Discussion questions from CS 7390 lecture

Date: 1/31/96

Presented by: Victor

Topic: Algorithm Animation

• The message passing components reuse is very close to object-oriented programming concepts. An algorithm animation system for an object oriented language myst be very natural for the programmers
• How animation components can be reused in smalltalk systems ?
• Most papers show typically a sorting example ( Why ? Historical ? Appropiate ? Benchmark) more general question: how well does it generalize to complex algorithm ? Should there be some 'benchmark', recognized with various specific and quantifiable measure
• Could someone please explain what is fundamentally different in having Animation program in Smalltalk ?
• In previous classes, it was mentioned that Balsa could only do discrete algorithm animations. Can't the view update routines be made to display continuous animations and if not, what implications does this have on the user ?
• In the Balsa paper, the example of Quicksort was given to explan how the different pieces of the system work together. Example "code" was given to each peice except for the modeller. What is involved in getting the modeller to produce output for the renderer and is all communication to the modeller handled in the event router ?
• The Balsa system provides for auto-genration of input but it seems limited to the case of sorting algorithms. Would an object oriented approach be more general, wher the object could generat its won random data, more suited to the problem domain ?
• What exactly is the MVC model ? How do they apply it specifically. They are very vague in the paper about what it si and how they use it.
• The Balsa II paper briefly mentions algorithm runtime analysis capabilities by counting loop entires. The smalltalk paper doens't mention anything. How can runtime analysis be animated (big - O notation) ?
• Could Balsa come closer to automatic animation by providing a large variery of librairs or external packages for implementing events that the event annotations simply hook into ?
• C++ makes a big deal about being "as efficient as C (as long as you don't use certain features)". Adding methods to objects to support animation seems like a convenient way to do things, byt you probabbly don't want all that extra baggage in the "real" code that gets used. Can one cleanly divide the code from the annotations in such a way that a user of a library can choose between the quick version or the slow but animated version ? I'm still a little confused by the M-V_C thing and the extensions proposed to it in the papers. Where does M-V-C break down ?
• The modeller , renderor approach taken by Marc Brown, it seems to me that is very similar to the small-talk Model-View Controller Architecture. I wasn't too sure what the differenc was . Is thery any special reason why he used such terminology ?
• Is it always possible to monitor smalltalk message to instrument it for algorithm animation ?