On Fri, May 21, 2021 at 4:36 PM Ralph Mellor <ralphdjmel...@gmail.com> wrote: > > * Things multiply. If *two* function arguments are junctions, and > each junction has, say, three elements, the function is set to be > called *six* times.
That should have been *nine* times. > * The combination has to follow precedence rules. I forget what > they are, but the general scheme is that there are two levels > and they do what's most natural. This will of course sound very > odd so I'll leave it as a thing to explore if you explore this. :) I just looked in the doc and didn't see discussion of this. And my investigations began to confuse me. ---- Here's a relatively gentle start focusing on the *operator precedence* of the *junction literal constructor operators*: ``` say 1 & 2 | 3 & 4; # any(all(1, 2), all(3, 4)) say 1 | 2 & 3 | 4; # any(1, all(2, 3), 4) ``` So the `&` junction op has higher operator precedence than `|`. This is covered in the doc, starting at: https://docs.raku.org/language/operators#Junctive_AND_(all)_precedence The compiler rejects mixing of `|` or `&` infix ops with `^` without parens to disambiguate, and there is no `none` infix op, so the above is the only precedence issue to consider for the junction literal creation infix ops. ---- Next comes combining two or more junctions with other ops / functions. The focus now is on "precedence" due to combining junctions with functions (or operators) that aren't themselves junction constructors: So this isn't about *operator* precedence but instead *junction data* precedence. I'll again start with a gentle example, switching to functions (methods) to avoid confusion with *operator* precedence: ``` say (1,2).all le (1, 2).one; # all(one(True, True), one(False, True)) say so (1,2).all le (1, 2).one; # False say (1,2).one le (1, 2).all; # all(one(True, True), one(False, True)) say so (1,2).one le (1, 2).all; # False ``` This indicates that `one` junctions have higher "precedence" than `all`. I'm going to stop at this point because I got confused about the exact "precedence" relationship between the four sorts of junctions and felt this was potentially a good launch point for others. If no one else does look into this I may return to this another day but no promises. -- love, raiph
