My aim is to provide C++ developers with a portable lightweight DOM implementation depending only on the STL and Boost libraries.
A few months ago, I had to write such a thing from scratch for Windows Mobile and SymbianOS. A preliminary implementation hence exists (see
MiniDOM/old-dom folder), but it heavily depends upon closed-source components that are owned by my employer.
The old version supported asynchronous loading of XML document, automatic decoding of various charsets (utf-8, utf-16/ucs-2, ucs4, latin1, latin9, win1252 and plain ASCII), and XPath queries.
For simplicity, and because the design of that old version sucked (really... if you don't believe me, get the source and read the DOMReader class!), the upcoming release will only support synchronous operations. I will re-introduce asynchronous loading once I add support for DOM events.
The main limitation comes from a design decision, in order to keep a simple memory management and get all the benefits of automatic memory reclaiming without having to use/write/depend upon a garbage collector. As a result, you must store the Document object somewhere while you want to manipulate its nodes and only keeping a child Node won't suffice.
- Right now, creating and manipulating DOM nodes should work. However, using the library for other things beside testing is not recommended right now (think of it as pre-alpha code).
- The DOMReader class has just been posted in trunk, and I have yet to test it. The old class worked and this is a port/boostifization/cleanup so there shouldn't be any big problem with it.
- The library compiles with Visual C++ 2005 SP1 on Windows and with GCC 4.1.2 on a Linux Gentoo box, so it might as well compile when using another compiler. Portability is a priority, but testing all the compilers is obviously not one at this stage of development.