/* PSPP - a program for statistical analysis.
- Copyright (C) 2006, 2009, 2010, 2011, 2012 Free Software Foundation, Inc.
+ Copyright (C) 2006, 2009, 2010, 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 <stdio.h>
#include <stdlib.h>
#include <string.h>
+#include <unicase.h>
#include <unigbrk.h>
#include "libpspp/assertion.h"
#include "gl/c-strcase.h"
#include "gl/localcharset.h"
+#include "gl/minmax.h"
#include "gl/xalloc.h"
#include "gl/relocatable.h"
#include "gl/xstrndup.h"
small. */
static ssize_t
try_recode (iconv_t conv,
- const char *ip, size_t inbytes,
- char *op_, size_t outbytes)
+ const char *in, size_t inbytes,
+ char *out_, size_t outbytes)
{
/* FIXME: Need to ensure that this char is valid in the target encoding */
const char fallbackchar = '?';
- char *op = op_;
+ char *out = out_;
+ int i;
/* 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);
- while (iconv (conv, (ICONV_CONST char **) &ip, &inbytes,
- &op, &outbytes) == -1)
- switch (errno)
- {
- case EINVAL:
- if (outbytes < 2)
- return -1;
- *op++ = fallbackchar;
- *op = '\0';
- return op - op_;
-
- case EILSEQ:
- if (outbytes == 0)
- return -1;
- *op++ = fallbackchar;
- outbytes--;
- ip++;
- inbytes--;
- break;
-
- case E2BIG:
- return -1;
-
- default:
- /* should never happen */
- fprintf (stderr, "Character conversion error: %s\n", strerror (errno));
- NOT_REACHED ();
- break;
- }
+ /* Do two rounds of iconv() calls:
+
+ - The first round does the bulk of the conversion using the
+ 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.
+ */
+ 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)
+ switch (errno)
+ {
+ case EINVAL:
+ if (outbytes < 2)
+ return -1;
+ *out++ = fallbackchar;
+ *out = '\0';
+ return out - out_;
+
+ case EILSEQ:
+ if (outbytes == 0)
+ return -1;
+ *out++ = fallbackchar;
+ outbytes--;
+ if (inp)
+ {
+ in++;
+ inbytes--;
+ }
+ break;
+
+ case E2BIG:
+ return -1;
+
+ default:
+ /* should never happen */
+ fprintf (stderr, "Character conversion error: %s\n",
+ strerror (errno));
+ NOT_REACHED ();
+ break;
+ }
+ }
if (outbytes == 0)
return -1;
- *op = '\0';
- return op - op_;
+ *out = '\0';
+ return out - out_;
}
/* Converts the string TEXT, which should be encoded in FROM-encoding, to a
dynamically allocated string in TO-encoding. Any characters which cannot be
converted will be represented by '?'.
- The returned string will be null-terminated and allocated on POOL.
+ The returned string will be null-terminated and allocated on POOL with
+ pool_malloc().
This function's behaviour differs from that of g_convert_with_fallback
provided by GLib. The GLib function will fail (returns NULL) if any part of
if ( (iconv_t) -1 == conv )
{
struct substring out;
- ss_alloc_substring_pool (&out, text, pool);
+
+ out.string = pool_malloc (pool, text.length + 1);
+ out.length = text.length;
+ memcpy (out.string, text.string, text.length);
+ out.string[out.length] = '\0';
return out;
}
return buffer;
}
\f
+/* UTF-8 functions that deal with uppercase/lowercase distinctions. */
+
+/* Returns a hash value for the N bytes of UTF-8 encoded data starting at S,
+ with lowercase and uppercase letters treated as equal, starting from
+ BASIS. */
+unsigned int
+utf8_hash_case_bytes (const char *s, size_t n, unsigned int basis)
+{
+ uint8_t folded_buf[2048];
+ size_t folded_len = sizeof folded_buf;
+ uint8_t *folded_s;
+ unsigned int hash;
+
+ folded_s = u8_casefold (CHAR_CAST (const uint8_t *, s), n,
+ NULL, UNINORM_NFKD, folded_buf, &folded_len);
+ if (folded_s != NULL)
+ {
+ hash = hash_bytes (folded_s, folded_len, basis);
+ if (folded_s != folded_buf)
+ free (folded_s);
+ }
+ else
+ {
+ if (errno == ENOMEM)
+ xalloc_die ();
+ hash = hash_bytes (s, n, basis);
+ }
+
+ return hash;
+}
+
+/* Returns a hash value for null-terminated UTF-8 string S, with lowercase and
+ uppercase letters treated as equal, starting from BASIS. */
+unsigned int
+utf8_hash_case_string (const char *s, unsigned int basis)
+{
+ return utf8_hash_case_bytes (s, strlen (s), basis);
+}
+
+/* Compares UTF-8 strings A and B case-insensitively.
+ Returns a negative value if A < B, zero if A == B, positive if A > B. */
+int
+utf8_strcasecmp (const char *a, const char *b)
+{
+ return utf8_strncasecmp (a, strlen (a), b, strlen (b));
+}
+
+/* Compares UTF-8 strings A (with length AN) and B (with length BN)
+ case-insensitively.
+ Returns a negative value if A < B, zero if A == B, positive if A > B. */
+int
+utf8_strncasecmp (const char *a, size_t an, const char *b, size_t bn)
+{
+ int result;
+
+ if (u8_casecmp (CHAR_CAST (const uint8_t *, a), an,
+ CHAR_CAST (const uint8_t *, b), bn,
+ NULL, UNINORM_NFKD, &result))
+ {
+ if (errno == ENOMEM)
+ xalloc_die ();
+
+ result = memcmp (a, b, MIN (an, bn));
+ if (result == 0)
+ result = an < bn ? -1 : an > bn;
+ }
+
+ return result;
+}
+
+static char *
+utf8_casemap (const char *s,
+ uint8_t *(*f) (const uint8_t *, size_t, const char *, uninorm_t,
+ uint8_t *, size_t *))
+{
+ char *result;
+ size_t size;
+
+ result = CHAR_CAST (char *,
+ f (CHAR_CAST (const uint8_t *, s), strlen (s) + 1,
+ NULL, NULL, NULL, &size));
+ if (result == NULL)
+ {
+ if (errno == ENOMEM)
+ xalloc_die ();
+
+ result = xstrdup (s);
+ }
+ return result;
+}
+
+char *
+utf8_to_upper (const char *s)
+{
+ return utf8_casemap (s, u8_toupper);
+}
+
+char *
+utf8_to_lower (const char *s)
+{
+ return utf8_casemap (s, u8_tolower);
+}
+\f
bool
get_encoding_info (struct encoding_info *e, const char *name)
{