c/c++ XML library question

Question

I know that a lot of c/c++ XML library questions have been asked already (I tried to read through all of them before getting to this).

Here are the things I'm going to need in my own project:

• Excellent performance
• SAX2
• Validation
• Open source
• Cross platform

I was going to use Xerces-C, but I see that a simple SAX2 setup with nothing going on in the filter is taking 5 seconds to run. (Perhaps I'm doing something wrong here?)

I would like to use libxml++, but as I tried to get it set up on my MacBook, there were some crazy dependencies that took me all the way back to gtk-doc, at which point I sort of tabled the idea.

So now I'm at libxml2. Is this the way to go? Have I missed an important option, bearing in mind the five requirements above? I don't mind using a (good) c-library like libxml2, but a c++ interface would be nice. (I don't like Xerces-C's API very much.)

I am willing to bend on the SAX2 requirement if comparable functionality is available.

Answer 1

If you need a whole host of other facilities, you can consider Qt, which has good XML support. Be warned though, it's WAY more than an XML processing library; it's a full blown application framework with support for GUIs, networking and a whole host of other things.

Qt

You can also try Poco. It's another application framework, but not as huge as Qt (i.e. no GUI-related things etc.)

Poco

Lastly, if you don't mind a C library, you can use Expat. It's not SAX per se, but writing code using Expat is somewhat like SAX. It has C++ wrappers, but they're not officially part of the project IIRC, and may not be as well-maintained or designed. I'm not too sure though.

Expat

Hope this helps!

EDIT: I misread your original post: not too sure about the validation features of these libraries, I've never used them before.

Answer 2

Having spent a goodly amount of time on this same problem, it was my conclusion that libxml2 is the best option available under your guidelines. The C interface is not too difficult to use and it's very fast.

There are some other good options for commercial libraries, but most of the other comparable open-source options are either painfully slow or are mired in a deep, annoying vat of dependency soup.

Answer 3

You say you need these things in your project, but don't give any idea of the pipeline. For example, we had a whole load of static XML files which needed to be loaded quickly, but only validated rarely. So validated using a separate process in batch (using RelaxNG as it was human writable markup ) and loaded the XML using expat. The system also used XMPP, so checked streaming input, but that didn't require validating against a schema (partly because it was streamed, and mostly because most of the possible errors were not expressible in a schema).