on using cl-do-symbols

on using cl-do-symbols

[Mark Oteiza]

For yuks I ported auth-source to cl-lib.  I encountered a warning I
don't quite understand when changing

  (loop for sym being the symbols of password-data
        ;; when the symbol name starts with auth-source-magic
        when (string-match (concat "^" auth-source-magic)
                           (symbol-name sym))
        ;; remove that key
        do (password-cache-remove (symbol-name sym)))

to

  (cl-do-symbols (sym password-data)
    (when (string-match (concat "^" auth-source-magic) (symbol-name sym))
      (password-cache-remove (symbol-name sym))))

The warning is that there is an unused binding for `sym'.  cl-do-symbols
does create an outer (let (sym) …) and then inside mapatoms, another
(lambda (sym) …). Does the outer let-binding in cl-do-symbols need to be there?

Re: on using cl-do-symbols

[Stefan Monnier]

> The warning is that there is an unused binding for `sym'.
> cl-do-symbols does create an outer (let (sym) …) and then inside
> mapatoms, another (lambda (sym) …).  Does the outer let-binding in
> cl-do-symbols need to be there?

I guess it is there so as to obey the CL definition which says:

    When result-form is evaluated, var is bound and has the value nil.

If that's the case, then the let-binding can be moved so it only
covers the `,(nth 2 spec)` part of the code.  Personally, I'd be happy
to disobey the CL specification in this regard since I find it
completely silly.


        Stefan

Re: on using cl-do-symbols

[Mark Oteiza]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
>> The warning is that there is an unused binding for `sym'.
>> cl-do-symbols does create an outer (let (sym) …) and then inside
>> mapatoms, another (lambda (sym) …).  Does the outer let-binding in
>> cl-do-symbols need to be there?
>
> I guess it is there so as to obey the CL definition which says:
>
>     When result-form is evaluated, var is bound and has the value nil.
>
> If that's the case, then the let-binding can be moved so it only
> covers the `,(nth 2 spec)` part of the code.  Personally, I'd be happy
> to disobey the CL specification in this regard since I find it
> completely silly.

I agree--I can't think of a reason why var should be part of
result-form.  Interestingly, in the dynamic binding part of dolist, var
is indeed set to nil.

Re: on using cl-do-symbols

[Stefan Monnier]

> Interestingly, in the dynamic binding part of dolist, var is indeed
> set to nil.

Yes, that's for backward compatibility (the lexical part doesn't need
to care about that backward compatibility ;-).


        Stefan