-/* PSPP - computes sample statistics.
- Copyright (C) 1997-9, 2000, 2006 Free Software Foundation, Inc.
- Written by Ben Pfaff <blp@gnu.org>.
+/* PSPP - a program for statistical analysis.
+ Copyright (C) 1997-9, 2000, 2006, 2009 Free Software Foundation, Inc.
- This program is free software; you can redistribute it and/or
- modify it under the terms of the GNU General Public License as
- published by the Free Software Foundation; either version 2 of the
- License, or (at your option) any later version.
+ This program is free software: you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation, either version 3 of the License, or
+ (at your option) any later version.
- This program is distributed in the hope that it will be useful, but
- WITHOUT ANY WARRANTY; without even the implied warranty of
- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
You should have received a copy of the GNU General Public License
- along with this program; if not, write to the Free Software
- Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
- 02110-1301, USA. */
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
#include <config.h>
#include <stdint.h>
#include <stdlib.h>
#include <time.h>
-#include <gsl/gsl_math.h>
-#include "calendar.h"
-#include "format.h"
-#include "settings.h"
-#include "variable.h"
+#include <data/calendar.h>
+#include <data/format.h>
+#include <data/settings.h>
+#include <data/value.h>
#include <libpspp/assertion.h>
#include <libpspp/float-format.h>
#include <libpspp/integer-format.h>
-#include <libpspp/magic.h>
#include <libpspp/message.h>
#include <libpspp/misc.h>
-#include <libpspp/misc.h>
#include <libpspp/str.h>
+#include <libpspp/pool.h>
+#include <libpspp/i18n.h>
#include "minmax.h"
static void rounder_format (const struct rounder *, int decimals,
char *output);
\f
-/* Format of integers in output (SET WIB). */
-static enum integer_format output_integer_format = INTEGER_NATIVE;
-
-/* Format of reals in output (SET WRB). */
-static enum float_format output_float_format = FLOAT_NATIVE_DOUBLE;
-
typedef void data_out_converter_func (const union value *,
const struct fmt_spec *,
char *);
char *);
static void output_hex (const void *, size_t bytes, char *);
\f
-/* Converts the INPUT value into printable form in the exactly
- FORMAT->W characters in OUTPUT according to format
- specification FORMAT. No null terminator is appended to the
- buffer. */
-void
-data_out (const union value *input, const struct fmt_spec *format,
- char *output)
-{
- static data_out_converter_func *const converters[FMT_NUMBER_OF_FORMATS] =
+
+static data_out_converter_func *const converters[FMT_NUMBER_OF_FORMATS] =
{
#define FMT(NAME, METHOD, IMIN, OMIN, IO, CATEGORY) output_##METHOD,
#include "format.def"
};
+/* Similar to data_out. Additionally recodes the output from
+ native form into the given legacy character ENCODING.
+ OUTPUT must be provided by the caller and must be at least
+ FORMAT->w long. No null terminator is appended to OUTPUT.
+*/
+void
+data_out_legacy (const union value *input, const char *encoding,
+ const struct fmt_spec *format, char *output)
+{
assert (fmt_check_output (format));
converters[format->type] (input, format, output);
+ if (0 != strcmp (encoding, LEGACY_NATIVE)
+ && fmt_get_category (format->type) != FMT_CAT_BINARY)
+ {
+ char *s = recode_string (encoding, LEGACY_NATIVE, output, format->w );
+ memcpy (output, s, format->w);
+ free (s);
+ }
}
-/* Returns the current output integer format. */
-enum integer_format
-data_out_get_integer_format (void)
-{
- return output_integer_format;
-}
+/* Converts the INPUT value into a UTF8 encoded string, according
+ to format specification FORMAT.
-/* Sets the output integer format to INTEGER_FORMAT. */
-void
-data_out_set_integer_format (enum integer_format integer_format)
-{
- output_integer_format = integer_format;
-}
+ VALUE must be the correct width for FORMAT, that is, its
+ width must be fmt_var_width(FORMAT).
-/* Returns the current output float format. */
-enum float_format
-data_out_get_float_format (void)
+ ENCODING must be the encoding of INPUT. Normally this can
+ be obtained by calling dict_get_encoding on the dictionary
+ with which INPUT is associated.
+
+ The return value is dynamically allocated, and must be freed
+ by the caller. If POOL is non-null, then the return value is
+ allocated on that pool.
+*/
+char *
+data_out_pool (const union value *input, const char *encoding,
+ const struct fmt_spec *format, struct pool *pool)
{
- return output_float_format;
+ char *output = xmalloc (format->w + 1);
+ char *t ;
+ assert (fmt_check_output (format));
+
+ converters[format->type] (input, format, output);
+ output[format->w] = '\0';
+
+ t = recode_string_pool (UTF8, encoding, output, format->w, pool);
+ free (output);
+ return t;
}
-/* Sets the output float format to FLOAT_FORMAT. */
-void
-data_out_set_float_format (enum float_format float_format)
+char *
+data_out (const union value *input, const char *encoding, const struct fmt_spec *format)
{
- output_float_format = float_format;
+ return data_out_pool (input, encoding, format, NULL);
}
+
\f
/* Main conversion functions. */
if (number == SYSMIS)
output_missing (format, output);
- else if (!gsl_finite (number))
+ else if (!isfinite (number))
output_infinite (number, format, output);
- else
+ else
{
- if (format->type != FMT_E && fabs (number) < 1.5 * power10 (format->w))
+ if (format->type != FMT_E && fabs (number) < 1.5 * power10 (format->w))
{
struct rounder r;
rounder_init (&r, number, format->d);
output_overflow (format, output);
else
{
- if (number < 0 && strspn (buf, "0") < format->w)
+ if (number < 0 && strspn (buf, "0") < format->w)
{
char *p = &buf[format->w - 1];
*p = "}JKLMNOPQR"[*p - '0'];
}
- memcpy (output, buf, format->w);
+ memcpy (output, buf, format->w);
}
}
uint64_t integer = fabs (number);
if (number < 0)
integer = -integer;
- output_binary_integer (integer, format->w, output_integer_format,
+ output_binary_integer (integer, format->w,
+ settings_get_output_integer_format (),
output);
}
}
|| number < 0 || number >= power256 (format->w))
memset (output, 0, format->w);
else
- output_binary_integer (number, format->w, output_integer_format, output);
+ output_binary_integer (number, format->w,
+ settings_get_output_integer_format (), output);
}
/* Outputs PIBHEX format. */
if (number == SYSMIS)
goto missing;
- if (fmt_get_category (format->type) == FMT_CAT_DATE)
+ if (fmt_get_category (format->type) == FMT_CAT_DATE)
{
if (number <= 0)
goto missing;
{
int ch = *template;
int count = 1;
- while (template[count] == ch)
+ while (template[count] == ch)
count++;
template += count;
-
+
switch (ch)
{
case 'd':
p += sprintf (p, "%02d", month);
else
{
- static const char *months[12] =
+ static const char *const months[12] =
{
"JAN", "FEB", "MAR", "APR", "MAY", "JUN",
"JUL", "AUG", "SEP", "OCT", "NOV", "DEC",
};
- p = stpcpy (p, months[month - 1]);
+ p = stpcpy (p, months[month - 1]);
}
break;
case 'y':
- if (count >= 4 || excess_width >= 2)
+ if (count >= 4 || excess_width >= 2)
{
if (year <= 9999)
p += sprintf (p, "%04d", year);
else if (format->type == FMT_DATETIME)
p = stpcpy (p, "****");
else
- goto overflow;
+ goto overflow;
}
- else
+ else
{
- int offset = year - get_epoch ();
+ int epoch = settings_get_epoch ();
+ int offset = year - epoch;
if (offset < 0 || offset > 99)
goto overflow;
- p += sprintf (p, "%02d", abs (year) % 100);
+ p += sprintf (p, "%02d", abs (year) % 100);
}
break;
case 'q':
p += sprintf (p, "%2d", (yday - 1) / 7 + 1);
break;
case 'D':
- if (number < 0)
+ if (number < 0)
*p++ = '-';
- number = fabs (number);
+ number = fabs (number);
p += sprintf (p, "%*.0f", count, floor (number / 60. / 60. / 24.));
number = fmod (number, 60. * 60. * 24.);
break;
case 'H':
if (number < 0)
*p++ = '-';
- number = fabs (number);
+ number = fabs (number);
p += sprintf (p, "%0*.0f", count, floor (number / 60. / 60.));
number = fmod (number, 60. * 60.);
break;
p += sprintf (p, "%02d", (int) floor (number / 60.));
number = fmod (number, 60.);
excess_width = format->w - (p - tmp);
- if (excess_width < 0)
+ if (excess_width < 0)
goto overflow;
if (excess_width == 3 || excess_width == 4
|| (excess_width >= 5 && format->d == 0))
int d = MIN (format->d, excess_width - 4);
int w = d + 3;
sprintf (p, ":%0*.*f", w, d, number);
- if (fmt_decimal_char (FMT_F) != '.')
+ if (settings_get_decimal_char (FMT_F) != '.')
{
char *cp = strchr (p, '.');
if (cp != NULL)
- *cp = fmt_decimal_char (FMT_F);
+ *cp = settings_get_decimal_char (FMT_F);
}
p += strlen (p);
}
default:
assert (count == 1);
*p++ = ch;
- break;
+ break;
}
}
- buf_copy_lpad (output, format->w, tmp, p - tmp);
+ buf_copy_lpad (output, format->w, tmp, p - tmp, ' ');
return;
overflow:
output_WKDAY (const union value *input, const struct fmt_spec *format,
char *output)
{
- static const char *weekdays[7] =
+ static const char *const weekdays[7] =
{
"SUNDAY", "MONDAY", "TUESDAY", "WEDNESDAY",
"THURSDAY", "FRIDAY", "SATURDAY",
};
if (input->f >= 1 && input->f < 8)
- buf_copy_str_rpad (output, format->w, weekdays[(int) input->f - 1]);
+ buf_copy_str_rpad (output, format->w, weekdays[(int) input->f - 1], ' ');
else
{
if (input->f != SYSMIS)
output_MONTH (const union value *input, const struct fmt_spec *format,
char *output)
{
- static const char *months[12] =
+ static const char *const months[12] =
{
"JANUARY", "FEBRUARY", "MARCH", "APRIL", "MAY", "JUNE",
"JULY", "AUGUST", "SEPTEMBER", "OCTOBER", "NOVEMBER", "DECEMBER",
};
if (input->f >= 1 && input->f < 13)
- buf_copy_str_rpad (output, format->w, months[(int) input->f - 1]);
+ buf_copy_str_rpad (output, format->w, months[(int) input->f - 1], ' ');
else
{
if (input->f != SYSMIS)
output_A (const union value *input, const struct fmt_spec *format,
char *output)
{
- memcpy (output, input->s, format->w);
+ memcpy (output, value_str (input, format->w), format->w);
}
/* Outputs AHEX format. */
output_AHEX (const union value *input, const struct fmt_spec *format,
char *output)
{
- output_hex (input->s, format->w, output);
+ output_hex (value_str (input, format->w), format->w / 2, output);
}
\f
/* Decimal and scientific formatting. */
allocate_space (int request, int max_width, int *width)
{
assert (*width <= max_width);
- if (request + *width <= max_width)
+ if (request + *width <= max_width)
{
*width += request;
return true;
output_decimal (const struct rounder *r, const struct fmt_spec *format,
bool require_affixes, char *output)
{
- const struct fmt_number_style *style = fmt_get_style (format->type);
+ const struct fmt_number_style *style =
+ settings_get_style (format->type);
+
int decimals;
- for (decimals = format->d; decimals >= 0; decimals--)
+ for (decimals = format->d; decimals >= 0; decimals--)
{
/* Formatted version of magnitude of NUMBER. */
char magnitude[64];
format->w, &width);
if (!add_affixes && require_affixes)
continue;
-
+
/* Check whether we should include grouping characters.
We need room for a complete set or we don't insert any at all.
We don't include grouping characters if decimal places were
/* Format the number's magnitude. */
rounder_format (r, decimals, magnitude);
-
+
/* Assemble number. */
p = output;
if (format->w > width)
if (i > 0 && (integer_digits - i) % 3 == 0)
*p++ = style->grouping;
*p++ = magnitude[i];
- }
+ }
}
if (decimals > 0)
{
output_scientific (double number, const struct fmt_spec *format,
bool require_affixes, char *output)
{
- const struct fmt_number_style *style = fmt_get_style (format->type);
+ const struct fmt_number_style *style =
+ settings_get_style (format->type);
int width;
int fraction_width;
bool add_affixes;
# flag does in the call to sprintf, below.) */
fraction_width = MIN (MIN (format->d + 1, format->w - width), 16);
if (format->type != FMT_E && fraction_width == 1)
- fraction_width = 0;
+ fraction_width = 0;
width += fraction_width;
/* Format (except suffix). */
{
char *cp = strchr (p, 'E') + 1;
long int exponent = strtol (cp, NULL, 10);
- if (abs (exponent) > 999)
+ if (abs (exponent) > 999)
return false;
sprintf (cp, "%+04ld", exponent);
}
p = mempset (p, ' ', ss_length (style->neg_suffix));
assert (p == buf + format->w);
+ memcpy (output, buf, format->w);
- buf_copy_str_lpad (output, format->w, buf);
return true;
}
\f
-#ifndef HAVE_ROUND
-/* Return X rounded to the nearest integer,
- rounding ties away from zero. */
-static double
-round (double x)
-{
- return x >= 0.0 ? floor (x + .5) : ceil (x - .5);
-}
-#endif /* !HAVE_ROUND */
-
/* Returns true if the magnitude represented by R should be
rounded up when chopped off at DECIMALS decimal places, false
if it should be rounded down. */
static bool
-should_round_up (const struct rounder *r, int decimals)
+should_round_up (const struct rounder *r, int decimals)
{
int digit = r->string[r->integer_digits + decimals + 1];
assert (digit >= '0' && digit <= '9');
{
assert (fabs (number) < 1e41);
assert (max_decimals >= 0 && max_decimals <= 16);
- if (max_decimals == 0)
+ if (max_decimals == 0)
{
/* Fast path. No rounding needed.
round_up assumes that fractional digits are present. */
sprintf (r->string, "%.0f.00", fabs (round (number)));
}
- else
+ else
{
/* Slow path.
-
+
This is more difficult than it really should be because
we have to make sure that numbers that are exactly
halfway between two representations are always rounded
(usually it rounds to even), so we have to fake it as
best we can, by formatting with extra precision and then
doing the rounding ourselves.
-
+
We take up to two rounds to format numbers. In the
first round, we obtain 2 digits of precision beyond
those requested by the user. If those digits are
exactly "50", then in a second round we format with as
many digits as are significant in a "double".
-
+
It might be better to directly implement our own
floating-point formatting routine instead of relying on
the system's sprintf implementation. But the classic
sprintf (r->string, "%.*f", format_decimals, fabs (number));
}
}
-
- if (r->string[0] == '0')
+
+ if (r->string[0] == '0')
memmove (r->string, &r->string[1], strlen (r->string));
r->leading_zeros = strspn (r->string, "0.");
*NEGATIVE is set to true; otherwise, it is set to false. */
static int
rounder_width (const struct rounder *r, int decimals,
- int *integer_digits, bool *negative)
+ int *integer_digits, bool *negative)
{
/* Calculate base measures. */
int width = r->integer_digits;
if (should_round_up (r, decimals))
{
/* Rounding up leading 9s adds a new digit (a 1). */
- if (r->leading_nines >= width)
+ if (r->leading_nines >= width)
{
width++;
- ++*integer_digits;
+ ++*integer_digits;
}
}
else
{
/* Rounding down. */
- if (r->leading_zeros >= width)
+ if (r->leading_zeros >= width)
{
/* All digits that remain after rounding are zeros.
Therefore we drop the negative sign. */
*negative = false;
- if (r->integer_digits == 0 && decimals == 0)
+ if (r->integer_digits == 0 && decimals == 0)
{
/* No digits at all are left. We need to display
at least a single digit (a zero). */
assert (width == 0);
width++;
*integer_digits = 1;
- }
+ }
}
}
return width;
indicated by rounder_width are written. No terminating null
is appended. */
static void
-rounder_format (const struct rounder *r, int decimals, char *output)
+rounder_format (const struct rounder *r, int decimals, char *output)
{
int base_width = r->integer_digits + (decimals > 0 ? decimals + 1 : 0);
- if (should_round_up (r, decimals))
+ if (should_round_up (r, decimals))
{
- if (r->leading_nines < base_width)
+ if (r->leading_nines < base_width)
{
/* Rounding up. This is the common case where rounding
up doesn't add an extra digit. */
char *p;
memcpy (output, r->string, base_width);
- for (p = output + base_width - 1; ; p--)
+ for (p = output + base_width - 1; ; p--)
{
assert (p >= output);
if (*p == '9')
*p = '0';
- else if (*p >= '0' && *p <= '8')
+ else if (*p >= '0' && *p <= '8')
{
(*p)++;
break;
assert (*p == '.');
}
}
- else
+ else
{
/* Rounding up leading 9s causes the result to be a 1
followed by a number of 0s, plus a decimal point. */
char *p = output;
*p++ = '1';
p = mempset (p, '0', r->integer_digits);
- if (decimals > 0)
+ if (decimals > 0)
{
*p++ = '.';
p = mempset (p, '0', decimals);
assert (p == output + base_width + 1);
}
}
- else
+ else
{
/* Rounding down. */
- if (r->integer_digits != 0 || decimals != 0)
+ if (r->integer_digits != 0 || decimals != 0)
{
/* Common case: just copy the digits. */
- memcpy (output, r->string, base_width);
+ memcpy (output, r->string, base_width);
}
- else
+ else
{
/* No digits remain. The output is just a zero. */
- output[0] = '0';
+ output[0] = '0';
}
}
}
/* Returns 10**X. */
static double PURE_FUNCTION
-power10 (int x)
+power10 (int x)
{
static const double p[] =
{
/* Returns 256**X. */
static double PURE_FUNCTION
-power256 (int x)
+power256 (int x)
{
- static const double p[] =
+ static const double p[] =
{
1.0,
256.0,
static void
output_infinite (double number, const struct fmt_spec *format, char *output)
{
- assert (!gsl_finite (number));
-
+ assert (!isfinite (number));
+
if (format->w >= 3)
{
const char *s;
else
s = "Unknown";
- buf_copy_str_lpad (output, format->w, s);
+ buf_copy_str_lpad (output, format->w, s, ' ');
}
- else
+ else
output_overflow (format, output);
}
{
memset (output, ' ', format->w);
- if (format->type != FMT_N)
+ if (format->type != FMT_N)
{
int dot_ofs = (format->type == FMT_PCT ? 2
: format->type == FMT_E ? 5
: 1);
- output[MAX (0, format->w - format->d - dot_ofs)] = '.';
+ output[MAX (0, format->w - format->d - dot_ofs)] = '.';
}
else
output[format->w - 1] = '.';
/* Formats OUTPUT for overflow given FORMAT. */
static void
-output_overflow (const struct fmt_spec *format, char *output)
+output_overflow (const struct fmt_spec *format, char *output)
{
memset (output, '*', format->w);
}
representable. On failure, OUTPUT is cleared to all zero
bytes. */
static bool
-output_bcd_integer (double number, int digits, char *output)
+output_bcd_integer (double number, int digits, char *output)
{
char decimal[64];
const char *src = decimal;
int i;
- for (i = 0; i < digits / 2; i++)
+ for (i = 0; i < digits / 2; i++)
{
int d0 = *src++ - '0';
int d1 = *src++ - '0';
- *output++ = (d0 << 4) + d1;
+ *output++ = (d0 << 4) + d1;
}
if (digits % 2)
*output = (*src - '0') << 4;
-
+
return true;
}
- else
+ else
{
memset (output, 0, DIV_RND_UP (digits, 2));
- return false;
+ return false;
}
}
given INTEGER_FORMAT. */
static void
output_binary_integer (uint64_t value, int bytes,
- enum integer_format integer_format, char *output)
+ enum integer_format integer_format, char *output)
{
integer_put (value, integer_format, output, bytes);
}
/* Converts the BYTES bytes in DATA to twice as many hexadecimal
digits in OUTPUT. */
static void
-output_hex (const void *data_, size_t bytes, char *output)
+output_hex (const void *data_, size_t bytes, char *output)
{
const uint8_t *data = data_;
size_t i;