/* 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)
#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,
+ struct string *output, const char *output_encoding)
{
assert (fmt_check_output (format));
+ if (format->type == FMT_A)
+ {
+ char *in = CHAR_CAST (char *, value_str (input, format->w));
+ 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,
+ ds_put_uninit (output, format->w));
+ else
+ {
+ char *utf8_encoded = data_out (input, input_encoding, format);
+ char *output_encoded = recode_string (output_encoding, UTF8,
+ utf8_encoded, -1);
+ ds_put_cstr (output, output_encoded);
+ free (output_encoded);
+ free (utf8_encoded);
+ }
+}
+
+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;
- converters[format->type] (input, format, output);
- if (0 != strcmp (encoding, C_ENCODING)
- && fmt_get_category (format->type) != FMT_CAT_BINARY)
+ 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.
+ 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. 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)
{
- 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 *, value_str (input, format->w));
+ return recode_string_pool (UTF8, encoding, in, format->w, pool);
+ }
+ else if (fmt_get_category (format->type) == FMT_CAT_BINARY)
+ {
+ char tmp[16];
+
+ assert (format->w + 1 <= sizeof tmp);
+ converters[format->type] (input, format, tmp);
+ return binary_to_utf8 (tmp, pool);
+ }
+ else
+ {
+ const struct fmt_number_style *style = settings_get_style (format->type);
+ size_t size = format->w + style->extra_bytes + 1;
+ char *output;
+
+ output = pool_alloc_unaligned (pool, size);
+ converters[format->type] (input, format, output);
+ return output;
+ }
+}
- output = xmalloc (format->w + style->extra_bytes + 1);
+/* 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, struct pool *pool)
+{
+
+ if (fmt_get_category (format->type) & (FMT_CAT_BASIC | FMT_CAT_CUSTOM))
+ {
+ const struct fmt_number_style *style = settings_get_style (format->type);
+ struct fmt_spec wide_format;
+ char tmp[128];
+ size_t size;
- converters[format->type] (input, format, output);
+ wide_format.type = format->type;
+ wide_format.w = 40;
+ wide_format.d = format->d;
- t = recode_string_pool (UTF8, encoding, output, format->w, pool);
- free (output);
- return t;
+ size = format->w + style->extra_bytes + 1;
+ if (size <= sizeof tmp)
+ {
+ output_number (input, &wide_format, tmp);
+ return pool_strdup (pool, tmp + strspn (tmp, " "));
+ }
+ }
+
+ return data_out_pool (input, encoding, format, pool);
}
char *
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);
if (input->f == SYSMIS)
output_missing (format, output);
else if (fabs (number) < power10 (format->w)
- && sprintf (buf, "%0*.0f", format->w,
+ && c_snprintf (buf, 128, "%0*.0f", format->w,
fabs (round (number))) == format->w)
{
if (number < 0 && strspn (buf, "0") < format->w)
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);
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':
{
int d = MIN (format->d, excess_width - 4);
int w = d + 3;
- sprintf (p, ":%0*.*f", w, d, number);
+ c_snprintf (p, 64, ":%0*.*f", w, d, number);
if (settings_get_decimal_char (FMT_F) != '.')
{
char *cp = strchr (p, '.');
}
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 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, char *output UNUSED)
{
- memcpy (output, value_str (input, format->w), format->w);
- output[format->w] = '\0';
+ NOT_REACHED ();
}
/* Outputs AHEX format. */
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. */
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));
}
}
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;