/* PSPP - a program for statistical analysis.
- Copyright (C) 2006, 2009, 2010, 2011, 2012, 2013, 2014 Free Software Foundation, Inc.
+ Copyright (C) 2006, 2009, 2010, 2011, 2012, 2013, 2014, 2015,
+ 2016, 2021 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 <errno.h>
#include <iconv.h>
#include <langinfo.h>
-#include <libintl.h>
#include <locale.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unicase.h>
#include <unigbrk.h>
+#include <uniwidth.h>
#include "libpspp/assertion.h"
#include "libpspp/compiler.h"
#include "libpspp/hmapx.h"
#include "libpspp/hash-functions.h"
+#include "libpspp/misc.h"
#include "libpspp/pool.h"
#include "libpspp/str.h"
#include "libpspp/version.h"
+#include "gl/c-ctype.h"
#include "gl/c-strcase.h"
#include "gl/localcharset.h"
+#include <gl/localename.h>
#include "gl/minmax.h"
#include "gl/xalloc.h"
#include "gl/relocatable.h"
#define _(msgid) gettext (msgid)
struct converter
- {
- char *tocode;
- char *fromcode;
- iconv_t conv;
- int error;
- };
+{
+ char *tocode;
+ char *fromcode;
+ iconv_t conv;
+ int null_char_width;
+};
static char *default_encoding;
static struct hmapx map;
/* A wrapper around iconv_open */
static struct converter *
-create_iconv__ (const char* tocode, const char* fromcode)
+create_iconv (const char* tocode, const char* fromcode, bool warn)
{
size_t hash;
struct hmapx_node *node;
hash = hash_string (tocode, hash_string (fromcode, 0));
HMAPX_FOR_EACH_WITH_HASH (converter, node, hash, &map)
- if (!strcmp (tocode, converter->tocode)
- && !strcmp (fromcode, converter->fromcode))
- return converter;
+ {
+ if (!converter)
+ return NULL;
+
+ if (!strcmp (tocode, converter->tocode)
+ && !strcmp (fromcode, converter->fromcode))
+ return converter;
+ }
converter = xmalloc (sizeof *converter);
converter->tocode = xstrdup (tocode);
converter->fromcode = xstrdup (fromcode);
converter->conv = iconv_open (tocode, fromcode);
- converter->error = converter->conv == (iconv_t) -1 ? errno : 0;
- hmapx_insert (&map, converter, hash);
-
- return converter;
-}
-
-static iconv_t
-create_iconv (const char* tocode, const char* fromcode)
-{
- struct converter *converter;
-
- converter = create_iconv__ (tocode, fromcode);
-
+ int error = converter->conv == (iconv_t) ~0 ? errno : 0;
/* I don't think it's safe to translate this string or to use messaging
as the converters have not yet been set up */
- if (converter->error && strcmp (tocode, fromcode))
+ if (error && strcmp (tocode, fromcode))
{
- fprintf (stderr,
- "Warning: "
- "cannot create a converter for `%s' to `%s': %s\n",
- fromcode, tocode, strerror (converter->error));
- converter->error = 0;
+ if (warn)
+ fprintf (stderr,
+ "Warning: "
+ "cannot create a converter for `%s' to `%s': %s\n",
+ fromcode, tocode, strerror (error));
+
+ free (converter->tocode);
+ free (converter->fromcode);
+ free (converter);
+
+ hmapx_insert (&map, NULL, hash);
+ return NULL;
}
- return converter->conv;
+ /* Find out how many bytes there are in a null char in the target
+ encoding */
+ iconv_t bconv = iconv_open (tocode, "ASCII");
+ if (bconv != (iconv_t) -1)
+ {
+ ICONV_CONST char inbuf[1] = "";
+ ICONV_CONST char *inptr = inbuf;
+ size_t inbytes = sizeof inbuf;
+
+ char outbuf[8];
+ char *outptr = outbuf;
+ size_t outbytes = sizeof outbuf;
+ if (-1 != iconv (bconv, &inptr, &inbytes, &outptr, &outbytes))
+ converter->null_char_width = outptr - outbuf;
+ iconv_close (bconv);
+ }
+
+ hmapx_insert (&map, converter, hash);
+
+ return converter;
}
+
/* Converts the single byte C from encoding FROM to TO, returning the first
byte of the result.
Returns the output length if successful, -1 if the output buffer is too
small. */
static ssize_t
-try_recode (iconv_t conv, char fallbackchar,
+try_recode (struct converter *cvtr, char fallbackchar,
const char *in, size_t inbytes,
char *out_, size_t outbytes)
{
char *out = out_;
- int i;
+ int i, j;
+
+ int null_bytes = cvtr->null_char_width;
/* Put the converter into the initial shift state, in case there was any
state information left over from its last usage. */
- iconv (conv, NULL, 0, NULL, 0);
+ iconv (cvtr->conv, NULL, 0, NULL, 0);
/* Do two rounds of iconv() calls:
- The first round does the bulk of the conversion using the
- caller-supplied input data..
+ caller-supplied input data..
- The second round flushes any leftover output. This has a real effect
- with input encodings that use combining diacritics, e.g. without the
- second round the last character tends to gets dropped when converting
- from windows-1258 to other encodings.
+ with input encodings that use combining diacritics, e.g. without the
+ second round the last character tends to gets dropped when converting
+ from windows-1258 to other encodings.
*/
for (i = 0; i < 2; i++)
{
ICONV_CONST char **inp = i ? NULL : (ICONV_CONST char **) ∈
size_t *inbytesp = i ? NULL : &inbytes;
- while (iconv (conv, inp, inbytesp, &out, &outbytes) == -1)
+ while (iconv (cvtr->conv, inp, inbytesp, &out, &outbytes) == -1)
switch (errno)
{
case EINVAL:
- if (outbytes < 2)
+ if (outbytes < null_bytes + 1)
return -E2BIG;
if (!fallbackchar)
return -EINVAL;
*out++ = fallbackchar;
- *out = '\0';
- return out - out_;
+ for (j = 0 ; j < null_bytes ; ++j)
+ *out++ = '\0';
+ return out - 1 - out_;
case EILSEQ:
if (outbytes == 0)
}
}
- if (outbytes == 0)
+ if (outbytes <= null_bytes - 1)
return -E2BIG;
- *out = '\0';
- return out - out_;
+ for (i = 0 ; i < null_bytes ; ++i)
+ *out++ = '\0';
+
+ return out - 1 - out_;
}
/* Converts the string TEXT, which should be encoded in FROM-encoding, to a
{
struct substring out;
- if ( text == NULL )
+ if (text == NULL)
return NULL;
- if ( length == -1 )
- length = strlen (text);
+ if (length == -1)
+ length = strlen (text);
out = recode_substring_pool (to, from, ss_buffer (text, length), pool);
return out.string;
fits or until HEAD_LEN reaches 0.
[*] Actually this function drops grapheme clusters instead of characters, so
- that, e.g. a Unicode character followed by a combining accent character
- is either completely included or completely excluded from HEAD_LEN. See
- UAX #29 at http://unicode.org/reports/tr29/ for more information on
- grapheme clusters.
+ that, e.g. a Unicode character followed by a combining accent character
+ is either completely included or completely excluded from HEAD_LEN. See
+ UAX #29 at http://unicode.org/reports/tr29/ for more information on
+ grapheme clusters.
A null ENCODING is treated as UTF-8.
Simple examples for encoding="UTF-8", max_len=6:
- head="abc", tail="xyz" => 3
- head="abcd", tail="xyz" => 3 ("d" dropped).
- head="abc", tail="uvwxyz" => 0 ("abc" dropped).
- head="abc", tail="tuvwxyz" => 0 ("abc" dropped).
+ head="abc", tail="xyz" => 3
+ head="abcd", tail="xyz" => 3 ("d" dropped).
+ head="abc", tail="uvwxyz" => 0 ("abc" dropped).
+ head="abc", tail="tuvwxyz" => 0 ("abc" dropped).
Examples for encoding="ISO-8859-1", max_len=6:
- head="éèä", tail="xyz" => 6
- (each letter in head is only 1 byte in ISO-8859-1 even though they
- each take 2 bytes in UTF-8 encoding)
+ head="éèä", tail="xyz" => 6
+ (each letter in head is only 1 byte in ISO-8859-1 even though they
+ each take 2 bytes in UTF-8 encoding)
*/
static size_t
utf8_encoding_concat__ (const char *head, size_t head_len,
included, even if TAIL by itself is longer than MAX_LEN in ENCODING.
[*] Actually this function drops grapheme clusters instead of characters, so
- that, e.g. a Unicode character followed by a combining accent character
- is either completely included or completely excluded from the returned
- string. See UAX #29 at http://unicode.org/reports/tr29/ for more
- information on grapheme clusters.
+ that, e.g. a Unicode character followed by a combining accent character
+ is either completely included or completely excluded from the returned
+ string. See UAX #29 at http://unicode.org/reports/tr29/ for more
+ information on grapheme clusters.
A null ENCODING is treated as UTF-8.
Simple examples for encoding="UTF-8", max_len=6:
- head="abc", tail="xyz" => "abcxyz"
- head="abcd", tail="xyz" => "abcxyz"
- head="abc", tail="uvwxyz" => "uvwxyz"
- head="abc", tail="tuvwxyz" => "tuvwxyz"
+ head="abc", tail="xyz" => "abcxyz"
+ head="abcd", tail="xyz" => "abcxyz"
+ head="abc", tail="uvwxyz" => "uvwxyz"
+ head="abc", tail="tuvwxyz" => "tuvwxyz"
Examples for encoding="ISO-8859-1", max_len=6:
- head="éèä", tail="xyz" => "éèäxyz"
- (each letter in HEAD is only 1 byte in ISO-8859-1 even though they
- each take 2 bytes in UTF-8 encoding)
+ head="éèä", tail="xyz" => "éèäxyz"
+ (each letter in HEAD is only 1 byte in ISO-8859-1 even though they
+ each take 2 bytes in UTF-8 encoding)
*/
char *
utf8_encoding_concat (const char *head, const char *tail,
return prefix_len + tail_len;
}
+/* Returns the number of display columns that would be occupied by the LENGTH
+ bytes of UTF-8 starting at S. */
+size_t
+utf8_count_columns (const char *s_, size_t length)
+{
+ const uint8_t *s = CHAR_CAST (const uint8_t *, s_);
+
+ size_t columns = 0;
+ for (int ofs = 0; ofs < length; )
+ {
+ ucs4_t uc;
+ ofs += u8_mbtouc (&uc, s + ofs, length - ofs);
+ if (uc != '\t')
+ {
+ int width = uc_width (uc, "UTF-8");
+ if (width > 0)
+ columns += width;
+ }
+ else
+ columns = ROUND_UP (columns + 1, 8);
+ }
+ return columns;
+}
+
+/* Returns the byte offset in LENGTH-byte UTF-8 string S that is N_COLUMNS
+ display columns into the string. */
+size_t
+utf8_columns_to_bytes (const char *s_, size_t length, size_t n_columns)
+{
+ const uint8_t *s = CHAR_CAST (const uint8_t *, s_);
+
+ size_t columns = 0;
+ int ofs;
+ for (ofs = 0; ofs < length && columns < n_columns; )
+ {
+ ucs4_t uc;
+ ofs += u8_mbtouc (&uc, s + ofs, length - ofs);
+ if (uc != '\t')
+ {
+ int width = uc_width (uc, "UTF-8");
+ if (width > 0)
+ columns += width;
+ }
+ else
+ columns = ROUND_UP (columns + 1, 8);
+ }
+ return ofs;
+}
+
/* Returns an allocated, null-terminated string, owned by the caller,
containing as many characters[*] from the beginning of S that would fit
within MAX_LEN bytes if the returned string were to be re-encoded in
ENCODING. Both S and the returned string are encoded in UTF-8.
[*] Actually this function drops grapheme clusters instead of characters, so
- that, e.g. a Unicode character followed by a combining accent character
- is either completely included or completely excluded from the returned
- string. See UAX #29 at http://unicode.org/reports/tr29/ for more
- information on grapheme clusters.
+ that, e.g. a Unicode character followed by a combining accent character
+ is either completely included or completely excluded from the returned
+ string. See UAX #29 at http://unicode.org/reports/tr29/ for more
+ information on grapheme clusters.
A null ENCODING is treated as UTF-8.
*/
struct pool *pool, struct substring *out)
{
size_t bufsize;
- iconv_t conv ;
+ struct converter *conv;
if (to == NULL)
to = default_encoding;
if (from == NULL)
from = default_encoding;
- conv = create_iconv (to, from);
+ conv = create_iconv (to, from, true);
- if ( (iconv_t) -1 == conv )
+ if (NULL == conv)
{
if (fallbackchar)
{
i18n_init (void)
{
setlocale (LC_ALL, "");
- bindtextdomain (PACKAGE, relocate(locale_dir));
+ char *allocated;
+ bindtextdomain (PACKAGE, relocate2 (locale_dir, &allocated));
+ free (allocated);
textdomain (PACKAGE);
assert (default_encoding == NULL);
default_encoding = xstrdup (enc);
}
+/* Return the ISO two letter code for the current LC_MESSAGES
+ locale category. */
+char *
+get_language (void)
+{
+ const char *localename = gl_locale_name (LC_MESSAGES, "LC_MESSAGES");
+ if (0 == strcmp (localename, "C"))
+ return NULL;
+ char *ln = xstrdup (localename);
+ char *end = strchr (ln, '_');
+ if (end)
+ *end = '\0';
+ return ln;
+}
+
/* Attempts to set the encoding from a locale name
- returns true if successfull.
+ returns true if successful.
This function does not (should not!) alter the current locale.
*/
bool
loc_encoding = xstrdup (locale_charset ());
- if ( 0 == strcmp (loc_encoding, c_encoding))
+ if (0 == strcmp (loc_encoding, c_encoding))
{
ok = false;
}
HMAPX_FOR_EACH (cvtr, node, &map)
{
+ if (cvtr == NULL)
+ continue;
free (cvtr->tocode);
free (cvtr->fromcode);
if (cvtr->conv != (iconv_t) -1)
{
iconv_t conv = iconv_open (UTF8, enc);
- if ( conv == (iconv_t) -1)
+ if (conv == (iconv_t) -1)
return false;
iconv_close (conv);
/* Return the system local's idea of the
- decimal seperator character */
+ decimal separator character */
char
get_system_decimal (void)
{
unsigned int
utf8_hash_case_string (const char *s, unsigned int basis)
{
- return utf8_hash_case_bytes (s, strlen (s), basis);
+ return utf8_hash_case_substring (ss_cstr (s), basis);
+}
+
+/* Returns a hash value for UTF-8 string S, with lowercase and uppercase
+ letters treated as equal, starting from BASIS. */
+unsigned int
+utf8_hash_case_substring (struct substring s, unsigned int basis)
+{
+ return utf8_hash_case_bytes (s.string, s.length, basis);
}
/* Compares UTF-8 strings A and B case-insensitively.
int
utf8_strcasecmp (const char *a, const char *b)
{
- return utf8_strncasecmp (a, strlen (a), b, strlen (b));
+ return utf8_sscasecmp (ss_cstr (a), ss_cstr (b));
+}
+
+int
+utf8_sscasecmp (struct substring a, struct substring b)
+{
+ return utf8_strncasecmp (a.string, a.length, b.string, b.length);
}
/* Compares UTF-8 strings A (with length AN) and B (with length BN)
return result;
}
+static bool
+is_all_digits (const uint8_t *s, size_t len)
+{
+ for (size_t i = 0; i < len; i++)
+ if (!c_isdigit (s[i]))
+ return false;
+ return true;
+}
+
+/* Compares UTF-8 strings A and B case-insensitively. If the strings end in a
+ number, then they are compared numerically. Returns a negative value if A <
+ B, zero if A == B, positive if A > B. */
+int
+utf8_strverscasecmp (const char *a, const char *b)
+{
+ /* Normalize A. */
+ uint8_t a_stub[64];
+ size_t a_len = sizeof a_stub;
+ uint8_t *a_norm = u8_casefold (CHAR_CAST (uint8_t *, a), strlen (a), NULL,
+ UNINORM_NFKD, a_stub, &a_len);
+
+ /* Normalize B. */
+ uint8_t b_stub[64];
+ size_t b_len = sizeof b_stub;
+ uint8_t *b_norm = u8_casefold (CHAR_CAST (uint8_t *, b), strlen (b), NULL,
+ UNINORM_NFKD, b_stub, &b_len);
+
+ int result;
+ if (!a_norm || !b_norm)
+ {
+ result = strcmp (a, b);
+ goto exit;
+ }
+
+ size_t len = MIN (a_len, b_len);
+ for (size_t i = 0; i < len; i++)
+ if (a_norm[i] != b_norm[i])
+ {
+ /* If both strings end in digits, compare them numerically. */
+ if (is_all_digits (&a_norm[i], a_len - i)
+ && is_all_digits (&b_norm[i], b_len - i))
+ {
+ /* Start by stripping leading zeros, since those don't matter for
+ numerical comparison. */
+ size_t ap, bp;
+ for (ap = i; ap < a_len; ap++)
+ if (a_norm[ap] != '0')
+ break;
+ for (bp = i; bp < b_len; bp++)
+ if (b_norm[bp] != '0')
+ break;
+
+ /* The number with more digits, if there is one, is larger. */
+ size_t a_digits = a_len - ap;
+ size_t b_digits = b_len - bp;
+ if (a_digits != b_digits)
+ result = a_digits > b_digits ? 1 : -1;
+ else
+ result = memcmp (&a_norm[ap], &b_norm[bp], a_digits);
+ }
+ else
+ result = a_norm[i] > b_norm[i] ? 1 : -1;
+ goto exit;
+ }
+ result = a_len < b_len ? -1 : a_len > b_len;
+
+exit:
+ if (a_norm != a_stub)
+ free (a_norm);
+ if (b_norm != b_stub)
+ free (b_norm);
+ return result;
+}
+
static char *
utf8_casemap (const char *s,
uint8_t *(*f) (const uint8_t *, size_t, const char *, uninorm_t,
{
return utf8_casemap (s, u8_tolower);
}
+
+char *
+utf8_to_title (const char *s)
+{
+ return utf8_casemap (s, u8_totitle);
+}
\f
bool
get_encoding_info (struct encoding_info *e, const char *name)
{
const struct substring in = SS_LITERAL_INITIALIZER (
- "\t\n\v\f\r "
- "!\"#$%&'()*+,-./0123456789:;<=>?@"
- "ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_`"
- "abcdefghijklmnopqrstuvwxyz{|}~");
+ "\t\n\v\f\r "
+ "!\"#$%&'()*+,-./0123456789:;<=>?@"
+ "ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_`"
+ "abcdefghijklmnopqrstuvwxyz{|}~");
struct substring out, cr, lf, space;
bool ok;
bool
is_encoding_supported (const char *encoding)
{
- return (create_iconv__ ("UTF-8", encoding)->conv != (iconv_t) -1
- && create_iconv__ (encoding, "UTF-8")->conv != (iconv_t) -1);
+ return (create_iconv ("UTF-8", encoding, false)
+ && create_iconv (encoding, "UTF-8", false));
}
/* Returns true if E is the name of a UTF-8 encoding.
static int n_categories;
static void SENTINEL (0)
-add_category (size_t *allocated_categories, const char *category, ...)
+ add_category (size_t *allocated_categories, const char *category, ...)
{
struct encoding_category *c;
const char *encodings[16];