On Sat, Oct 15, 2022 at 2:58 AM Jan-Pieter Jacobs <janpieter.jac...@gmail.com> wrote: > - The moment you apply verbs to arguments at the time of graph > construction, either monad or dyad, you loose the capability of reason > about the verb's other valence.
I think you should be careful here: Information is lost when you discard the verb's identity and/or the original arguments. You need that information not only to identify valence (which can be significant when the surrounding valence constraining context is ambiguous) but also to quantify the effect of an epsilon variation in the argument(s) (which is what calculus is "all about"). But at some point you need to work on finding the desired result and once you have done enough work on that issue, intermediate results can (and should) be discarded. The details, of course, should depend on the operation you're performing and any resource constraints you're working under. > It seems to me that adding them as equivalent classes later using rewrite > rules does not seem difficult and preserves the original as well. I think that you're right that rewrite rules are the way to go here. But -- at least in principle -- they can be applied as soon as their constraints are satisfied. That said, especially initially, simplicity is important for keeping things straight. (Once you've got things working and especially when you have a good test suite to help you keep track of your issues, blowing it up into higher levels of complexity can be an important intermediate step in rearranging things into an even simpler approach.) I hope this makes sense. Thanks, -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
