/* PSPP - a program for statistical analysis.
- Copyright (C) 2006, 2009, 2010, 2011 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 "libpspp/compiler.h"
#include "libpspp/hmapx.h"
#include "libpspp/hash-functions.h"
#include "libpspp/pool.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"
+#include "gettext.h"
+#define _(msgid) gettext (msgid)
+
struct converter
{
char *tocode;
char *fromcode;
iconv_t conv;
+ int error;
};
static char *default_encoding;
static struct hmapx map;
/* A wrapper around iconv_open */
-static iconv_t
-create_iconv (const char* tocode, const char* fromcode)
+static struct converter *
+create_iconv__ (const char* tocode, const char* fromcode)
{
size_t hash;
struct hmapx_node *node;
HMAPX_FOR_EACH_WITH_HASH (converter, node, hash, &map)
if (!strcmp (tocode, converter->tocode)
&& !strcmp (fromcode, converter->fromcode))
- return converter->conv;
+ 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);
+
/* 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 ( (iconv_t) -1 == converter->conv && 0 != strcmp (tocode, fromcode))
+ if (converter->error && strcmp (tocode, fromcode))
{
- const int err = errno;
fprintf (stderr,
"Warning: "
"cannot create a converter for `%s' to `%s': %s\n",
- fromcode, tocode, strerror (err));
+ fromcode, tocode, strerror (converter->error));
+ converter->error = 0;
}
return converter->conv;
Returns the output length if successful, -1 if the output buffer is too
small. */
static ssize_t
-try_recode (iconv_t conv,
- const char *ip, size_t inbytes,
- char *op_, size_t outbytes)
+try_recode (iconv_t conv, char fallbackchar,
+ 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 -E2BIG;
+ if (!fallbackchar)
+ return -EINVAL;
+ *out++ = fallbackchar;
+ *out = '\0';
+ return out - out_;
+
+ case EILSEQ:
+ if (outbytes == 0)
+ return -E2BIG;
+ if (!fallbackchar)
+ return -EILSEQ;
+ *out++ = fallbackchar;
+ outbytes--;
+ if (inp)
+ {
+ in++;
+ inbytes--;
+ }
+ break;
+
+ case E2BIG:
+ return -E2BIG;
+
+ default:
+ /* should never happen */
+ fprintf (stderr, "Character conversion error: %s\n",
+ strerror (errno));
+ NOT_REACHED ();
+ break;
+ }
+ }
if (outbytes == 0)
- return -1;
+ return -E2BIG;
- *op = '\0';
- return op - op_;
+ *out = '\0';
+ return out - out_;
}
/* Converts the string TEXT, which should be encoded in FROM-encoding, to a
else
{
size_t copy_len;
- size_t prev;
+ ucs4_t prev;
size_t ofs;
int mblen;
{
bool correct_result = false;
size_t copy_len;
- size_t prev;
+ ucs4_t prev;
size_t ofs;
int mblen;
return recode_string ("UTF-8", filename_encoding (), filename, -1);
}
-/* 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.
-
- 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
- the input string is not valid in the declared input encoding. This function
- however perseveres even in the presence of badly encoded input. */
-struct substring
-recode_substring_pool (const char *to, const char *from,
- struct substring text, struct pool *pool)
+static int
+recode_substring_pool__ (const char *to, const char *from,
+ struct substring text, char fallbackchar,
+ struct pool *pool, struct substring *out)
{
- size_t outbufferlength;
+ size_t bufsize;
iconv_t conv ;
if (to == NULL)
if ( (iconv_t) -1 == conv )
{
- struct substring out;
- ss_alloc_substring_pool (&out, text, pool);
- return out;
+ if (fallbackchar)
+ {
+ 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 0;
+ }
+ else
+ return EPROTO;
}
- for ( outbufferlength = 1 ; outbufferlength != 0; outbufferlength <<= 1 )
- if ( outbufferlength > text.length)
- {
- char *output = pool_malloc (pool, outbufferlength);
- ssize_t output_len = try_recode (conv, text.string, text.length,
- output, outbufferlength);
- if (output_len >= 0)
- return ss_buffer (output, output_len);
- pool_free (pool, output);
- }
+ for (bufsize = text.length + 1; bufsize > text.length; bufsize *= 2)
+ {
+ char *output = pool_malloc (pool, bufsize);
+ ssize_t retval;
+
+ retval = try_recode (conv, fallbackchar, text.string, text.length,
+ output, bufsize);
+ if (retval >= 0)
+ {
+ *out = ss_buffer (output, retval);
+ return 0;
+ }
+ pool_free (pool, output);
+
+ if (retval != -E2BIG)
+ return -retval;
+ }
NOT_REACHED ();
}
+/* 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 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
+ the input string is not valid in the declared input encoding. This function
+ however perseveres even in the presence of badly encoded input. */
+struct substring
+recode_substring_pool (const char *to, const char *from,
+ struct substring text, struct pool *pool)
+{
+ struct substring out;
+
+ recode_substring_pool__ (to, from, text, '?', pool, &out);
+ return out;
+}
+
+/* Converts the string TEXT, which should be encoded in FROM-encoding, to a
+ dynamically allocated string in TO-encoding. On success, returns 0, and the
+ converted null-terminated string, allocated from POOL with pool_malloc(), is
+ stored in *OUT. On failure, returns a positive errno value.
+
+ The function fails with an error if any part of the input string is not
+ valid in the declared input encoding. */
+int
+recode_pedantically (const char *to, const char *from,
+ struct substring text, struct pool *pool,
+ struct substring *out)
+{
+ int error;
+
+ error = recode_substring_pool__ (to, from, text, 0, pool, out);
+ if (error)
+ *out = ss_empty ();
+ return error;
+}
+\f
void
i18n_init (void)
{
- setlocale (LC_CTYPE, "");
- setlocale (LC_MESSAGES, "");
-#if HAVE_LC_PAPER
- setlocale (LC_PAPER, "");
-#endif
+ setlocale (LC_ALL, "");
bindtextdomain (PACKAGE, relocate(locale_dir));
textdomain (PACKAGE);
ok = false;
}
-
setlocale (LC_CTYPE, tmp);
free (tmp);
{
free (cvtr->tocode);
free (cvtr->fromcode);
- iconv_close (cvtr->conv);
+ if (cvtr->conv != (iconv_t) -1)
+ iconv_close (cvtr->conv);
free (cvtr);
}
{
char radix_char;
- char *ol = xstrdup (setlocale (LC_NUMERIC, NULL));
- setlocale (LC_NUMERIC, "");
-
#if HAVE_NL_LANGINFO
radix_char = nl_langinfo (RADIXCHAR)[0];
#else
}
#endif
- /* We MUST leave LC_NUMERIC untouched, since it would
- otherwise interfere with data_{in,out} */
- setlocale (LC_NUMERIC, ol);
- free (ol);
return radix_char;
}
snprintf (buffer, 16, "U+%04X", uc);
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)
+{
+ const struct substring in = SS_LITERAL_INITIALIZER (
+ "\t\n\v\f\r "
+ "!\"#$%&'()*+,-./0123456789:;<=>?@"
+ "ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_`"
+ "abcdefghijklmnopqrstuvwxyz{|}~");
+
+ struct substring out, cr, lf, space;
+ bool ok;
+
+ memset (e, 0, sizeof *e);
+
+ cr = recode_substring_pool (name, "UTF-8", ss_cstr ("\r"), NULL);
+ lf = recode_substring_pool (name, "UTF-8", ss_cstr ("\n"), NULL);
+ space = recode_substring_pool (name, "UTF-8", ss_cstr (" "), NULL);
+ ok = (cr.length >= 1
+ && cr.length <= MAX_UNIT
+ && cr.length == lf.length
+ && cr.length == space.length);
+ if (!ok)
+ {
+ fprintf (stderr, "warning: encoding `%s' is not supported.\n", name);
+ ss_dealloc (&cr);
+ ss_dealloc (&lf);
+ ss_dealloc (&space);
+ ss_alloc_substring (&cr, ss_cstr ("\r"));
+ ss_alloc_substring (&lf, ss_cstr ("\n"));
+ ss_alloc_substring (&space, ss_cstr (" "));
+ }
+
+ e->unit = cr.length;
+ memcpy (e->cr, cr.string, e->unit);
+ memcpy (e->lf, lf.string, e->unit);
+ memcpy (e->space, space.string, e->unit);
+
+ ss_dealloc (&cr);
+ ss_dealloc (&lf);
+ ss_dealloc (&space);
+
+ out = recode_substring_pool ("UTF-8", name, in, NULL);
+ e->is_ascii_compatible = ss_equals (in, out);
+ ss_dealloc (&out);
+
+ if (!e->is_ascii_compatible && e->unit == 1)
+ {
+ out = recode_substring_pool ("UTF-8", name, ss_cstr ("A"), NULL);
+ e->is_ebcdic_compatible = (out.length == 1
+ && (uint8_t) out.string[0] == 0xc1);
+ ss_dealloc (&out);
+ }
+ else
+ e->is_ebcdic_compatible = false;
+
+ return ok;
+}
+
+bool
+is_encoding_ascii_compatible (const char *encoding)
+{
+ struct encoding_info e;
+
+ get_encoding_info (&e, encoding);
+ return e.is_ascii_compatible;
+}
+
+bool
+is_encoding_ebcdic_compatible (const char *encoding)
+{
+ struct encoding_info e;
+
+ get_encoding_info (&e, encoding);
+ return e.is_ebcdic_compatible;
+}
+
+/* Returns true if iconv can convert ENCODING to and from UTF-8,
+ otherwise false. */
+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);
+}
+
+/* Returns true if E is the name of a UTF-8 encoding.
+
+ XXX Possibly we should test not E as a string but its properties via
+ iconv. */
+bool
+is_encoding_utf8 (const char *e)
+{
+ return ((e[0] == 'u' || e[0] == 'U')
+ && (e[1] == 't' || e[1] == 'T')
+ && (e[2] == 'f' || e[2] == 'F')
+ && ((e[3] == '8' && e[4] == '\0')
+ || (e[3] == '-' && e[4] == '8' && e[5] == '\0')));
+}
+\f
+static struct encoding_category *categories;
+static int n_categories;
+
+static void SENTINEL (0)
+add_category (size_t *allocated_categories, const char *category, ...)
+{
+ struct encoding_category *c;
+ const char *encodings[16];
+ va_list args;
+ int i, n;
+
+ /* Count encoding arguments. */
+ va_start (args, category);
+ n = 0;
+ while ((encodings[n] = va_arg (args, const char *)) != NULL)
+ {
+ const char *encoding = encodings[n];
+ if (!strcmp (encoding, "Auto") || is_encoding_supported (encoding))
+ n++;
+ }
+ assert (n < sizeof encodings / sizeof *encodings);
+ va_end (args);
+
+ if (n == 0)
+ return;
+
+ if (n_categories >= *allocated_categories)
+ categories = x2nrealloc (categories,
+ allocated_categories, sizeof *categories);
+
+ c = &categories[n_categories++];
+ c->category = category;
+ c->encodings = xmalloc (n * sizeof *c->encodings);
+ for (i = 0; i < n; i++)
+ c->encodings[i] = encodings[i];
+ c->n_encodings = n;
+}
+
+static void
+init_encoding_categories (void)
+{
+ static bool inited;
+ size_t alloc;
+
+ if (inited)
+ return;
+ inited = true;
+
+ alloc = 0;
+ add_category (&alloc, "Unicode", "UTF-8", "UTF-16", "UTF-16BE", "UTF-16LE",
+ "UTF-32", "UTF-32BE", "UTF-32LE", NULL_SENTINEL);
+ add_category (&alloc, _("Arabic"), "IBM864", "ISO-8859-6", "Windows-1256",
+ NULL_SENTINEL);
+ add_category (&alloc, _("Armenian"), "ARMSCII-8", NULL_SENTINEL);
+ add_category (&alloc, _("Baltic"), "ISO-8859-13", "ISO-8859-4",
+ "Windows-1257", NULL_SENTINEL);
+ add_category (&alloc, _("Celtic"), "ISO-8859-14", NULL_SENTINEL);
+ add_category (&alloc, _("Central European"), "IBM852", "ISO-8859-2",
+ "Mac-CentralEurope", "Windows-1250", NULL_SENTINEL);
+ add_category (&alloc, _("Chinese Simplified"), "GB18030", "GB2312", "GBK",
+ "HZ-GB-2312", "ISO-2022-CN", NULL_SENTINEL);
+ add_category (&alloc, _("Chinese Traditional"), "Big5", "Big5-HKSCS",
+ "EUC-TW", NULL_SENTINEL);
+ add_category (&alloc, _("Croatian"), "MacCroatian", NULL_SENTINEL);
+ add_category (&alloc, _("Cyrillic"), "IBM855", "ISO-8859-5", "ISO-IR-111",
+ "KOI8-R", "MacCyrillic", NULL_SENTINEL);
+ add_category (&alloc, _("Cyrillic/Russian"), "IBM866", NULL_SENTINEL);
+ add_category (&alloc, _("Cyrillic/Ukrainian"), "KOI8-U", "MacUkrainian",
+ NULL_SENTINEL);
+ add_category (&alloc, _("Georgian"), "GEOSTD8", NULL_SENTINEL);
+ add_category (&alloc, _("Greek"), "ISO-8859-7", "MacGreek", NULL_SENTINEL);
+ add_category (&alloc, _("Gujarati"), "MacGujarati", NULL_SENTINEL);
+ add_category (&alloc, _("Gurmukhi"), "MacGurmukhi", NULL_SENTINEL);
+ add_category (&alloc, _("Hebrew"), "IBM862", "ISO-8859-8-I", "Windows-1255",
+ NULL_SENTINEL);
+ add_category (&alloc, _("Hebrew Visual"), "ISO-8859-8", NULL_SENTINEL);
+ add_category (&alloc, _("Hindi"), "MacDevangari", NULL_SENTINEL);
+ add_category (&alloc, _("Icelandic"), "MacIcelandic", NULL_SENTINEL);
+ add_category (&alloc, _("Japanese"), "EUC-JP", "ISO-2022-JP", "Shift_JIS",
+ NULL_SENTINEL);
+ add_category (&alloc, _("Korean"), "EUC-KR", "ISO-2022-KR", "JOHAB", "UHC",
+ NULL_SENTINEL);
+ add_category (&alloc, _("Nordic"), "ISO-8859-10", NULL_SENTINEL);
+ add_category (&alloc, _("Romanian"), "ISO-8859-16", "MacRomanian",
+ NULL_SENTINEL);
+ add_category (&alloc, _("South European"), "ISO-8859-3", NULL_SENTINEL);
+ add_category (&alloc, _("Thai"), "ISO-8859-11", "TIS-620", "Windows-874",
+ NULL_SENTINEL);
+ add_category (&alloc, _("Turkish"), "IBM857", "ISO-8859-9", "Windows-1254",
+ NULL_SENTINEL);
+ add_category (&alloc, _("Vietnamese"), "TVCN", "VISCII", "VPS",
+ "Windows-1258", NULL_SENTINEL);
+ add_category (&alloc, _("Western European"), "ISO-8859-1", "ISO-8859-15",
+ "Windows-1252", "IBM850", "MacRoman", NULL_SENTINEL);
+}
+
+/* Returns an array of "struct encoding_category" that contains only the
+ categories and encodings that the system supports. */
+struct encoding_category *
+get_encoding_categories (void)
+{
+ init_encoding_categories ();
+ return categories;
+}
+
+/* Returns the number of elements in the array returned by
+ get_encoding_categories(). */
+size_t
+get_n_encoding_categories (void)
+{
+ init_encoding_categories ();
+ return n_categories;
+}