date: Tue, 11 Oct 2011 22:11:42 +0200 Subject: Fix misuse of PyUnicode_GET_SIZE, use PyUnicode_GET_LENGTH instead --- Python/formatter_unicode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'Python/formatter_unicode.c') diff --git a/Python/formatter_unicode.c b/Python/formatter_unicode.c index 037880068a..fda79fc657 100644 --- a/Python/formatter_unicode.c +++ b/Python/formatter_unicode.c @@ -693,7 +693,7 @@ format_string_internal(PyObject *value, const InternalFormatSpec *format) Py_ssize_t rpad; Py_ssize_t total; Py_ssize_t pos; - Py_ssize_t len = PyUnicode_GET_SIZE(value); + Py_ssize_t len = PyUnicode_GET_LENGTH(value); PyObject *result = NULL; int maxchar = 127; -- cgit v1.2.1