+ 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;
+}
+\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 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,
+ 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;