2023 02 21 meeting

Meeting 2023-02-21: How To Measure Our Performance

Jan asks: would a complete procedure be enough for a thesis? I say: yes.

His suggestion as to how to go about incompleteness at this time is sampling. Meaning, if you see an existential, try to stick some types in there and see what happens. I tried to do something like that in the past. His novelty is that you could sample not only from the stdlib or the package itself, but also from some database populated from many packages. This sounds a bit tricky to implement but perhaps working “better”.

What do we mean by “better”? Tricky question. I tried to explain Jan that rather than trying to improve (whatever that means) the procedure I wanted to first figure out how to measure its “goodness”. The approach I (want to) take is to match it with results of dynamic analysis (tests).

Jan is moderately receptive. His question is: what questions can we answer using it? I don’t have a convincing story here, but he seems to be receptive to talking about it and probably finding something.