Bug in diamond wand equality

[dtenny]

Hi there. Just a note that there is a small problem if you want to diamond wand substitutions without :USE-ing the arrow-macros package. For example

(ql:quickload :arrow-macros)
(arrow-macros:-<>> 1 (print <>))

Yields

The variable <> is unbound.
   [Condition of type UNBOUND-VARIABLE]

This is because has-diamond (and similar checks) are using (eq exp '<>) which fails if the symbol was interned in a different package than the :arrow-macros package.

Instead you probably need a check like this:

(and (symbolp exp)
     (string= exp "<>"))

[dtenny]

I suppose whether this is a bug is debatable, as I could package-qualify the <> symbol, e.g.
(arrow-macros:-<>> 3 (list arrow-macros:<> 1 2))

But that's kind of a drag in my book and it wasn't at all what I was expecting to to fix the unbound variable error problem.

[hipeta]

Thanks reporting.
I'm not native English speaker so some words or text may be wrong.

I suppose it's a specification issue.
If we use string= for checking <> symbol equality, it absorbs the package differences which the symbols belong to, but a:<> and b:<> are different things in common lisp.
It makes diamond-wands be "too strong" macro because it robs <> symbol semantics implicitly in all other packages.
I suppose it is unhealthy although <> symbol confliction rarely happens.

For example:

(ql:quickload :arrow-macros)

(defpackage a
  (:use :cl :arrow-macros)
  (:export :test))
(in-package :a)
(defmacro test (x)
  `(-<> 1 (+ 2 <>) (+ 3 <>) (+ ,x <>)))

(defpackage b
  (:use :cl :a))
(in-package :b)
(test (let ((<> 4)) <>))  

Using string= code, this causes an syntax error.
If package a and b are written by other persons, it will be bad behavior.
Therefore, I recommend using (import 'arrow-macros:<>), or including :use or :local-nicknames in your package definition.