Rapid Prototyping


Usually, if I have to code something small for work, I do it in JavaScript (because we have PC’s here): It’s quite easy to program if you know the basic C-Syntax and is also quite structured compared to VBScript. Second, it runs without an additional install and you have access to lots of automation objects of the system (aka ActiveX or COM).

Recently, I was working on a script that took some text file, parsed it and built an XML tree. Unfortunately, it was limited by the lack of power of the MSXML suite.


Ideally, I would have used LibXML2 because this is a really powerful library. Unfortunately, there is no ActiveX component that wraps LibXML – and doing it myself was not worth it – because of ActiveX, not LibXML2.

So, I had to look for alternatives, languages I know that have strong XML support of some kind. Came to mind (In no order of priority):[^1]

  • Mono/.NET* PHP* Python* Perl* C

Two of them are obvious candidates for the Trash,[^2] a third (PHP) was also quickly dropped because of a lack of a decent XML parser and a general bad gut-feeling of using a web development language for a desktop application.

From the remaining two (Mono and .NET are the same, basically), I choose Python. Why? There are several reasons:

  • I know more about Python coding than about C# or any CLR language
  • LibXML2 has excellent Python bindings
  • It’s about time to learn a new programming language
  • Python and LibXML2 are both cross-platform
  • Python has WebDAV support that I understand and can leverage

(The last point is particularly important with respect to the project I’m working on)

Building the libraries#

Because Python exists for the Mac (OS X) as well, it was obvious that I would do the main development on my yummy little Aluminium beast – and not on the big, noisy PC I got from my boss. (Not that I don’t value my Boss’ efforts to provide me with the best hardware I can dream of, but I love my command-line)

A glance at the LibXML2 web site showed that there are pre-compiled binaries available as [LibXML2 for Mac OS X][]. Obviously, there are no [Python bindings][] for the version “2.6.8” of the [LibXML2 for Mac OS X][]. Bummer, I had to build them myself.

To make it short: The only things you must pay attention to when you build LibXML2 and LibXSLT – including the Python bindings – is to make sure that LibXSLT and the bindings find the right version of the library: Panther already has LibXML2 2.5.4 installed and if you don’t pay attention, either one can fail.

The configure of LibXSLT required a --with-libxml-prefix=DIR where DIR points to the location where you installed LibXML2 (I choose /usr/local/ for all that stuff) and for the Python bindings, the easiest is to modiy the setup.py:

Just search the following piece of code in setup.py and comment-out the part where it includes /usr/include (The other paths might differ, depending on your system configuration):

# those are examined to find
# - libxml2/libxml/tree.h
# - iconv.h
# - libxslt/xsltconfig.hincludes_dir = ["/usr/include","/usr/local/include","/opt/include",os.path.join(ROOT,'include'),HOME];

It should look like this (more or less):

# those are examined to find
# - libxml2/libxml/tree.h
# - iconv.h
# - libxslt/xsltconfig.hincludes_dir = [# "/usr/include","/usr/local/include","/opt/include",os.path.join(ROOT,'include'),HOME];

Once you’ve made these changes, a sudo setup.py install should run smoothly.

Using the libraries#

Well, I didn’t want to build them, I wanted to use them. There’s a nifty [Python/XML][] tutorial over at [Kimbro Staken’s blog][] that covers the basics – that was enuff for me.

There is, btw, an awful lot of nifty stuff in Kimbro’s [XML category][] over at his site. Check it out!


Building the parsing part was maybe the hardest thing to do (I decided to create a dictionary with all the values etc. first and then to write-out these in XML) but it was much easier in Python than in JavaScript.

I had done this in approx 6hrs. Adding the XML output was a matter of 1hr maybe (Mainly due to my lack of knowledge on Python and libxml programming) and so, after less than one working day, I have a very working prototype of what shall become a Really Useful Technology® for my company.

The toughest part that remains so far is to add support for the WebDAV stuff I mentioned above. This will occupy me some more days I guess, but I am confident that the tools I use allow me to achieve what I want.

After all, Python has a very rigid syntax because it is so strict with respect to whitespace. It turned out that this is more of an advantage than a hassle. It results in clean code without any special effort: You just have to structure your code, otherwise it won’t work. Sweet.

Oh, and yes, I might’ve been bored lately … :)

[^1]Java isn’t on the list because it’s not my language of choice. I see it’s power but for rapid prototyping, it’s just not suitable – and I never really got around it’s syntax.

[^2]No, not .NET and Mono. It’s Perl and C, becaus I never understood Perl’s Syntax and because C is just too clumsy for what I wanted to do. (And C isn’t a real rapid prototyping language neither)

[LibXML2 for Mac OS X]: http://www.zveno.com/open_source/libxml2xslt.html[Python bindings]: http://xmlsoft.org/python.html [Python/XML]: http://www.xmldatabases.org/WK/blog/215_XML_Document_Construction_With_Python_and_libxml2.item [Kimbro Staken’s blog]: http://www.xmldatabases.org/WK/blog/ [XML category]: http://www.xmldatabases.org/WK/blog?t=category&a=XML


No comments. Be the first to add one!
Add a comment.
We'll never share your email with anyone else. We use the Gravatar system to pull in pictures based on an anonymous hash.
Once you submit your comment, it will be moderated and then show up here shortly after.