My thesis is slowly getting completed. Today I had a spare hour, so I decided to have a look at the bigger picture of the thesis in terms of statistics. There is a hard limit of 100,000 words for the thesis, and unfortunately mine is currently at 116,257... it looks like I have to do some serious editing work! (A lot of those words are from images, considered to be 250 words/image.)

To understand where I could most efficiently remove content, I made some statistics about the number of pages per chapter:



And also, the number of words per chapter:



It looks like I will have to do some serious editing in the Model chapter. These statistics are from the third draft of my thesis; I have one more iteration before I submit. I plan to submit in approximately four weeks time, and I'm pretty sure I can make it.

Today I did some research into why my Drools rules were running out of heap space in Java, even though there were not many rules (less than 300), and not many objects to iterate over (less than 10,000).

I found the problem was that I was using eval() in situations where it should not be used. So I wrote up an article with lots of pretty graphs.

Since I have started using StatSVN, I have made a script that will update the statistics automatically every week, which are then available online:

I've always been curious about obtaining useful (and pretty) statistics from my Subversion logs in my project, both the implementation side and the research side. But I hadn't found anything useful, until this afternoon.

Today I had a go with StatSVN, which analyses your development logs (and hammers the SVN server a little bit to get all the diffs) to create some pretty graphs.

So now you can see the development history of IAML as of r1808 (today).

The lines of code graph is interesting, although in this case, LOC is a measure of the physical lines of code - so this includes whitespace and comments. I also like the graph of developer activity.

Once I get an automated build script working, it would be pretty neat to add this as an automated task

Today, after talking with my supervisor, I decided to no longer try and submit a paper to the APCCM 09 conference. While I would have liked to attend, I don't think I could have submitted anything original enough to warrant inclusion in a modelling conference.

One of the key parts of my work is in validating that a RIA modelling language is actually better than existing languages or approaches. Because we are working with visual models, we cannot directly apply LOC metrics to different implementations; however, von Pilgrim (2007) discusses a really interesting quantitative approach to measuring the complexity of different model-driven development approaches.

Ideally, we want a transformation chain complexity closer to the solid line than the dashed line (image from the paper):



Unfortunately, it's difficult otherwise to evaluate our work because as we have shown before, there is no existing modelling language that addresses the requirements of RIAs; our only implementation of Ticket 2.0, for example, is directly in PHP/JS, and it's difficult (if not impossible) to compare visual and texual approaches in any meaningful way.