/* implements the string, long, and float formatters.  that is,

   string.__format__, etc. */

/* Before including this, you must include either:

   stringlib/unicodedefs.h

   stringlib/stringdefs.h

   Also, you should define the names:

   FORMAT_STRING

   FORMAT_LONG

   FORMAT_FLOAT

   to be whatever you want the public names of these functions to

   be.  These are the only non-static functions defined here.

*/

#define ALLOW_PARENS_FOR_SIGN 0

/*

    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(STRINGLIB_CHAR **ptr, STRINGLIB_CHAR *end,

                  Py_ssize_t *result)

{

    Py_ssize_t accumulator, digitval, oldaccumulator;

    int numdigits;

    accumulator = numdigits = 0;

    for (;;(*ptr)++, numdigits++) {

        if (*ptr >= end)

            break;

        digitval = STRINGLIB_TODECIMAL(**ptr);

        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(STRINGLIB_CHAR 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(STRINGLIB_CHAR c)

{

    switch (c) {

    case ' ': case '+': case '-':

#if ALLOW_PARENS_FOR_SIGN

    case '(':

#endif

        return 1;

    default:

        return 0;

    }

}

typedef struct {

    STRINGLIB_CHAR fill_char;

    STRINGLIB_CHAR align;

    int alternate;

    STRINGLIB_CHAR sign;

    Py_ssize_t width;

    Py_ssize_t precision;

    STRINGLIB_CHAR type;

} InternalFormatSpec;

/*

  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(STRINGLIB_CHAR *format_spec,

				  Py_ssize_t format_spec_len,

                                  InternalFormatSpec *format,

                                  char default_type)

{

    STRINGLIB_CHAR *ptr = format_spec;

    STRINGLIB_CHAR *end = format_spec + format_spec_len;

    /* end-ptr is used throughout this code to specify the length of

       the input string */

    Py_ssize_t specified_width;

    format->fill_char = '\0';

    format->align = '\0';

    format->alternate = 0;

    format->sign = '\0';

    format->width = -1;

    format->precision = -1;

    format->type = default_type;

    /* If the second char is an alignment token,

       then parse the fill char */

    if (end-ptr >= 2 && is_alignment_token(ptr[1])) {

        format->align = ptr[1];

        format->fill_char = ptr[0];

        ptr += 2;

    }

    else if (end-ptr >= 1 && is_alignment_token(ptr[0])) {

        format->align = ptr[0];

        ++ptr;

    }

    /* Parse the various sign options */

    if (end-ptr >= 1 && is_sign_element(ptr[0])) {

        format->sign = ptr[0];

        ++ptr;

#if ALLOW_PARENS_FOR_SIGN

        if (end-ptr >= 1 && ptr[0] == ')') {

            ++ptr;

        }

#endif

    }

    /* If the next character is #, we're in alternate mode.  This only

       applies to integers. */

    if (end-ptr >= 1 && ptr[0] == '#') {

	format->alternate = 1;

	++ptr;

    }

    /* The special case for 0-padding (backwards compat) */

    if (format->fill_char == '\0' && end-ptr >= 1 && ptr[0] == '0') {

        format->fill_char = '0';

        if (format->align == '\0') {

            format->align = '=';

        }

        ++ptr;

    }

    /* XXX add error checking */

    specified_width = get_integer(&ptr, end, &format->width);

    /* if specified_width 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 */

    if (specified_width == 0) {

        format->width = -1;

    }

    /* Parse field precision */

    if (end-ptr && ptr[0] == '.') {

        ++ptr;

        /* XXX add error checking */

        specified_width = get_integer(&ptr, end, &format->precision);

        /* not having a precision after a dot is an error */

        if (specified_width == 0) {

            PyErr_Format(PyExc_ValueError,

                         "Format specifier missing precision");

            return 0;

        }

    }

    /* Finally, parse the type field */

    if (end-ptr > 1) {

        /* invalid conversion spec */

        PyErr_Format(PyExc_ValueError, "Invalid conversion specification");

        return 0;

    }

    if (end-ptr == 1) {

        format->type = ptr[0];

        ++ptr;

    }

    return 1;

}

#if defined FORMAT_FLOAT || defined FORMAT_LONG

/************************************************************************/

/*********** common routines for numeric formatting *********************/

/************************************************************************/

/* 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 lsign;

    Py_ssize_t n_lsign;

    char rsign;

    Py_ssize_t n_rsign;

    Py_ssize_t n_total; /* just a convenience, it's derivable from the

                           other fields */

} NumberFieldWidths;

/* not all fields of format are used.  for example, precision is

   unused.  should this take discrete params in order to be more clear

   about what it does?  or is passing a single format parameter easier

   and more efficient enough to justify a little obfuscation? */

static void

calc_number_widths(NumberFieldWidths *spec, STRINGLIB_CHAR actual_sign,

		   Py_ssize_t n_prefix, Py_ssize_t n_digits,

		   const InternalFormatSpec *format)

{

    spec->n_lpadding = 0;

    spec->n_prefix = 0;

    spec->n_spadding = 0;

    spec->n_rpadding = 0;

    spec->lsign = '\0';

    spec->n_lsign = 0;

    spec->rsign = '\0';

    spec->n_rsign = 0;

    /* the output will look like:

       |                                                                    |

       | <lpadding> <lsign> <prefix> <spadding> <digits> <rsign> <rpadding> |

       |                                                                    |

       lsign and rsign are 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 */

    if (format->sign == '+') {

        /* always put a + or - */

        spec->n_lsign = 1;

        spec->lsign = (actual_sign == '-' ? '-' : '+');

    }

#if ALLOW_PARENS_FOR_SIGN

    else if (format->sign == '(') {

        if (actual_sign == '-') {

            spec->n_lsign = 1;

            spec->lsign = '(';

            spec->n_rsign = 1;

            spec->rsign = ')';

        }

    }

#endif

    else if (format->sign == ' ') {

        spec->n_lsign = 1;

        spec->lsign = (actual_sign == '-' ? '-' : ' ');

    }

    else {

        /* non specified, or the default (-) */

        if (actual_sign == '-') {

            spec->n_lsign = 1;

            spec->lsign = '-';

        }

    }

    spec->n_prefix = n_prefix;

    /* now the number of padding characters */

    if (format->width == -1) {

        /* no padding at all, nothing to do */

    }

    else {

        /* see if any padding is needed */

        if (spec->n_lsign + n_digits + spec->n_rsign +

	        spec->n_prefix >= format->width) {

            /* no padding needed, we're already bigger than the

               requested width */

        }

        else {

            /* determine which of left, space, or right padding is

               needed */

            Py_ssize_t padding = format->width -

		                    (spec->n_lsign + spec->n_prefix +

				     n_digits + spec->n_rsign);

            if (format->align == '<')

                spec->n_rpadding = padding;

            else if (format->align == '>')

                spec->n_lpadding = padding;

            else if (format->align == '^') {

                spec->n_lpadding = padding / 2;

                spec->n_rpadding = padding - spec->n_lpadding;

            }

            else if (format->align == '=')

                spec->n_spadding = padding;

            else

                spec->n_lpadding = padding;

        }

    }

    spec->n_total = spec->n_lpadding + spec->n_lsign + spec->n_prefix +

	    spec->n_spadding + n_digits + spec->n_rsign + spec->n_rpadding;

}

/* fill in the non-digit parts of a numbers's string representation,

   as determined in calc_number_widths().  returns the pointer to

   where the digits go. */

static STRINGLIB_CHAR *

fill_non_digits(STRINGLIB_CHAR *p_buf, const NumberFieldWidths *spec,

		STRINGLIB_CHAR *prefix, Py_ssize_t n_digits,

		STRINGLIB_CHAR fill_char)

{

    STRINGLIB_CHAR *p_digits;

    if (spec->n_lpadding) {

        STRINGLIB_FILL(p_buf, fill_char, spec->n_lpadding);

        p_buf += spec->n_lpadding;

    }

    if (spec->n_lsign == 1) {

        *p_buf++ = spec->lsign;

    }

    if (spec->n_prefix) {

	memmove(p_buf,

		prefix,

		spec->n_prefix * sizeof(STRINGLIB_CHAR));

	p_buf += spec->n_prefix;

    }

    if (spec->n_spadding) {

        STRINGLIB_FILL(p_buf, fill_char, spec->n_spadding);

        p_buf += spec->n_spadding;

    }

    p_digits = p_buf;

    p_buf += n_digits;

    if (spec->n_rsign == 1) {

        *p_buf++ = spec->rsign;

    }

    if (spec->n_rpadding) {

        STRINGLIB_FILL(p_buf, fill_char, spec->n_rpadding);

        p_buf += spec->n_rpadding;

    }

    return p_digits;

}

#endif /* FORMAT_FLOAT || FORMAT_LONG */

/************************************************************************/

/*********** string formatting ******************************************/

/************************************************************************/

static PyObject *

format_string_internal(PyObject *value, const InternalFormatSpec *format)

{

    Py_ssize_t width; /* total field width */

    Py_ssize_t lpad;

    STRINGLIB_CHAR *dst;

    STRINGLIB_CHAR *src = STRINGLIB_STR(value);

    Py_ssize_t len = STRINGLIB_LEN(value);

    PyObject *result = NULL;

    /* 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;

    }

    if (format->width >= 0) {

        width = format->width;

        /* but use at least len characters */

        if (len > width) {

            width = len;

        }

    }

    else {

        /* not specified, use all of the chars and no more */

        width = len;

    }

    /* allocate the resulting string */

    result = STRINGLIB_NEW(NULL, width);

    if (result == NULL)

        goto done;

    /* now write into that space */

    dst = STRINGLIB_STR(result);

    /* figure out how much leading space we need, based on the

       aligning */

    if (format->align == '>')

        lpad = width - len;

    else if (format->align == '^')

        lpad = (width - len) / 2;

    else

        lpad = 0;

    /* if right aligning, increment the destination allow space on the

       left */

    memcpy(dst + lpad, src, len * sizeof(STRINGLIB_CHAR));

    /* do any padding */

    if (width > len) {

        STRINGLIB_CHAR fill_char = format->fill_char;

        if (fill_char == '\0') {

            /* use the default, if not specified */

            fill_char = ' ';

        }

        /* pad on left */

        if (lpad)

            STRINGLIB_FILL(dst, fill_char, lpad);

        /* pad on right */

        if (width - len - lpad)

            STRINGLIB_FILL(dst + len + lpad, fill_char, width - len - lpad);

    }

done:

    return result;

}

/************************************************************************/

/*********** long formatting ********************************************/

/************************************************************************/

#if defined FORMAT_LONG || defined FORMAT_INT

typedef PyObject*

(*IntOrLongToString)(PyObject *value, int base);

static PyObject *

format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format,

			    IntOrLongToString tostring)

{

    PyObject *result = NULL;

    PyObject *tmp = NULL;

    STRINGLIB_CHAR *pnumeric_chars;

    STRINGLIB_CHAR numeric_char;

    STRINGLIB_CHAR sign = '\0';

    STRINGLIB_CHAR *p;

    Py_ssize_t n_digits;       /* count of digits need from the computed

                                  string */

    Py_ssize_t n_leading_chars;

    Py_ssize_t n_grouping_chars = 0; /* Count of additional chars to

					allocate, used for 'n'

					formatting. */

    Py_ssize_t n_prefix = 0;   /* Count of prefix chars, (e.g., '0x') */

    STRINGLIB_CHAR *prefix = NULL;

    NumberFieldWidths spec;

    long x;

    /* 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;

#ifdef Py_UNICODE_WIDE

        if (x < 0 || x > 0x10ffff) {

            PyErr_SetString(PyExc_OverflowError,

                            "%c arg not in range(0x110000) "

                            "(wide Python build)");

            goto done;

        }