X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flibpspp%2Fi18n.c;h=fbff1772a18b2748e21b2cb4a61a3b2f09734dae;hb=213a9dde780d4b31cdcf7dd14f5015633a89cec5;hp=db851217c80be27ae129a9c6252824929af51785;hpb=45a1201de864c76f5a8c78166dc4ff2ea4151e5d;p=pspp-builds.git

diff --git a/src/libpspp/i18n.c b/src/libpspp/i18n.c
index db851217..fbff1772 100644
--- a/src/libpspp/i18n.c
+++ b/src/libpspp/i18n.c
@@ -1,5 +1,5 @@
 /* PSPP - a program for statistical analysis.
-   Copyright (C) 2006 Free Software Foundation, Inc.
+   Copyright (C) 2006, 2009, 2010, 2011 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
@@ -15,210 +15,626 @@
    along with this program.  If not, see <http://www.gnu.org/licenses/>. */
 
 #include <config.h>
-#include <xalloc.h>
+
+#include "libpspp/i18n.h"
+
 #include <assert.h>
+#include <errno.h>
+#include <iconv.h>
+#include <langinfo.h>
+#include <libintl.h>
 #include <locale.h>
-#include <stdlib.h>
 #include <stdio.h>
+#include <stdlib.h>
 #include <string.h>
-#include <iconv.h>
-#include <errno.h>
-#include "assertion.h"
-
-#include "i18n.h"
-
-#include <localcharset.h>
-#include "xstrndup.h"
-
-#if HAVE_NL_LANGINFO
-#include <langinfo.h>
-#endif
-
-
-static char *locale = 0;
-static const char *charset;
-
-
-static iconv_t convertor[n_CONV];
-
+#include <unigbrk.h>
+
+#include "libpspp/assertion.h"
+#include "libpspp/hmapx.h"
+#include "libpspp/hash-functions.h"
+#include "libpspp/pool.h"
+#include "libpspp/str.h"
+#include "libpspp/version.h"
+
+#include "gl/c-strcase.h"
+#include "gl/localcharset.h"
+#include "gl/xalloc.h"
+#include "gl/relocatable.h"
+#include "gl/xstrndup.h"
+
+struct converter
+ {
+    char *tocode;
+    char *fromcode;
+    iconv_t conv;
+  };
+
+static char *default_encoding;
+static struct hmapx map;
 
 /* A wrapper around iconv_open */
 static iconv_t
 create_iconv (const char* tocode, const char* fromcode)
 {
-  iconv_t conv = iconv_open (tocode, fromcode);
+  size_t hash;
+  struct hmapx_node *node;
+  struct converter *converter;
+  assert (fromcode);
+
+  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->conv;
+
+  converter = xmalloc (sizeof *converter);
+  converter->tocode = xstrdup (tocode);
+  converter->fromcode = xstrdup (fromcode);
+  converter->conv = iconv_open (tocode, fromcode);
+  hmapx_insert (&map, converter, hash);
 
   /* I don't think it's safe to translate this string or to use messaging
-     as the convertors have not yet been set up */
-  if ( (iconv_t) -1 == conv && 0 != strcmp (tocode, fromcode))
+     as the converters have not yet been set up */
+  if ( (iconv_t) -1 == converter->conv && 0 != strcmp (tocode, fromcode))
     {
       const int err = errno;
       fprintf (stderr,
-	"Warning: cannot create a convertor for \"%s\" to \"%s\": %s\n",
-	fromcode, tocode, strerror (err));
+               "Warning: "
+               "cannot create a converter for `%s' to `%s': %s\n",
+               fromcode, tocode, strerror (err));
     }
 
-  return conv;
+  return converter->conv;
 }
 
-/* Return a string based on TEXT converted according to HOW.
-   If length is not -1, then it must be the number of bytes in TEXT.
-   The returned string must be freed when no longer required.
-*/
+/* Converts the single byte C from encoding FROM to TO, returning the first
+   byte of the result.
+
+   This function probably shouldn't be used at all, but some code still does
+   use it. */
+char
+recode_byte (const char *to, const char *from, char c)
+{
+  char x;
+  char *s = recode_string (to, from, &c, 1);
+  x = s[0];
+  free (s);
+  return x;
+}
+
+/* Similar to recode_string_pool, but allocates the returned value on the heap
+   instead of in a pool.  It is the caller's responsibility to free the
+   returned value. */
 char *
-recode_string (enum conv_id how,  const char *text, int length)
+recode_string (const char *to, const char *from,
+	       const char *text, int length)
 {
-  char *outbuf = 0;
-  size_t outbufferlength;
-  size_t result;
-  char *op ;
-  size_t inbytes = 0;
-  size_t outbytes ;
+  return recode_string_pool (to, from, text, length, NULL);
+}
 
+/* Returns the length, in bytes, of the string that a similar recode_string()
+   call would return. */
+size_t
+recode_string_len (const char *to, const char *from,
+                   const char *text, int length)
+{
+  char *s = recode_string (to, from, text, length);
+  size_t len = strlen (s);
+  free (s);
+  return len;
+}
+
+/* Uses CONV to convert the INBYTES starting at IP into the OUTBYTES starting
+   at OP, and appends a null terminator to the output.
+
+   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)
+{
   /* FIXME: Need to ensure that this char is valid in the target encoding */
   const char fallbackchar = '?';
+  char *op = op_;
+
+  /* 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;
+      }
+
+  if (outbytes == 0)
+    return -1;
+
+  *op = '\0';
+  return op - op_;
+}
+
+/* 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 '?'.
+
+   LENGTH should be the length of the string or -1, if null terminated.
+
+   The returned string will be 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. */
+char *
+recode_string_pool (const char *to, const char *from,
+                    const char *text, int length, struct pool *pool)
+{
+  struct substring out;
 
   if ( text == NULL )
     return NULL;
 
   if ( length == -1 )
-     length = strlen(text);
+     length = strlen (text);
 
-  assert (how < n_CONV);
+  out = recode_substring_pool (to, from, ss_buffer (text, length), pool);
+  return out.string;
+}
 
-  if (convertor[how] == (iconv_t) -1)
-    return xstrndup (text, length);
+/* Returns the name of the encoding that should be used for file names.
 
-  for ( outbufferlength = 1 ; outbufferlength != 0; outbufferlength <<= 1 )
-    if ( outbufferlength > length)
-      break;
+   This is meant to be the same encoding used by g_filename_from_uri() and
+   g_filename_to_uri() in GLib. */
+static const char *
+filename_encoding (void)
+{
+#if defined _WIN32 || defined __WIN32__
+  return "UTF-8";
+#else
+  return locale_charset ();
+#endif
+}
+
+static char *
+xconcat2 (const char *a, size_t a_len,
+          const char *b, size_t b_len)
+{
+  char *s = xmalloc (a_len + b_len + 1);
+  memcpy (s, a, a_len);
+  memcpy (s + a_len, b, b_len);
+  s[a_len + b_len] = '\0';
+  return s;
+}
 
-  outbuf = xmalloc(outbufferlength);
-  op = outbuf;
+/* Conceptually, this function concatenates HEAD_LEN-byte string HEAD and
+   TAIL_LEN-byte string TAIL, both encoded in UTF-8, then converts them to
+   ENCODING.  If the re-encoded result is no more than MAX_LEN bytes long, then
+   it returns HEAD_LEN.  Otherwise, it drops one character[*] from the end of
+   HEAD and tries again, repeating as necessary until the concatenated result
+   fits or until HEAD_LEN reaches 0.
 
-  outbytes = outbufferlength;
-  inbytes = length;
+   [*] 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.
 
-  do {
-    const char *ip = text;
-    result = iconv (convertor[how], (ICONV_CONST char **) &text, &inbytes,
-		   &op, &outbytes);
+   A null ENCODING is treated as UTF-8.
 
-    if ( -1 == result )
-      {
-	int the_error = errno;
-
-	switch (the_error)
-	  {
-	  case EILSEQ:
-	  case EINVAL:
-	    if ( outbytes > 0 )
-	      {
-		*op++ = fallbackchar;
-		outbytes--;
-		text++;
-		inbytes--;
-		break;
-	      }
-	    /* Fall through */
-	  case E2BIG:
-	    free (outbuf);
-	    outbufferlength <<= 1;
-	    outbuf = xmalloc (outbufferlength);
-	    op = outbuf;
-	    outbytes = outbufferlength;
-	    inbytes = length;
-	    text = ip;
-	    break;
-	  default:
-	    /* should never happen */
-	    NOT_REACHED ();
-	    break;
-	  }
-      }
-  } while ( -1 == result );
+   Sometimes this function has to actually construct the concatenated string to
+   measure its length.  When this happens, it sets *RESULTP to that
+   null-terminated string, allocated with malloc(), for the caller to use if it
+   needs it.  Otherwise, it sets *RESULTP to NULL.
 
-  if (outbytes == 0 )
-    {
-      char *const oldaddr = outbuf;
-      outbuf = xrealloc (outbuf, outbufferlength + 1);
+   Simple examples for encoding="UTF-8", max_len=6:
 
-      op += (outbuf - oldaddr) ;
+       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)
+*/
+static size_t
+utf8_encoding_concat__ (const char *head, size_t head_len,
+                        const char *tail, size_t tail_len,
+                        const char *encoding, size_t max_len,
+                        char **resultp)
+{
+  *resultp = NULL;
+  if (head_len == 0)
+    return 0;
+  else if (encoding == NULL || !c_strcasecmp (encoding, "UTF-8"))
+    {
+      if (head_len + tail_len <= max_len)
+        return head_len;
+      else if (tail_len >= max_len)
+        return 0;
+      else
+        {
+          size_t copy_len;
+          size_t prev;
+          size_t ofs;
+          int mblen;
+
+          copy_len = 0;
+          for (ofs = u8_mbtouc (&prev, CHAR_CAST (const uint8_t *, head),
+                                head_len);
+               ofs <= max_len - tail_len;
+               ofs += mblen)
+            {
+              ucs4_t next;
+
+              mblen = u8_mbtouc (&next,
+                                 CHAR_CAST (const uint8_t *, head + ofs),
+                                 head_len - ofs);
+              if (uc_is_grapheme_break (prev, next))
+                copy_len = ofs;
+
+              prev = next;
+            }
+
+          return copy_len;
+        }
     }
+  else
+    {
+      char *result;
+
+      result = (tail_len > 0
+                ? xconcat2 (head, head_len, tail, tail_len)
+                : CONST_CAST (char *, head));
+      if (recode_string_len (encoding, "UTF-8", result,
+                             head_len + tail_len) <= max_len)
+        {
+          *resultp = result != head ? result : NULL;
+          return head_len;
+        }
+      else
+        {
+          bool correct_result = false;
+          size_t copy_len;
+          size_t prev;
+          size_t ofs;
+          int mblen;
+
+          copy_len = 0;
+          for (ofs = u8_mbtouc (&prev, CHAR_CAST (const uint8_t *, head),
+                                head_len);
+               ofs <= head_len;
+               ofs += mblen)
+            {
+              ucs4_t next;
+
+              mblen = u8_mbtouc (&next,
+                                 CHAR_CAST (const uint8_t *, head + ofs),
+                                 head_len - ofs);
+              if (uc_is_grapheme_break (prev, next))
+                {
+                  if (result != head)
+                    {
+                      memcpy (result, head, ofs);
+                      memcpy (result + ofs, tail, tail_len);
+                      result[ofs + tail_len] = '\0';
+                    }
+
+                  if (recode_string_len (encoding, "UTF-8", result,
+                                         ofs + tail_len) <= max_len)
+                    {
+                      correct_result = true;
+                      copy_len = ofs;
+                    }
+                  else
+                    correct_result = false;
+                }
+
+              prev = next;
+            }
+
+          if (result != head)
+            {
+              if (correct_result)
+                *resultp = result;
+              else
+                free (result);
+            }
+
+          return copy_len;
+        }
+    }
+}
 
-  *op = '\0';
+/* Concatenates a prefix of HEAD with all of TAIL and returns the result as a
+   null-terminated string owned by the caller.  HEAD, TAIL, and the returned
+   string are all encoded in UTF-8.  As many characters[*] from the beginning
+   of HEAD are included as will fit within MAX_LEN bytes supposing that the
+   resulting string were to be re-encoded in ENCODING.  All of TAIL is always
+   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.
+
+   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"
+
+   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)
+*/
+char *
+utf8_encoding_concat (const char *head, const char *tail,
+                      const char *encoding, size_t max_len)
+{
+  size_t tail_len = strlen (tail);
+  size_t prefix_len;
+  char *result;
+
+  prefix_len = utf8_encoding_concat__ (head, strlen (head), tail, tail_len,
+                                       encoding, max_len, &result);
+  return (result != NULL
+          ? result
+          : xconcat2 (head, prefix_len, tail, tail_len));
+}
 
-  return outbuf;
+/* Returns the length, in bytes, of the string that would be returned by
+   utf8_encoding_concat() if passed the same arguments, but the implementation
+   is often more efficient. */
+size_t
+utf8_encoding_concat_len (const char *head, const char *tail,
+                          const char *encoding, size_t max_len)
+{
+  size_t tail_len = strlen (tail);
+  size_t prefix_len;
+  char *result;
+
+  prefix_len = utf8_encoding_concat__ (head, strlen (head), tail, tail_len,
+                                       encoding, max_len, &result);
+  free (result);
+  return prefix_len + tail_len;
 }
 
+/* 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.
 
-/* Returns the current PSPP locale */
-const char *
-get_pspp_locale (void)
+   [*] 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.
+
+   A null ENCODING is treated as UTF-8.
+*/
+char *
+utf8_encoding_trunc (const char *s, const char *encoding, size_t max_len)
 {
-  assert (locale);
-  return locale;
+  return utf8_encoding_concat (s, "", encoding, max_len);
 }
 
-/* Set the PSPP locale */
-void
-set_pspp_locale (const char *l)
+/* Returns the length, in bytes, of the string that would be returned by
+   utf8_encoding_trunc() if passed the same arguments, but the implementation
+   is often more efficient. */
+size_t
+utf8_encoding_trunc_len (const char *s, const char *encoding, size_t max_len)
+{
+  return utf8_encoding_concat_len (s, "", encoding, max_len);
+}
+
+/* Returns FILENAME converted from UTF-8 to the filename encoding.
+   On Windows the filename encoding is UTF-8; elsewhere it is based on the
+   current locale. */
+char *
+utf8_to_filename (const char *filename)
 {
-  char *current_locale;
-  const char *current_charset;
+  return recode_string (filename_encoding (), "UTF-8", filename, -1);
+}
+
+/* Returns FILENAME converted from the filename encoding to UTF-8.
+   On Windows the filename encoding is UTF-8; elsewhere it is based on the
+   current locale. */
+char *
+filename_to_utf8 (const char *filename)
+{
+  return recode_string ("UTF-8", filename_encoding (), filename, -1);
+}
 
-  free(locale);
-  locale = strdup(l);
+/* 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 '?'.
 
-  current_locale = setlocale (LC_CTYPE, 0);
-  current_charset = locale_charset ();
-  setlocale (LC_CTYPE, locale);
+   The returned string will be null-terminated and allocated on POOL.
 
-  charset = locale_charset ();
-  setlocale (LC_CTYPE, current_locale);
+   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)
+{
+  size_t outbufferlength;
+  iconv_t conv ;
+
+  if (to == NULL)
+    to = default_encoding;
 
-  iconv_close (convertor[CONV_PSPP_TO_UTF8]);
-  convertor[CONV_PSPP_TO_UTF8] = create_iconv ("UTF-8", charset);
+  if (from == NULL)
+    from = default_encoding;
 
-  iconv_close (convertor[CONV_SYSTEM_TO_PSPP]);
-  convertor[CONV_SYSTEM_TO_PSPP] = create_iconv (charset, current_charset);
+  conv = create_iconv (to, from);
 
-  iconv_close (convertor[CONV_UTF8_TO_PSPP]);
-  convertor[CONV_UTF8_TO_PSPP] = create_iconv (charset, "UTF-8");
+  if ( (iconv_t) -1 == conv )
+    {
+      struct substring out;
+      ss_alloc_substring_pool (&out, text, pool);
+      return out;
+    }
+
+  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);
+      }
+
+  NOT_REACHED ();
 }
 
 void
 i18n_init (void)
 {
-  assert (!locale) ;
-  locale = strdup (setlocale (LC_CTYPE, NULL));
+  setlocale (LC_CTYPE, "");
+  setlocale (LC_MESSAGES, "");
+#if HAVE_LC_PAPER
+  setlocale (LC_PAPER, "");
+#endif
+  bindtextdomain (PACKAGE, relocate(locale_dir));
+  textdomain (PACKAGE);
+
+  assert (default_encoding == NULL);
+  default_encoding = xstrdup (locale_charset ());
 
-  setlocale (LC_CTYPE, locale);
-  charset = locale_charset ();
+  hmapx_init (&map);
+}
 
-  convertor[CONV_PSPP_TO_UTF8]   = create_iconv ("UTF-8", charset);
-  convertor[CONV_SYSTEM_TO_PSPP] = create_iconv (charset, charset);
-  convertor[CONV_UTF8_TO_PSPP]   = create_iconv (charset, "UTF-8");
+const char *
+get_default_encoding (void)
+{
+  return default_encoding;
 }
 
+void
+set_default_encoding (const char *enc)
+{
+  free (default_encoding);
+  default_encoding = xstrdup (enc);
+}
+
+
+/* Attempts to set the encoding from a locale name
+   returns true if successfull.
+   This function does not (should not!) alter the current locale.
+*/
+bool
+set_encoding_from_locale (const char *loc)
+{
+  bool ok = true;
+  char *c_encoding;
+  char *loc_encoding;
+  char *tmp = xstrdup (setlocale (LC_CTYPE, NULL));
+
+  setlocale (LC_CTYPE, "C");
+  c_encoding = xstrdup (locale_charset ());
+
+  setlocale (LC_CTYPE, loc);
+  loc_encoding = xstrdup (locale_charset ());
+
+
+  if ( 0 == strcmp (loc_encoding, c_encoding))
+    {
+      ok = false;
+    }
+
+
+  setlocale (LC_CTYPE, tmp);
+
+  free (tmp);
+
+  if (ok)
+    {
+      free (default_encoding);
+      default_encoding = loc_encoding;
+    }
+  else
+    free (loc_encoding);
+
+  free (c_encoding);
+
+  return ok;
+}
 
 void
 i18n_done (void)
 {
-  int i;
-  free (locale);
-  locale = 0;
+  struct hmapx_node *node;
+  struct converter *cvtr;
 
-  for(i = 0 ; i < n_CONV; ++i )
+  HMAPX_FOR_EACH (cvtr, node, &map)
     {
-      if ( (iconv_t) -1 == convertor[i] )
-	continue;
-      iconv_close (convertor[i]);
+      free (cvtr->tocode);
+      free (cvtr->fromcode);
+      if (cvtr->conv != (iconv_t) -1)
+        iconv_close (cvtr->conv);
+      free (cvtr);
     }
+
+  hmapx_destroy (&map);
+
+  free (default_encoding);
+  default_encoding = NULL;
 }
 
 
 
+bool
+valid_encoding (const char *enc)
+{
+  iconv_t conv = iconv_open (UTF8, enc);
+
+  if ( conv == (iconv_t) -1)
+    return false;
+
+  iconv_close (conv);
+
+  return true;
+}
+
 
 /* Return the system local's idea of the
    decimal seperator character */
@@ -227,7 +643,7 @@ get_system_decimal (void)
 {
   char radix_char;
 
-  char *ol = setlocale (LC_NUMERIC, NULL);
+  char *ol = xstrdup (setlocale (LC_NUMERIC, NULL));
   setlocale (LC_NUMERIC, "");
 
 #if HAVE_NL_LANGINFO
@@ -243,6 +659,65 @@ get_system_decimal (void)
   /* We MUST leave LC_NUMERIC untouched, since it would
      otherwise interfere with data_{in,out} */
   setlocale (LC_NUMERIC, ol);
+  free (ol);
   return radix_char;
 }
 
+const char *
+uc_name (ucs4_t uc, char buffer[16])
+{
+  if (uc >= 0x20 && uc < 0x7f)
+    snprintf (buffer, 16, "`%c'", uc);
+  else
+    snprintf (buffer, 16, "U+%04X", uc);
+  return buffer;
+}
+
+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;
+  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);
+  ok = cr.length >= 1 && cr.length <= MAX_UNIT && cr.length == lf.length;
+  if (!ok)
+    {
+      fprintf (stderr, "warning: encoding `%s' is not supported.\n", name);
+      ss_dealloc (&cr);
+      ss_dealloc (&lf);
+      ss_alloc_substring (&cr, ss_cstr ("\r"));
+      ss_alloc_substring (&lf, ss_cstr ("\n"));
+    }
+
+  e->unit = cr.length;
+  memcpy (e->cr, cr.string, e->unit);
+  memcpy (e->lf, lf.string, e->unit);
+
+  ss_dealloc (&cr);
+  ss_dealloc (&lf);
+
+  out = recode_substring_pool ("UTF-8", name, in, NULL);
+  e->is_ascii_compatible = ss_equals (in, out);
+  ss_dealloc (&out);
+
+  return ok;
+}
+
+bool
+is_encoding_ascii_compatible (const char *encoding)
+{
+  struct encoding_info e;
+
+  get_encoding_info (&e, encoding);
+  return e.is_ascii_compatible;
+}