Any node can define namespaces:
<h:html xmlns:h = "html4" xmlns:h = "html4"> <body> <m:math xmlns:m = "mathml2">...</m:math> <h:span>...</h:span> <mml:math xmlns:m = "mathml2">...</mml:math> </body> </h:html>
so namespace lookup would propagate along the parents of each node until a matching namespace is found. This would mean that each node must store a (smart) pointer to a namespace mapping, in order to facilitate lookup (done during node construction). The node will also have a pointer to a namespace information object that looks something like:
class namespaceInfo { std::string url; std::string name; };
Have I got this wrong?
no, it's correct...on a conceptual level.
The point I'm trying to make is that libxml2 does all this for me already. The node itself doesn't store all the active namespaces, as that would be a horrible waste. You have to query them (and therefor you'll access the document and parent nodes). For example, if a node is moved from one place to another, its context will change and therefor some necessary adjustments have to be made notably concerning the namespaces. A look into the libxml2 source code convinced me that it's quite a tricky job...
I also like the xml::dom::document document = xml::dom::parse_file(argv[1]); style usage. Also: xml::dom::element e = doc.documentElement(); xml::dom::node n = e;
This is just my preferred style/usage, and does not have to be adopted.
you mean you vote for the 'nodes-are-references' style ?
This would, however, allow the implementation to be adapted to when memory management needs to be used and to switch between different management policies for the various implementations.
Can you elaborate ?
Regards,
Stefan_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
