Discussion:

• Sketch-based AToM³:
• Python Limitation: The number of arguments in a function in Python cannot exceed 255. This means a stroke cannot have more than 126-127 points in it! This is because a stroke is created using Tkinter Canvas' method create_line. I am already minimizing the number of points in a stroke by ignoring a point that is too close to its previous one (a point is too close if both of its x and y values are +/- 3 of the previous one).
  Professor Vangheluwe sugessted having a wrapper around Tkinter's Line object to get around this problem. In this wrapper if a line has more than 125 points or so we break it into more than one line and we keep a list of handlers of those lines.
• Recognizers: I've read few papers in this area and added them to the links page. The LADDER language seems to be the most inspiring one from what I read so far. However, my list of papers to read is far from being empty. LADDER suggests having a domain-independant recognizer that recognizes primitive shapes and then we supplement it by a domain-specific grammar to describe the objects that need to be recognized. Professor Vangheluwe suggested having a DChart model that describes the object. Maybe we can push this even further and actually be able to combine the context meta-model with that DChart and this way we can know right away if a drawing is allowed in the current context or not. If we were able to get this to work we will try to submit a paper to CHI about it.
• GUI-specific Formalism: This is the formalosm that we discussed in our previous meeting. At that time we talked about implementing this formalism and writing a paper about it. But the above mentioned idea for describing pattern recognitions seems to be promising. So we decided to try that idea first and if we saw that it is not feasible in the given amount of time (CHI deadline is in September) then we can revert back to publishing about this formalism.

Action Items:

• Python Limitation: Implement the work around described above to get around Python's limitation of having a maximum of 255 arguments for a function.
• Modelling patterns description: Try the idea described above in the Recognizers section and see how it works. We will meet on Saturday June 11 to discuss it. Also, I need to read the papers about GenGed and DiaGen to understand how LADDER's approach is different from theirs.