My code is in and done. Been for a little while now. She works like a charm :-D and I am very proud. I should take this time to thank Angie, Robert, Ted, Jeff and anyone else involved in Drupal for selecting me and giving me this opportunity. It has been a blast and what a way to spend the summer.

Drupal had 14 projects and all are ready for public testing : http://drupal.org/node/78942

I have taken the money that I have recieved and spent it on a robot (thats right, a robot). I figured it was time for a new one, this one will feature vision, speech recongition (just simple commands like "Robot, turn right"). O and it will be controllable from a web interface (perhaps a drupal module < evil laugh >).

Long time since I have posted in here. Lots has changed (though, surprisly still waiting on a check..hmm..). I have Drupalized the Movielens database and it has shown some HUGE performance problems. I think I have found them and they will be squashed!

I plan on implementing a query object much like the views does. This will allow modules that implement my recommendation api to edit the query through hooks.

Install process has been revamped...again! It runs through and installs incrimentally through cron runs. Scary I know but

But other then that, everything is grand.

I am ahead of schedule! Never thought I would be typing that but its true. My module works, fully. I have been messing with views intergration, which is what nuked my drupal installation. Just a test server with about 3k users and 1k votes. I have gotten the Movielens database ready to be 'drupalized'. Just haven't had the guts to do it. Thats 1 million votes. I know that when my module installs it is going to 'fart' and not finish loading all the data for a site that LARGE!

Personally, I really want other developers to pick up my engine, and develop modules to use it. Right now I have developed node_recommendation.module that utilizes the engine and recommends nodes and only nodes. But I know people use voting data for things other then nodes. And even if they are nodes, they can still break it down to vote only on say movies and not consider any other nodes. Pretty simple to do that with the votingapi.module and cre.module (content_type stored by the voting widget must be 'movies' and the content_type recommended by the module must have content_type of 'movies'). In fact, this ability should be a part of the node_recommendation.module.I shall do that now :-D ....

I am not a fan of setting goals. Put simply, I believe in setting one and only one goal, make it work right. But I think its apporiate to iron out what my goals should be for midterm evals. So...

As an experimental and academic project, I should first spend much time on researching some basic theory and some existing methods performing memory management optimization. Gathering information about the theory may take about 2 weeks time.

And then, I should get familiar with the archetecture of Mark's code. Luckily, Shed Skin has no more than 6k lines of code. The two main versions for research is ss0.0.10--the latest release without memory management--and ss0.0.6--with premature mm.

And my first code would be porting the escape analysis of ss0.0.6 to ss0.0.10, this would be a good b

THis summer, I'm doing a project called "Improvement to Shed Skin Project". SS project intends to develop a optimized compiler translates Python code to C++ code.

Python is a slow but productive language, C++ runs efficiently but is not as easy to code as scripted language such as Python or Ruby. So, SS can fill the gap between productive but slow Python and hard to code but efficiency C++.

My task is to optimize SS's memory management component. The original implementation is very trivial and doesn't make much sense. After the optimization, the generated C++ code can run even faster than present!

Well I have done several cvs commits. But the real big one is coming soon! Right now you can see my current code, which works PERFECTLY (or so I assume) with nodes. It is still very basic but it will and can recommend only certain types of nodes (i.e. story, or page, or whatever you want) and it can recommend the best nodes no matter the type.

It also has a function that recommends items that are similar to the current node.

This is how it is set up....

O man I got it to work! I now can get the top-n recommeded nodes for the current user! This is exciting! I created a _block() hook so someone could display it where ever.

Ok but it will recommend nodes that the user has already voted on but that is just a small detail.

Now we are getting close to having something to commit to CVS.

Google Summer of Code 2006 has hereby forced Teresa Thomas, the eternal blog-hater to start her own blog. Quite an achievement on their part. My only hope now is that as time goes on, I change my mind about the concept of "penning down every single little thought that crosses my head for the world to 'admire' ''.

Let me begin by introducing myself. The name is Teresa and nop, I am not always this bitter. I just completed my second year in Electrical and Computer engineering at the University of Toronto, Canada. This summer, other than SoC, I am researching in Integrated Optics at my uni. For SoC, I am working with Ubuntu and my project involves implementing synchronication plugins for the Evolution Calendar. I plan to start with plugins that will sync it with the BlackBerry and the Google calendar, and if time permits, add more to the list.