Are arrow macros valid places?

[phoe]

There is a question whether a call to e.g. -> or -<> is a valid place and therefore passable to setf.

What do you think about this? It should be possible to define proper setf expanders for those in order to deal with the fact that the PR from #3 turns them into let* forms which are not valid places.

[phoe]

The current implementation of arrows already makes it possible to use arrow forms as places with the exception of diamond macros with multiple diamonds (which is an extension, anyway). It might be possible to get this last one to work using a custom setf expansion; hence I'll keep this issue open.

[Harleqin]

I must admit that I never thought about this. I struggle a bit with coming up with use cases in the cases where the expansion has let bindings and/or conditionals, though.

[phoe]

Conditionals make it possible to set different places. As for named bindings, likely the same use cases as non-setf cases apply.

[phoe]

Example for conditionals:

BINDING-ARROWS> (let ((cons (cons (cons 1 2) 3)))
                  (setf (cond-> cons
                          (t car)
                          (t cdr))
                        42)
                  cons)
((1 . 42) . 3)

BINDING-ARROWS> (let ((cons (cons (cons 1 2) 3)))
                  (setf (cond-> cons
                          (nil car)
                          (t cdr))
                        42)
                  cons)
((1 . 2) . 42)

BINDING-ARROWS> (let ((cons (cons (cons 1 2) 3)))
                  (setf (cond-> cons
                          (t car)
                          (nil cdr))
                        42)
                  cons)
(42 . 3)

BINDING-ARROWS> (let ((cons (cons (cons 1 2) 3)))
                  (setf (cond-> cons
                          (nil car)
                          (nil cdr))
                        42)
                  cons)
42