More typing algorithms
I discussed today with Prof. Wadler what path to follow for the typing of the first prototypal Links language, so that I can start the database part of my work. Even though Rémy's and the updated Wand algorithm provide good typing results, they might be quite complicated to implement.
Wand's original algorithm is wrong because of one operator in the language it must support: it allows a 'with' operator on records that can both update an existing field or add a new field, two operations that have to be typed differently. Currently, Links is not supposed to support such an operator and CPL doesn't provide at all such a complex record typing mechanism.
I therefore started today looking for yet another typing algorithm that would be at the same time simple enough (like Wand's original algorithm), correct and powerful. I have plenty of papers to look at, and I am by far not finished yet. I will publish what I find in tomorrow's post here.