/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2006, 2009, 2011 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2006, 2009, 2011, 2012, 2013, 2014 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
#include <config.h>
-#include "data-out.h"
+#include "data/data-out.h"
#include <ctype.h>
#include <float.h>
#include <stdint.h>
#include <stdlib.h>
#include <time.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/message.h>
-#include <libpspp/misc.h>
-#include <libpspp/str.h>
-#include <libpspp/pool.h>
-#include <libpspp/i18n.h>
-
-#include "minmax.h"
+#include <unistr.h>
+
+#include "data/calendar.h"
+#include "data/format.h"
+#include "data/settings.h"
+#include "data/value.h"
+#include "libpspp/assertion.h"
+#include "libpspp/cast.h"
+#include "libpspp/float-format.h"
+#include "libpspp/i18n.h"
+#include "libpspp/integer-format.h"
+#include "libpspp/message.h"
+#include "libpspp/misc.h"
+#include "libpspp/pool.h"
+#include "libpspp/str.h"
+
+#include "gl/minmax.h"
+#include "gl/c-snprintf.h"
#include "gettext.h"
#define _(msgid) gettext (msgid)
bool negative; /* Is the number negative? */
};
-static void rounder_init (struct rounder *, double number, int max_decimals);
+static void rounder_init (struct rounder *, const struct fmt_number_style *,
+ double number, int max_decimals);
static int rounder_width (const struct rounder *, int decimals,
int *integer_digits, bool *negative);
static void rounder_format (const struct rounder *, int decimals,
\f
typedef void data_out_converter_func (const union value *,
const struct fmt_spec *,
- char *);
+ const struct fmt_settings *, char *);
#define FMT(NAME, METHOD, IMIN, OMIN, IO, CATEGORY) \
static data_out_converter_func output_##METHOD;
#include "format.def"
static bool output_decimal (const struct rounder *, const struct fmt_spec *,
+ const struct fmt_number_style *,
bool require_affixes, char *);
static bool output_scientific (double, const struct fmt_spec *,
+ const struct fmt_number_style *,
bool require_affixes, char *);
static double power10 (int) PURE_FUNCTION;
#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.
-*/
+/* Converts the INPUT value, encoded in INPUT_ENCODING, according to format
+ specification FORMAT, appending the output to OUTPUT in OUTPUT_ENCODING.
+ However, binary formats (FMT_P, FMT_PK, FMT_IB, FMT_PIB, FMT_RB) yield the
+ binary results, which may not be properly encoded for OUTPUT_ENCODING.
+
+ VALUE must be the correct width for FORMAT, that is, its width must be
+ fmt_var_width(FORMAT).
+
+ INPUT_ENCODING can normally be obtained by calling dict_get_encoding() on
+ the dictionary with which INPUT is associated. ENCODING is only important
+ when FORMAT's type is FMT_A. */
void
-data_out_legacy (const union value *input, const char *encoding,
- const struct fmt_spec *format, char *output)
+data_out_recode (const union value *input, const char *input_encoding,
+ const struct fmt_spec *format,
+ const struct fmt_settings *settings,
+ struct string *output, const char *output_encoding)
{
assert (fmt_check_output (format));
+ if (format->type == FMT_A)
+ {
+ char *in = CHAR_CAST (char *, input->s);
+ char *out = recode_string (output_encoding, input_encoding,
+ in, format->w);
+ ds_put_cstr (output, out);
+ free (out);
+ }
+ else if (fmt_get_category (format->type) == FMT_CAT_BINARY)
+ converters[format->type] (input, format, settings,
+ ds_put_uninit (output, format->w));
+ else
+ {
+ char *utf8_encoded = data_out (input, input_encoding, format, settings);
+ char *output_encoded = recode_string (output_encoding, UTF8,
+ utf8_encoded, -1);
+ ds_put_cstr (output, output_encoded);
+ free (output_encoded);
+ free (utf8_encoded);
+ }
+}
- converters[format->type] (input, format, output);
- if (0 != strcmp (encoding, C_ENCODING)
- && fmt_get_category (format->type) != FMT_CAT_BINARY)
+static char *
+binary_to_utf8 (const char *in, struct pool *pool)
+{
+ uint8_t *out = pool_alloc_unaligned (pool, strlen (in) * 2 + 1);
+ uint8_t *p = out;
+
+ while (*in != '\0')
{
- char *s = recode_string (encoding, C_ENCODING, output, format->w );
- memcpy (output, s, format->w);
- free (s);
+ uint8_t byte = *in++;
+ int mblen = u8_uctomb (p, byte, 2);
+ assert (mblen > 0);
+ p += mblen;
}
+ *p = '\0';
+
+ return CHAR_CAST (char *, out);
}
-/* Converts the INPUT value into a UTF8 encoded string, according
- to format specification FORMAT.
+/* Converts the INPUT value into a UTF-8 encoded string, according to format
+ specification FORMAT.
- VALUE must be the correct width for FORMAT, that is, its
- width must be fmt_var_width(FORMAT).
+ VALUE must be the correct width for FORMAT, that is, its width must be
+ fmt_var_width(FORMAT).
- 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.
+ INPUT_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. INPUT_ENCODING is only important when FORMAT's type is FMT_A.
- 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.
-*/
+ 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)
+data_out_pool (const union value *input, const char *input_encoding,
+ const struct fmt_spec *format,
+ const struct fmt_settings *settings, struct pool *pool)
{
- const struct fmt_number_style *style = settings_get_style (format->type);
- char *output;
- char *t ;
assert (fmt_check_output (format));
+ if (format->type == FMT_A)
+ {
+ char *in = CHAR_CAST (char *, input->s);
+ return recode_string_pool (UTF8, input_encoding, in, format->w, pool);
+ }
+ else if (fmt_get_category (format->type) == FMT_CAT_BINARY)
+ {
+ char tmp[16];
- output = xmalloc (format->w + style->extra_bytes + 1);
+ assert (format->w + 1 <= sizeof tmp);
+ converters[format->type] (input, format, settings, tmp);
+ return binary_to_utf8 (tmp, pool);
+ }
+ else
+ {
+ const struct fmt_number_style *style = fmt_settings_get_style (
+ settings, format->type);
+ size_t size = format->w + style->extra_bytes + 1;
+ char *output;
+
+ output = pool_alloc_unaligned (pool, size);
+ converters[format->type] (input, format, settings, output);
+ return output;
+ }
+}
- converters[format->type] (input, format, output);
+/* Like data_out_pool(), except that for basic numeric formats (F, COMMA, DOT,
+ COLLAR, PCT, E) and custom currency formats are formatted as wide as
+ necessary to fully display the selected number of decimal places. */
+char *
+data_out_stretchy (const union value *input, const char *encoding,
+ const struct fmt_spec *format,
+ const struct fmt_settings *settings, struct pool *pool)
+{
- t = recode_string_pool (UTF8, encoding, output, format->w, pool);
- free (output);
- return t;
+ if (fmt_get_category (format->type) & (FMT_CAT_BASIC | FMT_CAT_CUSTOM))
+ {
+ const struct fmt_number_style *style
+ = fmt_settings_get_style (settings, format->type);
+ struct fmt_spec wide_format;
+ char tmp[128];
+ size_t size;
+
+ wide_format.type = format->type;
+ wide_format.w = 40;
+ wide_format.d = format->d;
+
+ size = format->w + style->extra_bytes + 1;
+ if (size <= sizeof tmp)
+ {
+ output_number (input, &wide_format, settings, tmp);
+ return pool_strdup (pool, tmp + strspn (tmp, " "));
+ }
+ }
+
+ return data_out_pool (input, encoding, format, settings, pool);
}
char *
-data_out (const union value *input, const char *encoding, const struct fmt_spec *format)
+data_out (const union value *input, const char *input_encoding,
+ const struct fmt_spec *format, const struct fmt_settings *settings)
{
- return data_out_pool (input, encoding, format, NULL);
+ return data_out_pool (input, input_encoding, format, settings, NULL);
}
\f
CCE formats. */
static void
output_number (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings, char *output)
{
double number = input->f;
output_infinite (number, format, output);
else
{
+ const struct fmt_number_style *style =
+ fmt_settings_get_style (settings, format->type);
+
if (format->type != FMT_E && fabs (number) < 1.5 * power10 (format->w))
{
struct rounder r;
- rounder_init (&r, number, format->d);
+ rounder_init (&r, style, number, format->d);
- if (output_decimal (&r, format, true, output)
- || output_scientific (number, format, true, output)
- || output_decimal (&r, format, false, output))
+ if (output_decimal (&r, format, style, true, output)
+ || output_scientific (number, format, style, true, output)
+ || output_decimal (&r, format, style, false, output))
return;
}
- if (!output_scientific (number, format, false, output))
+ if (!output_scientific (number, format, style, false, output))
output_overflow (format, output);
}
}
/* Outputs N format. */
static void
output_N (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
double number = input->f * power10 (format->d);
if (input->f == SYSMIS || number < 0)
char buf[128];
number = fabs (round (number));
if (number < power10 (format->w)
- && sprintf (buf, "%0*.0f", format->w, number) == format->w)
+ && c_snprintf (buf, 128, "%0*.0f", format->w, number) == format->w)
memcpy (output, buf, format->w);
else
output_overflow (format, output);
/* Outputs Z format. */
static void
output_Z (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
double number = input->f * power10 (format->d);
char buf[128];
if (input->f == SYSMIS)
output_missing (format, output);
- else if (fabs (number) >= power10 (format->w)
- || sprintf (buf, "%0*.0f", format->w,
- fabs (round (number))) != format->w)
- output_overflow (format, output);
- else
+ else if (fabs (number) < power10 (format->w)
+ && c_snprintf (buf, 128, "%0*.0f", format->w,
+ fabs (round (number))) == format->w)
{
if (number < 0 && strspn (buf, "0") < format->w)
{
memcpy (output, buf, format->w);
output[format->w] = '\0';
}
+ else
+ output_overflow (format, output);
}
/* Outputs P format. */
static void
output_P (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
if (output_bcd_integer (fabs (input->f * power10 (format->d)),
format->w * 2 - 1, output)
/* Outputs PK format. */
static void
output_PK (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
output_bcd_integer (input->f * power10 (format->d), format->w * 2, output);
}
/* Outputs IB format. */
static void
output_IB (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
double number = round (input->f * power10 (format->d));
if (input->f == SYSMIS
/* Outputs PIB format. */
static void
output_PIB (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
double number = round (input->f * power10 (format->d));
if (input->f == SYSMIS
/* Outputs PIBHEX format. */
static void
output_PIBHEX (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
double number = round (input->f);
if (input->f == SYSMIS)
/* Outputs RB format. */
static void
output_RB (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
double d = input->f;
memcpy (output, &d, format->w);
/* Outputs RBHEX format. */
static void
output_RBHEX (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
double d = input->f;
DATETIME, TIME, and DTIME formats. */
static void
output_date (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings, char *output)
{
double number = input->f;
int year, month, day, yday;
- const char *template = fmt_date_template (format->type);
- size_t template_width = strlen (template);
- int excess_width = format->w - template_width;
+ const char *template = fmt_date_template (format->type, format->w);
char tmp[64];
char *p = tmp;
- assert (format->w >= template_width);
if (number == SYSMIS)
goto missing;
while (*template != '\0')
{
+ int excess_width;
+
int ch = *template;
int count = 1;
while (template[count] == ch)
}
break;
case 'y':
- if (count >= 4 || excess_width >= 2)
+ if (count >= 4)
{
if (year <= 9999)
p += sprintf (p, "%04d", year);
- else if (format->type == FMT_DATETIME)
+ else if (format->type == FMT_DATETIME
+ || format->type == FMT_YMDHMS)
p = stpcpy (p, "****");
else
goto overflow;
}
else
{
- int epoch = settings_get_epoch ();
+ int epoch = fmt_settings_get_epoch (settings);
int offset = year - epoch;
if (offset < 0 || offset > 99)
goto overflow;
if (number < 0)
*p++ = '-';
number = fabs (number);
- p += sprintf (p, "%*.0f", count, floor (number / 60. / 60. / 24.));
+ p += c_snprintf (p, 64, "%*.0f", count, floor (number / 60. / 60. / 24.));
number = fmod (number, 60. * 60. * 24.);
break;
case 'H':
if (number < 0)
*p++ = '-';
number = fabs (number);
- p += sprintf (p, "%0*.0f", count, floor (number / 60. / 60.));
+ p += c_snprintf (p, 64, "%0*.0f", count, floor (number / 60. / 60.));
number = fmod (number, 60. * 60.);
break;
case 'M':
+ if (number < 0)
+ *p++ = '-';
+ number = fabs (number);
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
+ || (format->type == FMT_MTIME && excess_width < 3))
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 (settings_get_decimal_char (FMT_F) != '.')
+ c_snprintf (p, 64, ":%0*.*f", w, d, number);
+ if (settings->decimal != '.')
{
char *cp = strchr (p, '.');
if (cp != NULL)
- *cp = settings_get_decimal_char (FMT_F);
+ *cp = settings->decimal;
}
p += strlen (p);
}
- break;
- case 'X':
- *p++ = ' ';
- break;
+ goto done;
default:
assert (count == 1);
*p++ = ch;
}
}
+ done:
buf_copy_lpad (output, format->w, tmp, p - tmp, ' ');
output[format->w] = '\0';
return;
/* Outputs WKDAY format. */
static void
output_WKDAY (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
static const char *const weekdays[7] =
{
/* Outputs MONTH format. */
static void
output_MONTH (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
static const char *const months[12] =
{
/* Outputs A format. */
static void
-output_A (const union value *input, const struct fmt_spec *format,
- char *output)
+output_A (const union value *input UNUSED,
+ const struct fmt_spec *format UNUSED,
+ const struct fmt_settings *settings UNUSED, char *output UNUSED)
{
- memcpy (output, value_str (input, format->w), format->w);
- output[format->w] = '\0';
+ NOT_REACHED ();
}
/* Outputs AHEX format. */
static void
output_AHEX (const union value *input, const struct fmt_spec *format,
- char *output)
+ const struct fmt_settings *settings UNUSED, char *output)
{
- output_hex (value_str (input, format->w), format->w / 2, output);
+ output_hex (input->s, format->w / 2, output);
}
\f
/* Decimal and scientific formatting. */
}
/* Tries to compose the number represented by R, in the style of
- FORMAT, into OUTPUT. Returns true if successful, false on
- failure, which occurs if FORMAT's width is too narrow. If
+ FORMAT and STYLE, into OUTPUT. Returns true if successful, false on
+ failure, which cocurs if FORMAT's width is too narrow. If
REQUIRE_AFFIXES is true, then the prefix and suffix specified
by FORMAT's style must be included; otherwise, they may be
omitted to make the number fit. */
static bool
output_decimal (const struct rounder *r, const struct fmt_spec *format,
- bool require_affixes, char *output)
+ const struct fmt_number_style *style, bool require_affixes,
+ char *output)
{
- const struct fmt_number_style *style =
- settings_get_style (format->type);
-
int decimals;
for (decimals = format->d; decimals >= 0; decimals--)
return false;
}
-/* Formats NUMBER into OUTPUT in scientific notation according to
- the style of the format specified in FORMAT. */
+/* Formats NUMBER into OUTPUT in scientific notation according to FORMAT and
+ STYLE. */
static bool
output_scientific (double number, const struct fmt_spec *format,
+ const struct fmt_number_style *style,
bool require_affixes, char *output)
{
- const struct fmt_number_style *style =
- settings_get_style (format->type);
int width;
int fraction_width;
bool add_affixes;
- char buf[64], *p;
+ char *p;
/* Allocate minimum required space. */
width = 6 + style->neg_suffix.width;
/* Figure out number of characters we can use for the fraction,
if any. (If that turns out to be 1, then we'll output a
decimal point without any digits following; that's what the
- # flag does in the call to sprintf, below.) */
+ # flag does in the call to c_snprintf, below.) */
fraction_width = MIN (MIN (format->d + 1, format->w - width), 16);
if (format->type != FMT_E && fraction_width == 1)
fraction_width = 0;
width += fraction_width;
/* Format (except suffix). */
- p = buf;
+ p = output;
if (width < format->w)
p = mempset (p, ' ', format->w - width);
if (number < 0)
if (add_affixes)
p = stpcpy (p, style->prefix.s);
if (fraction_width > 0)
- sprintf (p, "%#.*E", fraction_width - 1, fabs (number));
+ c_snprintf (p, 64, "%#.*E", fraction_width - 1, fabs (number));
else
- sprintf (p, "%.0E", fabs (number));
+ c_snprintf (p, 64, "%.0E", fabs (number));
/* The C locale always uses a period `.' as a decimal point.
Translate to comma if necessary. */
{
char *cp = strchr (p, 'E') + 1;
long int exponent = strtol (cp, NULL, 10);
- if (abs (exponent) > 999)
+ if (labs (exponent) > 999)
return false;
sprintf (cp, "%+04ld", exponent);
}
return digit >= '5';
}
-/* Initializes R for formatting the magnitude of NUMBER to no
+/* Initializes R for formatting the magnitude of NUMBER with STYLE to no
more than MAX_DECIMAL decimal places. */
static void
-rounder_init (struct rounder *r, double number, int max_decimals)
+rounder_init (struct rounder *r, const struct fmt_number_style *style,
+ double number, int max_decimals)
{
assert (fabs (number) < 1e41);
assert (max_decimals >= 0 && max_decimals <= 16);
We append ".00" to the integer representation because
round_up assumes that fractional digits are present. */
- sprintf (r->string, "%.0f.00", fabs (round (number)));
+ c_snprintf (r->string, 64, "%.0f.00", fabs (round (number)));
}
else
{
numbers does not hint how to do what we want, and it's
not obvious how to change their algorithms to do so. It
would also be a lot of work. */
- sprintf (r->string, "%.*f", max_decimals + 2, fabs (number));
+ c_snprintf (r->string, 64, "%.*f", max_decimals + 2, fabs (number));
if (!strcmp (r->string + strlen (r->string) - 2, "50"))
{
int binary_exponent, decimal_exponent, format_decimals;
decimal_exponent = binary_exponent * 3 / 10;
format_decimals = (DBL_DIG + 1) - decimal_exponent;
if (format_decimals > max_decimals + 2)
- sprintf (r->string, "%.*f", format_decimals, fabs (number));
+ c_snprintf (r->string, 64, "%.*f", format_decimals, fabs (number));
}
}
- if (r->string[0] == '0')
+ if (r->string[0] == '0' && !style->include_leading_zero)
memmove (r->string, &r->string[1], strlen (r->string));
r->leading_zeros = strspn (r->string, "0.");
r->leading_nines = strspn (r->string, "9.");
r->integer_digits = strchr (r->string, '.') - r->string;
+ assert (r->integer_digits < 64);
+ assert (r->integer_digits >= 0);
r->negative = number < 0;
}
if (number != SYSMIS
&& number >= 0.
&& number < power10 (digits)
- && sprintf (decimal, "%0*.0f", digits, round (number)) == digits)
+ && c_snprintf (decimal, 64, "%0*.0f", digits, round (number)) == digits)
{
const char *src = decimal;
int i;