X-Git-Url: https://git.lyx.org/gitweb/?a=blobdiff_plain;f=src%2Fsupport%2Funicode.h;h=17b95dd29079c7ad69ac7efa110647dd0cdd17a1;hb=cc2835392294fb1e82a89b2ec3803f78c9183f18;hp=8fa4ff77d95c8bb69c1f40c19e0cf92d2c92a091;hpb=6d5a52b8e36152cc8a1e37f08848102f2a81c1d6;p=lyx.git

diff --git a/src/support/unicode.h b/src/support/unicode.h
index 8fa4ff77d9..17b95dd290 100644
--- a/src/support/unicode.h
+++ b/src/support/unicode.h
@@ -1,9 +1,10 @@
+// -*- C++ -*-
 /**
  * \file unicode.h
  * This file is part of LyX, the document processor.
  * Licence details can be found in the file COPYING.
  *
- * \author Lars Gullik Bjønnes
+ * \author Lars Gullik Bjønnes
  *
  * Full author contact details are available in file CREDITS.
  *
@@ -13,49 +14,62 @@
 #ifndef LYX_SUPPORT_UNICODE_H
 #define LYX_SUPPORT_UNICODE_H
 
-#include "support/types.h"
-
-#include <boost/scoped_ptr.hpp>
+#include "support/strfwd.h"
+#include "support/unique_ptr.h"
 
+#include <cstddef>
 #include <string>
 #include <vector>
 
 
 namespace lyx {
 
+/**
+ * Wrapper for iconv(3).
+ *
+ * According to the POSIX standard, all specified functions are thread-safe,
+ * with some exceptions. The iconv() function is not listed as an exception:
+ * http://pubs.opengroup.org/onlinepubs/9699919799/xrat/V4_xsh_chap02.html#tag_22_02_09_09
+ * http://man7.org/linux/man-pages/man7/pthreads.7.html
+ *
+ * Therefore, you can use as many instances of this class in parallel as you
+ * like. However, you need to ensure that each instance is only used by one
+ * thread at any given time. If this condition is not met you get nasty
+ * mixtures of different thread data as in bug 7240.
+ *
+ * From a performance point of view it is best to use one static instance
+ * per thread for each in/out encoding pair. This can e.g. be achieved by
+ * using helpers for thread-local storage such as QThreadStorage or
+ * boost::thread_specific_ptr. A single static instance protected by a mutex
+ * would work as well, and might be preferrable for exotic encoding pairs.
+ * Creating local IconvProcessor instances should be avoided because of the
+ * overhead in iconv_open().
+ */
 class IconvProcessor
 {
-public:
-	IconvProcessor(
-		char const * tocode = "",
-		char const * fromcode = "");
-	/// copy constructor needed because of pimpl_
-	IconvProcessor(IconvProcessor const &);
-	/// assignment operator needed because of pimpl_
-	IconvProcessor & operator=(IconvProcessor const &);
-	/// destructor (needs to be implemented in the .C file because the
-	/// boost::scoped_ptr destructor needs a fully defined type
-	~IconvProcessor();
-
-	/// convert any data from \c fromcode to \c tocode unicode format.
-	/// \return the number of bytes of the converted output buffer.
-	int convert(
-		char const * in_buffer,
-		size_t in_size,
-		char * out_buffer,
-		size_t max_out_size);
-private:
 	/// open iconv.
 	/// \return true if the processor is ready to use.
 	bool init();
-
-	std::string tocode_;
-	std::string fromcode_;
-
-	struct Private;
-	::boost::scoped_ptr<Private> pimpl_;
+	std::string const tocode_;
+	std::string const fromcode_;
+	struct Handler;
+	unique_ptr<Handler> h_;
+public:
+	IconvProcessor(std::string tocode, std::string fromcode);
+	/// convert any data from \c fromcode to \c tocode unicode format.
+	/// \return the number of bytes of the converted output buffer.
+	int convert(char const * in_buffer, size_t in_size,
+		char * out_buffer, size_t max_out_size);
+	/// target encoding
+	std::string to() const { return tocode_; }
+	// required by g++ 4.6
+	IconvProcessor(IconvProcessor && other);
 };
 
+/// Get the global IconvProcessor instance of the current thread for
+/// utf8->ucs4 conversions
+IconvProcessor & utf8ToUcs4();
+
 // A single codepoint conversion for utf8_to_ucs4 does not make
 // sense, so that function is left out.
 
@@ -71,6 +85,10 @@ std::vector<char_type> utf16_to_ucs4(unsigned short const * s, size_t ls);
 
 std::vector<unsigned short> ucs4_to_utf16(char_type const * s, size_t ls);
 
+/// Get the global IconvProcessor instance of the current thread for
+/// ucs4->utf8 conversions
+IconvProcessor & ucs4ToUtf8();
+
 // ucs4_to_utf8
 
 std::vector<char> ucs4_to_utf8(char_type c);
@@ -86,8 +104,8 @@ eightbit_to_ucs4(char const * s, size_t ls, std::string const & encoding);
 
 /// convert \p s from ucs4 to encoding \p encoding.
 /// \p encoding must be a valid iconv 8bit encoding
-std::vector<char>
-ucs4_to_eightbit(char_type const * ucs4str, size_t ls, std::string const & encoding);
+std::vector<char> ucs4_to_eightbit(char_type const * ucs4str,
+	size_t ls, std::string const & encoding);
 
 /// convert ucs4 character \p c to encoding \p encoding.
 /// \p encoding must be a valid iconv 8bit encoding
@@ -99,6 +117,8 @@ void ucs4_to_multibytes(char_type ucs4, std::vector<char> & out,
 
 extern char const * ucs4_codeset;
 
+/// How many bytes does one UCS4 code point use at most in encoding \p encoding?
+int max_encoded_bytes(std::string const & encoding);
 
 } // namespace lyx