QuoteOriginally posted by: ametranoIt should be a screencast, since Luigi's talk was an hands-on live session. It's delayed because Luigi is probably adding the voice-over.I look forward to see how he will handle his accidental off-stage falling (or more accurately in-stage falling): it was so acrobatic to almost appear as staged ;-)It's a screencast. I've recorded my screen, but the voice was recorded from the microphone line, so I have to put the two together.I'm not doctoring the sound. Brownie points to the first who finds the time at which I find the hole in the stage.

Ok, the screencast is online.

Opa! @ 21.17. I guess it was quite entertaining, since everyone was clapping.Interesting discussion @1.00 and @51.30. Thanks for making it available.The whole Python things looks very cool.Is Python+NumPy+whatever is a better alternative to Matlab for prototyping?As I understand it, Python has the following advantages:0) Open source. Huge user base. Subject to scrutiny from many experts/advanced users.1) More powerful syntax. Better support for data structures and algorithms, input/output..2) Easier interoperability with native code (and Java/C# ?).3) Boost.Python4) Tons of open source high quality libraries.5) Interoperability with R.6) What else?

A general question: do the developers of QL do code inspection/review after they have gotten the original code working?An example is the Cubic interpolation in QL which has not implemented Akima and the code for Hyman89 which has many if else in a loop, leading to jumps and less efficient code?And it took me a finite amount to time to 're-engineer' the code.An UML diagram would have worked wonders...

QuoteA general question: do the developers of QL do code inspection/review after they have gotten the original code working?I made some (very few) contributions to QuantLib and my code were reviewed (by Ferdinando, I believe). The code was change based on the review and we discussed test suite conditions as well.Other guys may be had different experiences depending of the amount of code involved.

We reversed engineered code and models and now we can go in forward drive mode by just adding new classes. UML Pictures are nice because they help us document the design rationale. Now all I need to do is implement a few virtual functions. C++ or C# is almost the same code.