--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,11 @@
 2009-02-21  Bruno Haible  <bruno@clisp.org>
 
+	New module 'uninorm/u8-normalize'.
+	* lib/uninorm/u8-normalize.c: New file.
+	* lib/uninorm/normalize-internal.h: New file.
+	* lib/uninorm/u-normalize-internal.h: New file.
+	* modules/uninorm/u8-normalize: New file.
+
 	New module 'uninorm/decompose-internal'.
 	* lib/uninorm/decompose-internal.c: New file.
 	* modules/uninorm/decompose-internal: New file.

new file mode 100644
--- /dev/null
+++ b/lib/uninorm/normalize-internal.h
@@ -0,0 +1,34 @@
+/* Normalization of Unicode strings.
+   Copyright (C) 2009 Free Software Foundation, Inc.
+   Written by Bruno Haible <bruno@clisp.org>, 2009.
+
+   This program is free software: you can redistribute it and/or modify it
+   under the terms of the GNU Lesser General Public License as published
+   by the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+#include <stddef.h>
+
+#include "unitypes.h"
+
+/* Complete definition of normalization form descriptor.  */
+struct unicode_normalization_form
+{
+  /* Bit mask containing meta-information.  */
+  unsigned int description;
+  #define NF_IS_COMPAT_DECOMPOSING  (1 << 0)
+  #define NF_IS_COMPOSING           (1 << 1)
+  /* Function that decomposes a Unicode character.  */
+  int (*decomposer) (ucs4_t uc, ucs4_t *decomposition);
+  /* Function that combines two Unicode characters, a starter and another
+     character.  */
+  ucs4_t (*composer) (ucs4_t uc1, ucs4_t uc2);
+};

new file mode 100644
--- /dev/null
+++ b/lib/uninorm/u-normalize-internal.h
@@ -0,0 +1,352 @@
+/* Decomposition and composition of Unicode strings.
+   Copyright (C) 2009 Free Software Foundation, Inc.
+   Written by Bruno Haible <bruno@clisp.org>, 2009.
+
+   This program is free software: you can redistribute it and/or modify it
+   under the terms of the GNU Lesser General Public License as published
+   by the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+UNIT *
+FUNC (uninorm_t nf, const UNIT *s, size_t n,
+      UNIT *resultbuf, size_t *lengthp)
+{
+  int (*decomposer) (ucs4_t uc, ucs4_t *decomposition) = nf->decomposer;
+  ucs4_t (*composer) (ucs4_t uc1, ucs4_t uc2) = nf->composer;
+
+  /* The result being accumulated.  */
+  UNIT *result;
+  size_t length;
+  size_t allocated;
+  /* The buffer for sorting.  */
+  #define SORTBUF_PREALLOCATED 64
+  struct ucs4_with_ccc sortbuf_preallocated[2 * SORTBUF_PREALLOCATED];
+  struct ucs4_with_ccc *sortbuf; /* array of size 2 * sortbuf_allocated */
+  size_t sortbuf_allocated;
+  size_t sortbuf_count;
+
+  /* Initialize the accumulator.  */
+  if (resultbuf == NULL)
+    {
+      result = NULL;
+      allocated = 0;
+    }
+  else
+    {
+      result = resultbuf;
+      allocated = *lengthp;
+    }
+  length = 0;
+
+  /* Initialize the buffer for sorting.  */
+  sortbuf = sortbuf_preallocated;
+  sortbuf_allocated = SORTBUF_PREALLOCATED;
+  sortbuf_count = 0;
+
+  {
+    const UNIT *s_end = s + n;
+
+    for (;;)
+      {
+	int count;
+	ucs4_t decomposed[UC_DECOMPOSITION_MAX_LENGTH];
+	int decomposed_count;
+	int i;
+
+	if (s < s_end)
+	  {
+	    /* Fetch the next character.  */
+	    count = U_MBTOUC_UNSAFE (&decomposed[0], s, s_end - s);
+	    decomposed_count = 1;
+
+	    /* Decompose it, recursively.
+	       It would be possible to precompute the recursive decomposition
+	       and store it in a table.  But this would significantly increase
+	       the size of the decomposition tables, because for example for
+	       U+1FC1 the recursive canonical decomposition and the recursive
+	       compatibility decomposition are different.  */
+	    {
+	      int curr;
+
+	      for (curr = 0; curr < decomposed_count; )
+		{
+		  /* Invariant: decomposed[0..curr-1] is fully decomposed, i.e.
+		     all elements are atomic.  */
+		  ucs4_t curr_decomposed[UC_DECOMPOSITION_MAX_LENGTH];
+		  int curr_decomposed_count;
+
+		  curr_decomposed_count = decomposer (decomposed[curr], curr_decomposed);
+		  if (curr_decomposed_count >= 0)
+		    {
+		      /* Move curr_decomposed[0..curr_decomposed_count-1] over
+			 decomposed[curr], making room.  It's not worth using
+			 memcpy() here, since the counts are so small.  */
+		      int shift = curr_decomposed_count - 1;
+
+		      if (shift < 0)
+			abort ();
+		      if (shift > 0)
+			{
+			  int j;
+
+			  decomposed_count += shift;
+			  if (decomposed_count > UC_DECOMPOSITION_MAX_LENGTH)
+			    abort ();
+			  for (j = decomposed_count - 1 - shift; j > curr; j--)
+			    decomposed[j + shift] = decomposed[j];
+			}
+		      for (; shift >= 0; shift--)
+			decomposed[curr + shift] = curr_decomposed[shift];
+		    }
+		  else
+		    {
+		      /* decomposed[curr] is atomic.  */
+		      curr++;
+		    }
+		}
+	    }
+	  }
+	else
+	  {
+	    count = 0;
+	    decomposed_count = 0;
+	  }
+
+	i = 0;
+	for (;;)
+	  {
+	    ucs4_t uc;
+	    int ccc;
+
+	    if (s < s_end)
+	      {
+		/* Fetch the next character from the decomposition.  */
+		if (i == decomposed_count)
+		  break;
+		uc = decomposed[i];
+		ccc = uc_combining_class (uc);
+	      }
+	    else
+	      {
+		/* End of string reached.  */
+		uc = 0;
+		ccc = 0;
+	      }
+
+	    if (ccc == 0)
+	      {
+	        size_t j;
+
+		/* Apply the canonical ordering algorithm to the accumulated
+		   sequence of characters.  */
+		if (sortbuf_count > 1)
+		  gl_uninorm_decompose_merge_sort_inplace (sortbuf, sortbuf_count,
+							   sortbuf + sortbuf_count);
+
+		if (composer != NULL)
+		  {
+		    /* Attempt to combine decomposed characters, as specified
+		       in the Unicode Standard Annex #15 "Unicode Normalization
+		       Forms".  We need to check
+			 1. whether the first accumulated character is a
+			    "starter" (i.e. has ccc = 0).  This is usually the
+			    case.  But when the string starts with a
+			    non-starter, the sortbuf also starts with a
+			    non-starter.  Btw, this check could also be
+			    omitted, because the composition table has only
+			    entries (code1, code2) for which code1 is a
+			    starter; if the first accumulated character is not
+			    a starter, no lookup will succeed.
+			 2. If the sortbuf has more than one character, check
+			    for each of these characters that are not "blocked"
+			    from the starter (i.e. have a ccc that is higher
+			    than the ccc of the previous character) whether it
+			    can be combined with the first character.
+			 3. If only one character is left in sortbuf, check
+			    whether it can be combined with the next character
+			    (also a starter).  */
+		    if (sortbuf_count > 0 && sortbuf[0].ccc == 0)
+		      {
+			for (j = 1; j < sortbuf_count; )
+			  {
+			    if (sortbuf[j].ccc > sortbuf[j - 1].ccc)
+			      {
+				ucs4_t combined =
+				  composer (sortbuf[0].code, sortbuf[j].code);
+				if (combined)
+				  {
+				    size_t k;
+
+				    sortbuf[0].code = combined;
+				    /* sortbuf[0].ccc = 0, still valid.  */
+				    for (k = j + 1; k < sortbuf_count; k++)
+				      sortbuf[k - 1] = sortbuf[k];
+				    sortbuf_count--;
+				    continue;
+				  }
+			      }
+			    j++;
+			  }
+			if (s < s_end && sortbuf_count == 1)
+			  {
+			    ucs4_t combined =
+			      composer (sortbuf[0].code, uc);
+			    if (combined)
+			      {
+				uc = combined;
+				ccc = 0;
+				/* uc could be further combined with subsequent
+				   characters.  So don't put it into sortbuf[0] in
+				   this round, only in the next round.  */
+				sortbuf_count = 0;
+			      }
+			  }
+		      }
+		  }
+
+		for (j = 0; j < sortbuf_count; j++)
+		  {
+		    ucs4_t muc = sortbuf[j].code;
+
+		    /* Append muc to the result accumulator.  */
+		    if (length < allocated)
+		      {
+			int ret =
+			  U_UCTOMB (result + length, muc, allocated - length);
+			if (ret == -1)
+			  {
+			    errno = EINVAL;
+			    goto fail;
+			  }
+			if (ret >= 0)
+			  {
+			    length += ret;
+			    goto done_appending;
+			  }
+		      }
+		    {
+		      size_t old_allocated = allocated;
+		      size_t new_allocated = 2 * old_allocated;
+		      if (new_allocated < 64)
+			new_allocated = 64;
+		      if (new_allocated < old_allocated) /* integer overflow? */
+			abort ();
+		      {
+			UNIT *larger_result;
+			if (result == NULL)
+			  {
+			    larger_result =
+			      (UNIT *) malloc (new_allocated * sizeof (UNIT));
+			    if (larger_result == NULL)
+			      {
+				errno = ENOMEM;
+				goto fail;
+			      }
+			  }
+			else if (result == resultbuf)
+			  {
+			    larger_result =
+			      (UNIT *) malloc (new_allocated * sizeof (UNIT));
+			    if (larger_result == NULL)
+			      {
+				errno = ENOMEM;
+				goto fail;
+			      }
+			    U_CPY (larger_result, resultbuf, length);
+			  }
+			else
+			  {
+			    larger_result =
+			      (UNIT *) realloc (result, new_allocated * sizeof (UNIT));
+			    if (larger_result == NULL)
+			      {
+				errno = ENOMEM;
+				goto fail;
+			      }
+			  }
+			result = larger_result;
+			allocated = new_allocated;
+			{
+			  int ret =
+			    U_UCTOMB (result + length, muc, allocated - length);
+			  if (ret == -1)
+			    {
+			      errno = EINVAL;
+			      goto fail;
+			    }
+			  if (ret < 0)
+			    abort ();
+			  length += ret;
+			  goto done_appending;
+			}
+		      }
+		    }
+		   done_appending: ;
+		  }
+
+		/* sortbuf is now empty.  */
+		sortbuf_count = 0;
+	      }
+
+	    if (!(s < s_end))
+	      /* End of string reached.  */
+	      break;
+
+	    /* Append (uc, ccc) to sortbuf.  */
+	    if (sortbuf_count == sortbuf_allocated)
+	      {
+		struct ucs4_with_ccc *new_sortbuf;
+
+		sortbuf_allocated = 2 * sortbuf_allocated;
+		if (sortbuf_allocated < sortbuf_count) /* integer overflow? */
+		  abort ();
+		new_sortbuf =
+		  (struct ucs4_with_ccc *) malloc (2 * sortbuf_allocated * sizeof (struct ucs4_with_ccc));
+		memcpy (new_sortbuf, sortbuf,
+			sortbuf_count * sizeof (struct ucs4_with_ccc));
+		if (sortbuf != sortbuf_preallocated)
+		  free (sortbuf);
+		sortbuf = new_sortbuf;
+	      }
+	    sortbuf[sortbuf_count].code = uc;
+	    sortbuf[sortbuf_count].ccc = ccc;
+	    sortbuf_count++;
+
+	    i++;
+	  }
+
+	if (!(s < s_end))
+	  /* End of string reached.  */
+	  break;
+
+	s += count;
+      }
+  }
+
+  if (sortbuf_count > 0)
+    abort ();
+  if (sortbuf != sortbuf_preallocated)
+    free (sortbuf);
+
+  *lengthp = length;
+  return result;
+
+ fail:
+  {
+    int saved_errno = errno;
+    if (sortbuf != sortbuf_preallocated)
+      free (sortbuf);
+    if (result != resultbuf)
+      free (result);
+    errno = saved_errno;
+  }
+  return NULL;
+}

new file mode 100644
--- /dev/null
+++ b/lib/uninorm/u8-normalize.c
@@ -0,0 +1,38 @@
+/* Normalization of UTF-8 strings.
+   Copyright (C) 2009 Free Software Foundation, Inc.
+   Written by Bruno Haible <bruno@clisp.org>, 2009.
+
+   This program is free software: you can redistribute it and/or modify it
+   under the terms of the GNU Lesser General Public License as published
+   by the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
+
+#include <config.h>
+
+/* Specification.  */
+#include "uninorm.h"
+
+#include <errno.h>
+#include <stddef.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "unistr.h"
+#include "unictype.h"
+#include "normalize-internal.h"
+#include "decompose-internal.h"
+
+#define FUNC u8_normalize
+#define UNIT uint8_t
+#define U_MBTOUC_UNSAFE u8_mbtouc_unsafe
+#define U_UCTOMB u8_uctomb
+#define U_CPY u8_cpy
+#include "u-normalize-internal.h"

new file mode 100644
--- /dev/null
+++ b/modules/uninorm/u8-normalize
@@ -0,0 +1,31 @@
+Description:
+Normalization of UTF-8 strings.
+
+Files:
+lib/uninorm/u8-normalize.c
+lib/uninorm/normalize-internal.h
+lib/uninorm/u-normalize-internal.h
+
+Depends-on:
+uninorm/base
+unistr/u8-mbtouc-unsafe
+unistr/u8-uctomb
+unistr/u8-cpy
+unictype/combining-class
+uninorm/decompose-internal
+
+configure.ac:
+gl_MODULE_INDICATOR([uninorm/u8-normalize])
+
+Makefile.am:
+lib_SOURCES += uninorm/u8-normalize.c
+
+Include:
+"uninorm.h"
+
+License:
+LGPL
+
+Maintainer:
+Bruno Haible
+