Collada loader now handles texture-names with escape characters.

Names like "my%20texture.png" will now load "my texture.png". Collada filenames are in xs:anyURI format.
xs:anyURI is used in more places, but we don't support any other file-loading inside Collada so far, so that was the most important place to fix.
Also added/fixed a few comments.


git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@5793 dfc29bdd-3216-0410-991c-e03cc46cb475

changes.txt

@@ -1,5 +1,6 @@
 --------------------------
 Changes in 1.9 (not yet released)
+- .dae/Collada reader/writer now handle whitespace in texture-filenames with escape characters (handle them as xs:anyURI).
 - Add function string::insert.
 - EditBox now still allows overwriting characters when the text-length is at max.
 - Bugfix: CMatrix4::transformPlane was calculating the wrong plane-normal before. 
@@ -9,7 +10,7 @@ Changes in 1.9 (not yet released)
 - Bugfix: CCameraSceneNode resets the IsOrthogonal flag to false now when it recalculates a projection matrix.
 - SViewFrustum::setFrom and SViewFrustum constructor now use a parameter to allow to set the correct near clipping plane when the projection matrix doesn't use a target depth range of 0 to z, but for example -z to z. So OGL projections matrices can now also use it.
 - Remove code to read boundingbox element in Collada reader as it's not in Collada specification.
-- .dae/Collade reader now converts from Collada's right-handed to Irrlicht's left handed coordinate system (switching z to -z)
+- .dae/Collada reader now converts from Collada's right-handed to Irrlicht's left handed coordinate system (switching z to -z)
 - Add irr::string::eraseTrailingFloatZeros to kick out trailing 0's for strings generated from floats.
 - .dae/Collada writer now converts from Irrlicht's left-handed to Collada's right-handed coordinate system (switching z to -z)
 - Switch Collada writer to utf8 xml's.

include/IColladaMeshWriter.h

@@ -154,6 +154,7 @@ namespace scene
 	//! Callback interface to use custom names on collada writing.
 	/** You can either modify names and id's written to collada or you can use
 	this interface to just find out which names are used on writing.
+	Names are often used later as xs:anyURI, so avoid whitespace, '#' and '%' in the names.
 	*/
 	class IColladaMeshWriterNames  : public virtual IReferenceCounted
 	{
@@ -165,7 +166,7 @@ namespace scene
 		/** Note that names really must be unique here per mesh-pointer, so
 		mostly it's a good idea to return the nameForMesh from
 		IColladaMeshWriter::getDefaultNameGenerator(). Also names must follow
-		the xs::NCName standard to be valid, you can run them through
+		the xs:NCName standard to be valid, you can run them through
 		IColladaMeshWriter::toNCName to ensure that.
 		\param mesh Pointer to the mesh which needs a name
 		\param instance When E_COLLADA_GEOMETRY_WRITING is not ECGI_PER_MESH then
@@ -177,7 +178,7 @@ namespace scene
 		/** Note that names really must be unique here per node-pointer, so
 		mostly it's a good idea to return the nameForNode from
 		IColladaMeshWriter::getDefaultNameGenerator(). Also names must follow
-		the xs::NCName standard to be valid, you can run them through
+		the xs:NCName standard to be valid, you can run them through
 		IColladaMeshWriter::toNCName to ensure that.
 		*/
 		virtual irr::core::stringc nameForNode(const scene::ISceneNode* node) = 0;
@@ -190,7 +191,7 @@ namespace scene
 		instances per node are identical between different nodes you can reduce
 		the number of exported materials using that knowledge by using identical
 		names for such shared materials.
-		Names must follow the xs::NCName standard to be valid, you can run them
+		Names must follow the xs:NCName standard to be valid, you can run them
 		through IColladaMeshWriter::toNCName to ensure that.
 		*/
 		virtual irr::core::stringc nameForMaterial(const video::SMaterial & material, int materialId, const scene::IMesh* mesh, const scene::ISceneNode* node) = 0;
@@ -382,7 +383,7 @@ namespace scene
 			return DefaultNameGenerator;
 		}
 
-		//! Restrict the characters of oldString a set of allowed characters in xs::NCName and add the prefix.
+		//! Restrict the characters of oldString a set of allowed characters in xs:NCName and add the prefix.
 		/** A tool function to help when using a custom name generator to generative valid names for collada names and id's. */
 		virtual irr::core::stringc toNCName(const irr::core::stringc& oldString, const irr::core::stringc& prefix=irr::core::stringc("_NC_")) const = 0;
 

source/Irrlicht/CColladaFileLoader.cpp

@@ -332,6 +332,22 @@ CColladaFileLoader::CColladaFileLoader(scene::ISceneManager* smgr,
 	setDebugName("CColladaFileLoader");
 	#endif
 
+	// Escape characters, see https://www.w3schools.com/tags/ref_urlencode.asp
+	// TODO: There should be more, but usually people just escape the space character anyway.
+	//       And I'm not sure if our xml files are utf-8 or windows-1252, so let's avoid the ambiguous ones.
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL(' ', "%20"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('"', "%22"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('#', "%23"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('$', "%24"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('%', "%25"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('&', "%26"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('\'', "%27"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('(', "%28"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL(')', "%29"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('/', "%2F"));
+	EscapeCharsAnyURI.push_back(EscapeCharacterURL('\\', "%5C"));
+
+
 	TextureLoader = new CMeshTextureLoader( FileSystem, SceneManager->getVideoDriver() );
 }
 
@@ -1209,6 +1225,7 @@ void CColladaFileLoader::readImage(io::IXMLReaderUTF8* reader)
 					reader->read();
 					image.Source = reader->getNodeData();
 					image.Source.trim();
+					unescape(image.Source);
 					image.SourceIsFilename=true;
 				}
 				else
@@ -2742,8 +2759,8 @@ void CColladaFileLoader::clearData()
 //! changes the XML URI into an internal id
 void CColladaFileLoader::uriToId(core::stringc& str)
 {
-	// currently, we only remove the # from the begin if there
-	// because we simply don't support referencing other files.
+	// Currently, we only remove the # from the beginning
+	// as we don't support referencing other files.
 	if (!str.size())
 		return;
 
@@ -2942,6 +2959,47 @@ core::matrix4 CColladaFileLoader::flipZAxis(const core::matrix4& m)
 	return matrix;
 }
 
+void CColladaFileLoader::unescape(irr::core::stringc& uri)
+{
+	u32 len = uri.size();
+	for (u32 i=0; i<len-1; ++i)
+	{
+		if (uri[i] == '%' )
+		{
+			if (uri[i+1] == '%')
+			{
+				++i;
+				continue;
+			}
+
+			for (u32 e = 0; e < EscapeCharsAnyURI.size(); ++e)
+			{
+				const irr::core::stringc& escapeString = EscapeCharsAnyURI[e].Escape;
+				const u32 escapeLen = escapeString.size();
+				bool equals = true;
+				for ( u32 c = 1; c<escapeLen; ++c)	// string compare (and we already know first on fits as always '%')
+				{
+					if ( uri[i+c] != escapeString[c] )
+					{
+						equals = false;
+						break;
+					}
+				}
+
+				if ( equals )
+				{
+					uri[i] = EscapeCharsAnyURI[e].Character;
+					// TODO: core::string has no erase function which erases more than one char at a time currently
+					for ( u32 a=0;a<escapeLen-1; ++a)
+						uri.erase(i+1);
+					len -= escapeLen-1;
+					break;
+				}
+			}
+		}
+	}
+}
+
 
 } // end namespace scene
 } // end namespace irr

source/Irrlicht/CColladaFileLoader.h

@@ -344,6 +344,9 @@ private:
 	//! Note that function is symmetric (no difference if called before or after a transpose).
 	core::matrix4 flipZAxis(const core::matrix4& m);
 
+	//! replace escape characters with the unescaped ones
+	void unescape(irr::core::stringc& uri);
+
 	scene::ISceneManager* SceneManager;
 	io::IFileSystem* FileSystem;
 
@@ -369,6 +372,19 @@ private:
 	core::array< core::array<irr::scene::IMeshBuffer*> > MeshesToBind;
 
 	bool CreateInstances;
+
+	struct EscapeCharacterURL
+	{
+		EscapeCharacterURL(irr::c8 c, const irr::c8* e)
+			: Character(c)
+		{
+			Escape = e;
+		}
+
+		irr::c8 Character;		// unescaped (like ' ')
+		irr::core::stringc Escape;	// escaped (like '%20')
+	};
+	irr::core::array<EscapeCharacterURL> EscapeCharsAnyURI;
 };
 
 

source/Irrlicht/CColladaMeshWriter.cpp

@@ -178,7 +178,7 @@ irr::core::stringc CColladaMeshWriterNames::nameForMesh(const scene::IMesh* mesh
 irr::core::stringc CColladaMeshWriterNames::nameForNode(const scene::ISceneNode* node)
 {
 	irr::core::stringc name;
-	// Prefix, because xs::ID can't start with a number, also nicer name
+	// Prefix, because xs:ID can't start with a number, also nicer name
 	if ( node && node->getType() == ESNT_LIGHT )
 		name = "light";
 	else
@@ -1225,7 +1225,7 @@ bool CColladaMeshWriter::isXmlNameChar(c8 c) const
 		 */
 }
 
-// Restrict the characters to a set of allowed characters in xs::NCName.
+// Restrict the characters to a set of allowed characters in xs:NCName.
 irr::core::stringc CColladaMeshWriter::toNCName(const irr::core::stringc& oldString, const irr::core::stringc& prefix) const
 {
 	irr::core::stringc result(prefix);	// help to ensure id starts with a valid char and reduce chance of name-conflicts
@@ -1260,7 +1260,7 @@ const irr::core::stringc* CColladaMeshWriter::findGeometryNameForNode(ISceneNode
 	return &colladaMesh.findGeometryNameForNode(node);
 }
 
-// Restrict the characters to a set of allowed characters in xs::anyURI
+// Restrict the characters to a set of allowed characters in xs:anyURI
 irr::core::stringc CColladaMeshWriter::pathToURI(const irr::io::path& path) const
 {
 	irr::core::stringc result;
@@ -1280,7 +1280,7 @@ irr::core::stringc CColladaMeshWriter::pathToURI(const irr::io::path& path) cons
 	}
 	result.append(path);
 
-	// Make correct URI (without whitespaces)
+	// Make correct URI (without whitespace)
 	u32 len = result.size();
 	for (u32 i=0; i<len; ++i)
 	{
@@ -1404,7 +1404,7 @@ void CColladaMeshWriter::writeMaterialEffect(const irr::core::stringc& materialf
 		//          <init_from>internal_texturename</init_from>
 					Writer->writeElement("init_from", false);
 					irr::io::path p(FileSystem->getRelativeFilename(layer.Texture->getName().getPath(), Directory));
-					Writer->writeText(toNCName(irr::core::stringc(p)).c_str());
+					Writer->writeText(toNCName(irr::core::stringc(p)).c_str());	// same ID for internal name as in writeLibraryImages
 					Writer->writeClosingTag("init_from");
 					Writer->writeLineBreak();
 

source/Irrlicht/CColladaMeshWriter.h

@@ -93,7 +93,7 @@ public:
 	//! writes a mesh
 	virtual bool writeMesh(io::IWriteFile* file, scene::IMesh* mesh, s32 flags=EMWF_NONE) _IRR_OVERRIDE_;
 
-	// Restrict the characters of oldString a set of allowed characters in xs::NCName and add the prefix.
+	// Restrict the characters of oldString a set of allowed characters in xs:NCName and add the prefix.
 	virtual irr::core::stringc toNCName(const irr::core::stringc& oldString, const irr::core::stringc& prefix=irr::core::stringc("_NC_")) const _IRR_OVERRIDE_;
 
 	//! After export you can find out which name had been used for writing the geometry for this node.
@@ -269,8 +269,8 @@ protected:
 			Escape = e;
 		}
 
-		irr::c8 Character;
-		irr::core::stringc Escape;
+		irr::c8 Character;		// unescaped (like ' ')
+		irr::core::stringc Escape;	// escaped (like '%20')
 	};
 	irr::core::array<EscapeCharacterURL> EscapeCharsAnyURI;
 };