Re: Code for cond*

[JD Smith <jdtsmith@gmail.com>]

> On Jan 25, 2024, at 8:57 AM, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
> 
>>> [1] I regularly convince myself that it’s such low hanging fruit,
>>> there must in fact already BE a cond-let, and I go hunting for it.
>>> The obvious interface seems like such a straightforward extension of
>>> if/when-let, that there would be absolutely nothing new to learn:
>>> 
>>> (cond-let
>>> (((var value)
>>>   (dvar (derived-from var))
>>>   ((has-the-right-stuff-p dvar)))
>>>  (cons 'correct dvar))
>>> 
>>> (((foo value2)
>>>   (bar (1- foo))
>>>   ((< bar 0)))
>>>  (cons 'incorrect bar))
>>> 
>>> (t nil))
> 
> Nice.  Would `var` and `dvar` scope over just the first branch or also the
> other ones?
> Technically, there is no reason for them not to be visible in the other
> branches, so that's what I would go for.

It seems most intuitive to me for each clause to have its own scope, but mostly because active bindings made in sibling forms at the same depth seem non-intuitive and easy to overlook.

> Personally, I was thinking of a syntax like
> 
>    (my-cond
>     (:let var value)
>     (:let dvar (derived-from var))
>     ((has-the-right-stuff-p dvar)
>      (cons 'correct dvar))
> 
>     (:let foo value2)
>     (:let bar (1- foo))
>     ((< bar 0)
>      (cons 'incorrect bar))
> 
>     (t nil))
> 
> with the intention of replacing `cond` without introducing
> any incompatibility.

This is an interesting approach, and makes it much clearer when new bindings appear.  Mine is a different idea.  In my cond-let approach, just as for when-let, each binding in a clause’s binding group is tested, one by one, and the clause only succeeds if all the bindings are non-nil.  So var and dvar must be non-nil, and (has-the-right-stuff-p dvar) too.  To ensure this, in your version, I’d need something like:

(:let var value)
(:let dvar (and var (derived-from var))
((and var dvar (has-the-right-stuff-p dvar)) …)

This would get repetitive and error prone for a large number of bindings, not to mention awkward if some “bindings” in the middle are actually bound-variable-free tests like ((has-the-right-stuff-p dvar)).

The hypothetical cond-let makes it easy to build an arbitrary grouping of sequential when's/when-let's which short circuit as soon as the first clause body executes, no matter what that body returns.

So:

(cond-let
 (simple0 body0) ; a normal cond-style clause
 (bindings1 body1)
 (bindings2 body2)
 (bindings3 body3)
 (t fallthrough-body))

would be equivalent to:

(catch 'cond-let
  (when simple0 (throw 'cond-let body0))
  (when-let (bindings1) (throw 'cond-let body1))
  (when-let (bindings2) (throw 'cond-let body2))
  (when-let (bindings3) (throw 'cond-let body3))
  (throw 'cond-let fallthrough-body))

But a lot less wordy/repetitive ;).