Fix wxFileSystem::FileNameToURL() for Unicode file names.

[wxWidgets.git] / src / common / textfile.cpp

diff --git a/src/common/textfile.cpp b/src/common/textfile.cpp
index 40f5d4b44d10a3edb7ba7e9df6c13dc22043402d..91a86d5fb4b0e1ee6a856d87d30de9264a3ebd9d 100644 (file)
--- a/src/common/textfile.cpp
+++ b/src/common/textfile.cpp
@@ -64,7 +64,7 @@ bool wxTextFile::OnOpen(const wxString &strBufferName, wxTextBufferOpenMode Open
     switch ( OpenMode )
     {
         default:
     switch ( OpenMode )
     {
         default:

-            wxFAIL_MSG( _T("unknown open mode in wxTextFile::Open") );
+            wxFAIL_MSG( wxT("unknown open mode in wxTextFile::Open") );

             // fall through
 
         case ReadAccess :
             // fall through
 
         case ReadAccess :


@@ -86,68 +86,125 @@ bool wxTextFile::OnClose()
 }
 
 
 }
 
 

-bool wxTextFile::OnRead(wxMBConv& conv)
+bool wxTextFile::OnRead(const wxMBConv& conv)

 {
 {

-    // file should be opened and we must be in it's beginning
-    wxASSERT( m_file.IsOpened() && m_file.Tell() == 0 );
+    // file should be opened
+    wxASSERT_MSG( m_file.IsOpened(), wxT("can't read closed file") );

 
     // read the entire file in memory: this is not the most efficient thing to
 
     // read the entire file in memory: this is not the most efficient thing to

-    // do but there is no good way to avoid it in Unicode build because if we
-    // read the file block by block we can't convert each block to Unicode
+    // do it but there is no good way to avoid it in Unicode build because if
+    // we read the file block by block we can't convert each block to Unicode

     // separately (the last multibyte char in the block might be only partially
     // read and so the conversion would fail) and, as the file contents is kept
     // in memory by wxTextFile anyhow, it shouldn't be a big problem to read
     // the file entirely
     // separately (the last multibyte char in the block might be only partially
     // read and so the conversion would fail) and, as the file contents is kept
     // in memory by wxTextFile anyhow, it shouldn't be a big problem to read
     // the file entirely

-    const size_t bufSize = (size_t)(m_file.Length() + 4 /* for trailing NULs */ );
-    size_t bufPos = 0;
-    wxCharBuffer buf(bufSize - 1 /* it adds 1 internally */);
+    size_t bufSize = 0;

 
 

-    char block[1024];
-    for ( bool eof = false; !eof; )
+    // number of bytes to (try to) read from disk at once
+    static const size_t BLOCK_SIZE = 4096;
+
+    wxCharBuffer buf;
+
+    // first determine if the file is seekable or not and so whether we can
+    // determine its length in advance
+    wxFileOffset fileLength;

     {
     {

-        // try to read up to the size of the entire block
-        ssize_t nRead = m_file.Read(block, WXSIZEOF(block));
+        wxLogNull logNull;
+        fileLength = m_file.Length();
+    }

 
 

-        if ( nRead == wxInvalidOffset )
-        {
-            // read error (error message already given in wxFile::Read)
+    // some non-seekable files under /proc under Linux pretend that they're
+    // seekable but always return 0; others do return an error
+    const bool seekable = fileLength != wxInvalidOffset && fileLength != 0;
+    if ( seekable )
+    {
+        // we know the required length, so set the buffer size in advance
+        bufSize = fileLength;
+        if ( !buf.extend(bufSize) )

             return false;
             return false;

-        }

 
 

-        eof = nRead == 0;
-        if ( eof )
+        // if the file is seekable, also check that we're at its beginning
+        wxASSERT_MSG( m_file.Tell() == 0, wxT("should be at start of file") );
+
+        char *dst = buf.data();
+        for ( size_t nRemaining = bufSize; nRemaining > 0; )

         {
         {

-            // append 4 trailing NUL bytes: this is needed to ensure that the
-            // string is going to be NUL-terminated, whatever is the encoding
-            // used (even UTF-32)
-            block[0] =
-            block[1] =
-            block[2] =
-            block[3] = '\0';
-            nRead = 4;
+            size_t nToRead = BLOCK_SIZE;
+
+            // the file size could have changed, avoid overflowing the buffer
+            // even if it did
+            if ( nToRead > nRemaining )
+                nToRead = nRemaining;
+
+            ssize_t nRead = m_file.Read(dst, nToRead);
+
+            if ( nRead == wxInvalidOffset )
+            {
+                // read error (error message already given in wxFile::Read)
+                return false;
+            }
+
+            if ( nRead == 0 )
+            {
+                // this file can't be empty because we checked for this above
+                // so this must be the end of file
+                break;
+            }
+
+            dst += nRead;
+            nRemaining -= nRead;

         }
 
         }
 

-        // this shouldn't happen but don't overwrite the buffer if it does
-        wxCHECK_MSG( bufPos + nRead <= bufSize, false,
-                     _T("read more than file length?") );
+        wxASSERT_MSG( dst - buf.data() == (wxFileOffset)bufSize,
+                      wxT("logic error") );
+    }
+    else // file is not seekable
+    {
+        char block[BLOCK_SIZE];
+        for ( ;; )
+        {
+            ssize_t nRead = m_file.Read(block, WXSIZEOF(block));
+
+            if ( nRead == wxInvalidOffset )
+            {
+                // read error (error message already given in wxFile::Read)
+                return false;
+            }
+
+            if ( nRead == 0 )
+            {
+                // if no bytes have been read, presumably this is a
+                // valid-but-empty file
+                if ( bufSize == 0 )
+                    return true;
+
+                // otherwise we've finished reading the file
+                break;
+            }

 
 

-        // append to the buffer
-        memcpy(buf.data() + bufPos, block, nRead);
-        bufPos += nRead;
+            // extend the buffer for new data
+            if ( !buf.extend(bufSize + nRead) )
+                return false;
+
+            // and append it to the buffer
+            memcpy(buf.data() + bufSize, block, nRead);
+            bufSize += nRead;
+        }

     }
 
     }
 

-    const wxString str(buf, conv);
+    const wxString str(buf, conv, bufSize);

 
 

-    // this doesn't risk to happen in ANSI build
+    // there's no risk of this happening in ANSI build

 #if wxUSE_UNICODE
     if ( bufSize > 4 && str.empty() )
     {
 #if wxUSE_UNICODE
     if ( bufSize > 4 && str.empty() )
     {

-        wxLogError(_("Failed to convert file contents to Unicode."));
+        wxLogError(_("Failed to convert file \"%s\" to Unicode."), GetName());

         return false;
     }
 #endif // wxUSE_UNICODE
 
         return false;
     }
 #endif // wxUSE_UNICODE
 

-    free(buf.release()); // we don't need this memory any more
+    // we don't need this memory any more
+    buf.reset();

 
 
     // now break the buffer in lines
 
 
     // now break the buffer in lines


@@ -167,7 +224,15 @@ bool wxTextFile::OnRead(wxMBConv& conv)
                 // could be a DOS or Unix EOL
                 if ( chLast == '\r' )
                 {
                 // could be a DOS or Unix EOL
                 if ( chLast == '\r' )
                 {

-                    AddLine(wxString(lineStart, p - 1), wxTextFileType_Dos);
+                    if ( p - 1 >= lineStart )
+                    {
+                        AddLine(wxString(lineStart, p - 1), wxTextFileType_Dos);
+                    }
+                    else
+                    {
+                        // there were two line endings, so add an empty line:
+                        AddLine(wxEmptyString, wxTextFileType_Dos);
+                    }

                 }
                 else // bare '\n', Unix style
                 {
                 }
                 else // bare '\n', Unix style
                 {


@@ -192,7 +257,15 @@ bool wxTextFile::OnRead(wxMBConv& conv)
                 if ( chLast == '\r' )
                 {
                     // Mac line termination
                 if ( chLast == '\r' )
                 {
                     // Mac line termination

-                    AddLine(wxString(lineStart, p - 1), wxTextFileType_Mac);
+                    if ( p - 1 >= lineStart )
+                    {
+                        AddLine(wxString(lineStart, p - 1), wxTextFileType_Mac);
+                    }
+                    else
+                    {
+                        // there were two line endings, so add an empty line:
+                        AddLine(wxEmptyString, wxTextFileType_Mac);
+                    }

                     lineStart = p;
                 }
         }
                     lineStart = p;
                 }
         }


@@ -211,7 +284,7 @@ bool wxTextFile::OnRead(wxMBConv& conv)
 }
 
 
 }
 
 

-bool wxTextFile::OnWrite(wxTextFileType typeNew, wxMBConv& conv)
+bool wxTextFile::OnWrite(wxTextFileType typeNew, const wxMBConv& conv)

 {
     wxFileName fn = m_strBufferName;
 
 {
     wxFileName fn = m_strBufferName;