Fixes bug #320666:

2005-11-07  Federico Mena Quintero  <federico@ximian.com>

	Fixes bug #320666:

	Instead of doing a bsearch() for every gunichar to map it to a
	PangoScript, use a precomputed table for the first 8192 code
	points.  Also, remember the last script that we computed on each
	invocation; this will also help CJK and the other scripts above
	U+2000.

	This table also holds information on whether the characters in it
	are paired characters.  We can use this to avoid doing the
	expensive get_pair_index() call most of the time.

	Many thanks to Matthias Clasen for his suggestions for this patch.

	* tools/gen-easy-scripts-table.c: New program to generate
	pango_easy_scripts_table.

	* tools/Makefile.am: Build gen-easy-scripts-table.

	* pango/pango-easy-scripts-table.h: New file with a mapping of the
	first 8192 Unicode characters to their corresponding scripts.  The
	table also says whether each character has a paired char or not.

	* pango/Makefile.am (libpango_1_0_la_SOURCES): Add pango-easy-scripts-table.h.

	* pango/pango-script-table.h: Remove everything below U+2000, and
	add a note to that effect.

	* pango/pango-script.c (pango_script_for_unichar_with_last_index):
	New function.  This is the old pango_script_for_unichar(), but it
	lets the caller keep around the computed index in
	pango_script_table.  This works under the assumption that a
	character is likely to be in the same script block as the
	preceding character in a string.
	(pango_script_for_unichar): First, do a quick check against the
	pango_easy_scripts_table.  Then, do the expensive check with
	pango_script_for_unichar_with_last_index().
	(pango_script_iter_next): If the character is within the easy
	script range, find out if it is a paired character by using
	PANGO_PAIRED_CHAR_FLAG.
	(struct _PangoScriptIter): Add a last_index_for_script_lookup
	field.  We use this to maintain the last-lookup index from
	pango_script_for_unichar_with_last_index().
	(pango_script_iter_next): If the character is not within the easy
	script range, use pango_script_for_unichar_with_last_index(), and
	store the index in the last_index_for_script_lookup field of the
	PangoScriptIter.

1 files changed, 32 insertions, 31 deletions

diff --git a/pango/Makefile.am b/pango/Makefile.am
index da89b91e..7800eb76 100644
--- a/pango/Makefile.am
+++ b/pango/Makefile.am
@@ -57,37 +57,38 @@ endif
 pango-win32-res.o: pango.rc
 	$(WINDRES) $< $@
 
-libpango_1_0_la_SOURCES =      \
-	break.c 	       \
-	ellipsize.c     	\
-	fonts.c 	       \
-	glyphstring.c 	       \
-	mapping.c 	       \
-	modules.c 	       \
-	pango-attributes.c     \
-	pango-color.c	       \
-	pango-color-table.h    \
-	pango-context.c        \
-	pango-coverage.c       \
-	pango-engine.c	       \
-	pango-engine-private.h \
-	pango-fontmap.c	       \
-	pango-fontset.c	       \
-	pango-glyph-item.c     \
-	pango-glyph-item-private.h     \
-	pango-impl-utils.h     \
-	pango-item.c	       \
-	pango-layout.c	       \
-	pango-layout-private.h	       \
-	pango-markup.c	       \
-	pango-renderer.c       \
-	pango-script.c	       \
-	pango-script-lang-table.h \
-	pango-script-table.h   \
-	pango-tabs.c	       \
-	pango-utils.c	       \
-	reorder-items.c        \
-	shape.c	               \
+libpango_1_0_la_SOURCES =			\
+	break.c					\
+	ellipsize.c				\
+	fonts.c					\
+	glyphstring.c				\
+	mapping.c				\
+	modules.c				\
+	pango-attributes.c			\
+	pango-color.c				\
+	pango-color-table.h			\
+	pango-context.c				\
+	pango-coverage.c			\
+	pango-easy-scripts-table.h		\
+	pango-engine.c				\
+	pango-engine-private.h			\
+	pango-fontmap.c				\
+	pango-fontset.c				\
+	pango-glyph-item.c			\
+	pango-glyph-item-private.h		\
+	pango-impl-utils.h			\
+	pango-item.c				\
+	pango-layout.c				\
+	pango-layout-private.h			\
+	pango-markup.c				\
+	pango-renderer.c			\
+	pango-script.c				\
+	pango-script-lang-table.h		\
+	pango-script-table.h			\
+	pango-tabs.c				\
+	pango-utils.c				\
+	reorder-items.c				\
+	shape.c					\
 	pango-enum-types.c
 
 pango_headers =		   \