Re: Code for cond*

[Stefan Monnier <monnier@iro.umontreal.ca>]

> Maybe you are right, but I don't think I have seen an example of it.
> Hmm, was this the purpose of proposing the cl-lambda pattern operator?

No, see:

    (defmacro byte-optimize--pcase (exp &rest cases)
      ;; When we do
      ;;
      ;;     (pcase EXP
      ;;       (`(if ,exp ,then ,else) (DO-TEST))
      ;;       (`(plus ,e2 ,e2)        (DO-ADD))
      ;;       (`(times ,e2 ,e2)       (DO-MULT))
      ;;       ...)
      ;;
      ;; we usually don't want to fall back to the default case if
      ;; the value of EXP is of a form like `(if E1 E2)' or `(plus E1)'
      ;; or `(times E1 E2 E3)', instead we either want to signal an error
      ;; that EXP has an unexpected shape, or we want to carry on as if
      ;; it had the right shape (ignore the extra data and pretend the missing
      ;; data is nil) because it should simply never happen.
      ;;
      ;; The macro below implements the second option by rewriting patterns
      ;; like `(if ,exp ,then ,else)'
      ;; to   `(if . (or `(,exp ,then ,else) pcase--dontcare))'.
      ;;
      ;; The resulting macroexpansion is also significantly cleaner/smaller/faster.
      [...])

in `byte-opt.el`.

> Basically, I think allowing a shorter list is something that I think
> would be desirable for one level of list structure whereas `cdr-safe'
> seemed to make sense for nested levels.

I can't think of places where it would make sense to apply something
like `cdr-safe` to nested lists, so it's hard for me to judge.

>   > - As you say for "one specific list, it is hard to find anything simpler
>   >   than `. ,_'", yet you added `cdr-safe` which applies to a single
>   >   pattern, and thus typically to a single list, unless the pattern is
>   >   a big (or ...) pattern which are rather unusual.
> I can't follow that.  Why do you say "typically to a single list"?  I
> don't think that is true -- it applies to all sublists of the pattern
> used within it.

Most patterns have a single list.  As you see in `byte-optimize--pcase`,
I need that kind of functionality for most/all branches.


        Stefan