#include <config.h>
#include "support/docstream.h"
+#include "support/lstrings.h"
#include "support/unicode.h"
+#include <algorithm>
#include <cerrno>
#include <cstdio>
#include <cstring>
using lyx::ucs4_codeset;
+
+#if defined(_MSC_VER) && (_MSC_VER >= 1600)
+std::locale::id numpunct<lyx::char_type>::id;
+
+namespace std {
+// Implementation of numpunct<lyx::char_type> defined in numpunct_lyx_char_type.h
+typedef basic_string<lyx::char_type> string_type;
+
+string_type numpunct<lyx::char_type>::truename() const
+{
+ return lyx::from_ascii(numpunct<char>::truename());
+}
+
+string_type numpunct<lyx::char_type>::falsename() const
+{
+ return lyx::from_ascii(numpunct<char>::falsename());
+}
+
+} // namespace std
+
+#endif // _MSC_VER >= 1600
+
+
namespace {
// We use C IO throughout this file, because the facets might be used with
} else
out_cd_ = (iconv_t)(-1);
}
+ string const & encoding() const { return encoding_; }
protected:
virtual ~iconv_codecvt_facet()
{
}
virtual int do_max_length() const throw()
{
- // FIXME: this information should be transferred to lib/encodings
- // UTF8 uses at most 4 bytes to represent one UCS4 code point
- // (see RFC 3629). RFC 2279 specifies 6 bytes, but that
- // information is outdated, and RFC 2279 has been superseded by
- // RFC 3629.
- // The CJK encodings use (different) multibyte representation as well.
- // All other encodings encode one UCS4 code point in one byte
- // (and can therefore only encode a subset of UCS4)
- // Note that BIG5 and SJIS do not work with LaTeX (see lib/encodings).
- // Furthermore, all encodings that use shifting (like SJIS) do not work with
- // iconv_codecvt_facet.
- if (encoding_ == "UTF-8" ||
- encoding_ == "GB" ||
- encoding_ == "EUC-TW")
- return 4;
- else if (encoding_ == "EUC-JP")
- return 3;
- else if (encoding_ == "ISO-2022-JP")
- return 8;
- else if (encoding_ == "BIG5" ||
- encoding_ == "EUC-KR" ||
- encoding_ == "EUC-CN" ||
- encoding_ == "SJIS" ||
- encoding_ == "GBK")
- return 2;
- else
- return 1;
+ return lyx::max_encoded_bytes(encoding_);
}
private:
/// Do the actual conversion. The interface is equivalent to that of
inbytesleft, to, outbytesleft);
if (converted == (size_t)(-1)) {
switch(errno) {
+ case 0:
+ // As strange as it may seem, this
+ // does happen on windows when parsing
+ // comments with accented chars in
+ // tex2lyx. See the following thread
+ // for details
+ // http://thread.gmane.org/gmane.editors.lyx.devel/117636
+ break;
case EINVAL:
case E2BIG:
return base::partial;
}
+ifdocstream::ifdocstream(SetEnc const & enc) : base()
+{
+ setEncoding(*this, enc.encoding, in);
+}
+
+
ifdocstream::ifdocstream(const char* s, ios_base::openmode mode,
string const & encoding)
: base()
}
+ofdocstream::ofdocstream(SetEnc const & enc) : base()
+{
+ setEncoding(*this, enc.encoding, out);
+}
+
+
ofdocstream::ofdocstream(const char* s, ios_base::openmode mode,
string const & encoding)
: base()
if (has_facet<iconv_codecvt_facet>(os.rdbuf()->getloc())) {
// This stream must be a file stream, since we never imbue
// any other stream with a locale having a iconv_codecvt_facet.
+ iconv_codecvt_facet const & facet =
+ use_facet<iconv_codecvt_facet>(os.rdbuf()->getloc());
+
+ // FIXME Changing the codecvt facet of an open file is allowed,
+ // but unsafe for facets that use internal state (see the thread
+ // "iostreams: Does imbue() need to be called before open()?"
+ // in comp.std.c++.
+ // Currently it seems to work with gcc and MSVC, but not with
+ // clang on OS X.
+ // Avoid imbueing with the same encoding again if possible.
+ if (facet.encoding() == e.encoding)
+ return os;
+
// Flush the stream so that all pending output is written
// with the old encoding.
os.flush();
+
locale locale(os.rdbuf()->getloc(),
new iconv_codecvt_facet(e.encoding, ios_base::out));
- // FIXME Does changing the codecvt facet of an open file
- // stream always work? It does with gcc 4.1, but I have read
- // somewhere that it does not with MSVC.
- // What does the standard say?
os.imbue(locale);
}
return os;
}
-//CHECKME: I just copied the code above, and have no idea whether it
-//is correct... (JMarc)
idocstream & operator<<(idocstream & is, SetEnc e)
{
if (has_facet<iconv_codecvt_facet>(is.rdbuf()->getloc())) {
// This stream must be a file stream, since we never imbue
// any other stream with a locale having a iconv_codecvt_facet.
- // Flush the stream so that all pending output is written
- // with the old encoding.
- //is.flush();
+ iconv_codecvt_facet const & facet =
+ use_facet<iconv_codecvt_facet>(is.rdbuf()->getloc());
+
+ // FIXME Changing the codecvt facet of an open file is allowed,
+ // but unsafe for facets that use internal state (see the thread
+ // "iostreams: Does imbue() need to be called before open()?"
+ // in comp.std.c++.
+ // Currently it seems to work with gcc and MSVC, but not with
+ // clang on OS X.
+ // Avoid imbueing with the same encoding again if possible.
+ if (facet.encoding() == e.encoding)
+ return is;
+
locale locale(is.rdbuf()->getloc(),
new iconv_codecvt_facet(e.encoding, ios_base::in));
- // FIXME Does changing the codecvt facet of an open file
- // stream always work? It does with gcc 4.1, but I have read
- // somewhere that it does not with MSVC.
- // What does the standard say?
is.imbue(locale);
}
return is;
}
+
#if ! defined(USE_WCHAR_T) && defined(__GNUC__)
// We get undefined references to these virtual methods. This looks like
// a bug in gcc. The implementation here does not do anything useful, since
return true;
}
-#if __GNUC__ == 3 && __GNUC_MINOR__ < 4
-
-template<>
-int codecvt<lyx::char_type, char, mbstate_t>::do_length(
- mbstate_t const &, const char *, const char *, size_t) const
-{
- return 1;
-}
-
-#else
-
template<>
int codecvt<lyx::char_type, char, mbstate_t>::do_length(
mbstate_t &, const char *, const char *, size_t) const
return 1;
}
-#endif
-
template<>
int codecvt<lyx::char_type, char, mbstate_t>::do_max_length() const throw()
{
projects / lyx.git / blobdiff