Use ICU/boost.locale for case-insensitive searching

Do proper unicode case-folding for case-insensitive searching rather
than converting only ascii characters to lowercase. The Turkish 'i' is
still not handled correctly (since it's the only place where
case-folding is locale-dependent), but that's probably not worth caring
about as long as we don't have a Turkish UI translation.

This affects both the find/replace dialog and the select lines dialog.

Closes #1342.

aegisub/Makefile.inc.in

@@ -76,7 +76,7 @@ CPPFLAGS_WX        = @WX_CPPFLAGS@
 LIBS_WX            = @WX_LIBS@ -lz
 
 CPPFLAGS_BOOST     = @BOOST_CPPFLAGS@
-LIBS_BOOST         = @BOOST_FILESYSTEM_LDFLAGS@ @BOOST_FILESYSTEM_LIBS@ @BOOST_REGEX_LDFLAGS@ @BOOST_REGEX_LIBS@ @BOOST_SYSTEM_LDFLAGS@ @BOOST_SYSTEM_LIBS@
+LIBS_BOOST         = @BOOST_FILESYSTEM_LDFLAGS@ @BOOST_FILESYSTEM_LIBS@ @BOOST_LOCALE_LIBS@ @BOOST_REGEX_LIBS@ @BOOST_SYSTEM_LIBS@
 
 CFLAGS_FFMS2       = @FFMS2_CFLAGS@
 CFLAGS_FFTW3       = @FFTW3_CFLAGS@

aegisub/build/tests/tests.vcxproj

@@ -44,6 +44,7 @@
     <ClCompile Include="$(SrcDir)tests\dialogue_lexer.cpp" />
     <ClCompile Include="$(SrcDir)tests\hotkey.cpp" />
     <ClCompile Include="$(SrcDir)tests\iconv.cpp" />
+    <ClCompile Include="$(SrcDir)tests\ifind.cpp" />
     <ClCompile Include="$(SrcDir)tests\keyframe.cpp" />
     <ClCompile Include="$(SrcDir)tests\line_iterator.cpp" />
     <ClCompile Include="$(SrcDir)tests\line_wrap.cpp" />

aegisub/build/tests/tests.vcxproj.filters

@@ -35,6 +35,9 @@
     <ClCompile Include="$(SrcDir)tests\iconv.cpp">
       <Filter>Tests</Filter>
     </ClCompile>
+    <ClCompile Include="$(SrcDir)tests\ifind.cpp">
+      <Filter>Tests</Filter>
+    </ClCompile>
     <ClCompile Include="$(SrcDir)tests\keyframe.cpp">
       <Filter>Tests</Filter>
     </ClCompile>

aegisub/configure.ac

@@ -248,6 +248,7 @@ PKG_CHECK_MODULES(FONTCONFIG, fontconfig >= fontconfig_required_version,
 libext=a
 BOOST_REQUIRE([boost_required_version])
 BOOST_FILESYSTEM
+BOOST_LOCALE
 BOOST_REGEX
 
 ########

aegisub/libaegisub/common/util.cpp

@@ -18,8 +18,45 @@
 
 #include "libaegisub/util.h"
 
+#include "libaegisub/exception.h"
+
+#include <boost/locale/boundary.hpp>
+#include <boost/locale/conversion.hpp>
+#include <boost/range/algorithm_ext.hpp>
 #include <ctime>
 
+namespace {
+const size_t bad_pos = (size_t)-1;
+const std::pair<size_t, size_t> bad_match(bad_pos, bad_pos);
+
+template<typename Iterator>
+size_t advance_both(Iterator& folded, Iterator& raw) {
+	size_t len;
+	if (*folded == *raw) {
+		len = folded->length();
+		++folded;
+	}
+	else {
+		// This character was changed by case folding, so refold it and eat the
+		// appropriate number of characters from folded
+		len = boost::locale::fold_case(raw->str()).size();
+		for (size_t folded_consumed = 0; folded_consumed < len; ++folded)
+			folded_consumed += folded->length();
+	}
+
+	++raw;
+	return len;
+}
+
+std::pair<size_t, size_t> find_range(std::string const& haystack, std::string const& needle, size_t start = 0) {
+	const size_t match_start = haystack.find(needle, start);
+	if (match_start == std::string::npos)
+		return bad_match;
+	return std::make_pair(match_start, match_start + needle.size());
+}
+
+}
+
 namespace agi { namespace util {
 
 std::string strftime(const char *fmt, const tm *tmptr) {
@@ -33,4 +70,66 @@ std::string strftime(const char *fmt, const tm *tmptr) {
 	return buff;
 }
 
+std::pair<size_t, size_t> ifind(std::string const& haystack, std::string const& needle) {
+	const auto folded_hs = boost::locale::fold_case(haystack);
+	const auto folded_n = boost::locale::fold_case(needle);
+	auto match = find_range(folded_hs, folded_n);
+	if (match == bad_match || folded_hs == haystack)
+		return match;
+
+	// We have a match, but the position is an index into the folded string
+	// and we want an index into the unfolded string.
+
+	using namespace boost::locale::boundary;
+	const ssegment_index haystack_characters(character, begin(haystack), end(haystack));
+	const ssegment_index folded_characters(character, begin(folded_hs), end(folded_hs));
+	const size_t haystack_char_count = boost::distance(haystack_characters);
+	const size_t folded_char_count = boost::distance(folded_characters);
+
+	// As of Unicode 6.2, case folding can never reduce the number of
+	// characters, and can only reduce the number of bytes with UTF-8 when
+	// increasing the number of characters. As a result, iff the bytes and
+	// characters are unchanged, no folds changed the size of any characters
+	// and our indices are correct.
+	if (haystack.size() == folded_hs.size() && haystack_char_count == folded_char_count)
+		return match;
+
+	const auto map_folded_to_raw = [&]() -> std::pair<size_t, size_t> {
+		size_t start = -1;
+
+		// Iterate over each pair of characters and refold each character which was
+		// changed by folding, so that we can find the corresponding positions in
+		// the unfolded string
+		auto folded_it = begin(folded_characters);
+		auto haystack_it = begin(haystack_characters);
+		size_t folded_pos = 0;
+
+		while (folded_pos < match.first)
+			folded_pos += advance_both(folded_it, haystack_it);
+		// If we overshot the start then the match started in the middle of a
+		// character which was folded to multiple characters
+		if (folded_pos > match.first)
+			return bad_match;
+
+		start = distance(begin(haystack), begin(*haystack_it));
+
+		while (folded_pos < match.second)
+			folded_pos += advance_both(folded_it, haystack_it);
+		if (folded_pos > match.second)
+			return bad_match;
+
+		return std::make_pair(start, distance(begin(haystack), begin(*haystack_it)));
+	};
+
+	auto ret = map_folded_to_raw();
+	while (ret == bad_match) {
+		// Found something, but it was an invalid match so retry from the next character
+		match = find_range(folded_hs, folded_n, match.first + 1);
+		if (match == bad_match) return match;
+		ret = map_folded_to_raw();
+	}
+
+	return ret;
+}
+
 } }

aegisub/libaegisub/include/libaegisub/util.h

@@ -41,6 +41,16 @@ namespace agi {
 	/// @return The strftime-formatted string
 	std::string strftime(const char *fmt, const tm *tmptr = nullptr);
 
+	/// Case-insensitive find with proper case folding
+	/// @param haystack String to search
+	/// @param needle String to look for
+	/// @return make_pair(-1,-1) if `needle` could not be found, or a range equivalent to `needle` in `haystack` if it could
+	///
+	/// `needle` and `haystack` must both be in Normalization Form D. The size
+	/// of the match might be different from the size of `needle`, since it's
+	/// based on the unfolded length.
+	std::pair<size_t, size_t> ifind(std::string const& haystack, std::string const& needle);
+
 	struct delete_ptr {
 		template<class T>
 		void operator()(T* ptr) const {

aegisub/src/dialog_search_replace.cpp

@@ -36,6 +36,7 @@
 #include <wx/checkbox.h>
 #include <wx/combobox.h>
 #include <wx/radiobox.h>
+#include <wx/msgdlg.h>
 #include <wx/sizer.h>
 #include <wx/stattext.h>
 #include <wx/textctrl.h>

aegisub/src/search_replace_engine.cpp

@@ -25,8 +25,9 @@
 #include "text_selection_controller.h"
 
 #include <libaegisub/of_type_adaptor.h>
+#include <libaegisub/util.h>
 
-#include <boost/algorithm/string/case_conv.hpp>
+#include <boost/locale.hpp>
 
 #include <wx/msgdlg.h>
 
@@ -43,6 +44,14 @@ auto get_dialogue_field(SearchReplaceSettings::Field field) -> decltype(&AssDial
 	throw agi::InternalError("Bad field for search", 0);
 }
 
+std::string const& get_normalized(const AssDialogue *diag, decltype(&AssDialogue::Text) field) {
+	auto& value = const_cast<AssDialogue*>(diag)->*field;
+	auto normalized = boost::locale::normalize(value.get());
+	if (normalized != value)
+		value = normalized;
+	return value.get();
+}
+
 typedef std::function<MatchState (const AssDialogue*, size_t)> matcher;
 
 class noop_accessor {
@@ -54,7 +63,7 @@ public:
 
 	std::string get(const AssDialogue *d, size_t s) {
 		start = s;
-		return (d->*field).get().substr(s);
+		return get_normalized(d, field).substr(s);
 	}
 
 	MatchState make_match_state(size_t s, size_t e, boost::u32regex *r = nullptr) {
@@ -87,7 +96,7 @@ public:
 	skip_tags_accessor(SearchReplaceSettings::Field f) : field(get_dialogue_field(f)), start(0) { }
 
 	std::string get(const AssDialogue *d, size_t s) {
-		auto const& str = (d->*field).get();
+		auto const& str = get_normalized(d, field);
 		parse_str(str);
 
 		std::string out;
@@ -156,25 +165,25 @@ matcher get_matcher(SearchReplaceSettings const& settings, Accessor&& a) {
 		};
 	}
 
-	bool full_match_only = settings.exact_match;
+	const bool full_match_only = settings.exact_match;
-	bool match_case = settings.match_case;
+	const bool match_case = settings.match_case;
 	std::string look_for = settings.find;
+
 	if (!settings.match_case)
-		boost::to_lower(look_for);
+		look_for = boost::locale::fold_case(look_for);
 
 	return [=](const AssDialogue *diag, size_t start) mutable -> MatchState {
-		auto str = a.get(diag, start);
+		const auto str = a.get(diag, start);
 		if (full_match_only && str.size() != look_for.size())
 			return MatchState();
 
-		if (!match_case)
+		if (match_case) {
-			boost::to_lower(str);
+			const auto pos = str.find(look_for);
+			return pos == std::string::npos ? MatchState() : a.make_match_state(pos, pos + look_for.size());
+		}
 
-		size_t pos = str.find(look_for);
+		const auto pos = agi::util::ifind(str, look_for);
-		if (pos == std::string::npos)
+		return pos.first == bad_pos ? MatchState() : a.make_match_state(pos.first, pos.second);
-			return MatchState();
-
-		return a.make_match_state(pos, pos + look_for.size());
 	};
 }
 

aegisub/tests/Makefile

@@ -23,6 +23,7 @@ SRC = \
 		tests/dialogue_lexer.cpp \
 		tests/hotkey.cpp \
 		tests/iconv.cpp \
+		tests/ifind.cpp \
 		tests/keyframe.cpp \
 		tests/line_iterator.cpp \
 		tests/line_wrap.cpp \

aegisub/tests/support/main.cpp

@@ -18,8 +18,11 @@
 #include <libaegisub/fs.h>
 #include <libaegisub/log.h>
 
+#include <boost/locale/generator.hpp>
+
 int main(int argc, char **argv) {
 	agi::dispatch::Init([](agi::dispatch::Thunk f) { });
+	std::locale::global(boost::locale::generator().generate(""));
 
 	int retval;
 	agi::log::log = new agi::log::LogSink;

aegisub/tests/tests/ifind.cpp

@@ -0,0 +1,75 @@
+// Copyright (c) 2013, Thomas Goyne <plorkyeran@aegisub.org>
+//
+// Permission to use, copy, modify, and distribute this software for any
+// purpose with or without fee is hereby granted, provided that the above
+// copyright notice and this permission notice appear in all copies.
+//
+// THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+// WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+// MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+// ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+// WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+// ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+// OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+//
+// Aegisub Project http://www.aegisub.org/
+
+#include <libaegisub/util.h>
+
+#include "main.h"
+
+#define IFIND(haystack, needle) \
+	std::pair<size_t, size_t> pos; \
+	ASSERT_NO_THROW(pos = agi::util::ifind(haystack, needle))
+
+#define EXPECT_IFIND(haystack, needle, s, e) \
+	do { \
+		IFIND(haystack, needle); \
+		EXPECT_EQ((size_t)s, pos.first); \
+		EXPECT_EQ((size_t)e, pos.second); \
+	} while(false)
+
+#define EXPECT_NO_MATCH(haystack, needle) \
+	do { \
+		IFIND(haystack, needle); \
+		EXPECT_EQ((size_t)-1, pos.first); \
+		EXPECT_EQ((size_t)-1, pos.second); \
+	} while(false)
+
+TEST(lagi_ifind, basic_match) {
+	EXPECT_IFIND(" a ", "a", 1, 2);
+	EXPECT_IFIND(" a ", "A", 1, 2);
+	EXPECT_NO_MATCH(" a ", "b");
+}
+
+TEST(lagi_ifind, sharp_s_matches_ss) {
+	// lowercase
+	EXPECT_IFIND(" \xC3\x9F ", "ss", 1, 3);
+	EXPECT_IFIND(" ss ", "\xC3\x9F", 1, 3);
+
+	// uppercase
+	EXPECT_IFIND(" \xE1\xBA\x9E ", "ss", 1, 4);
+	EXPECT_IFIND(" ss ", "\xE1\xBA\x9E", 1, 3);
+}
+
+TEST(lagi_ifind, no_partial_match_on_decomposed_character) {
+	EXPECT_NO_MATCH("s\xEF\xAC\x86", "ss"); // LATIN SMALL LIGATURE ST
+	EXPECT_NO_MATCH("\xEF\xAC\x86t", "tt");
+	EXPECT_NO_MATCH(" \xE1\xBA\x9E ", "s");
+	EXPECT_NO_MATCH("\xE1\xBA\x9E", "s");
+	EXPECT_IFIND(" \xE1\xBA\x9E s ", "s", 5, 6);
+	EXPECT_IFIND("s\xE1\xBA\x9E", "ss", 1, 4);
+	EXPECT_IFIND("s\xE1\xBA\x9E", "sss", 0, 4);
+	EXPECT_IFIND("\xE1\xBA\x9Es", "sss", 0, 4);
+	EXPECT_IFIND("\xEF\xAC\x86", "st", 0, 3);
+}
+
+TEST(lagi_ifind, correct_index_with_expanded_character_before_match) {
+	// U+0587 turns into U+0565 U+0582, all of which are two bytes in UTF-8
+	EXPECT_IFIND(" \xD6\x87 a ", "a", 4, 5);
+}
+
+TEST(lagi_ifind, correct_index_with_shrunk_character_before_match) {
+	// U+FB00 turns into "ff", which is one byte shorter in UTF-8
+	EXPECT_IFIND(" \xEF\xAC\x80 a ", "a", 5, 6);
+}