Re: macros and macroexpand

[Philip Kaludercic]

Heime <heimeborgia@protonmail.com> writes:

> Sent with Proton Mail secure email.
>
> ------- Original Message -------
> On Monday, August 7th, 2023 at 11:46 PM, Yuri Khan <yuri.v.khan@gmail.com> 
> wrote:
>
>
>> On Mon, 7 Aug 2023 at 18:04, Heime heimeborgia@protonmail.com wrote:
>> 
>> > I have made a macro and know that they are supposed to return
>> > expanded code for use. Still I cannot understand the need to
>> > call "macroexpand". Should't the macro already perform the
>> > expansion ?
>> 
>> 
>> You should be posting small examples of code that you’re trying,
>> otherwise, there is high chance people will either misunderstand you
>> or just disregard your questions as ill-posed.
>> 
>> ----
>> 
>> When you define a macro, you indeed write the definition similarly to
>> a function that returns expanded code.
>> 
>> (defmacro foo (&rest body)
>> `(bar ,@body)) When you evaluate a form that references a macro,
>> Elisp will (1) expand the macro, and (2) evaluate the result of the
>> expansion: (foo 'quux) ⇒ Debugger entered--Lisp error:
>> (void-function bar) On the other hand, calling ‘macroexpand’ on a
>> data representation of that form will just return the expansion
>> result: (macroexpand '(foo 'quux)) ⇒ (bar 'quux) In this example, I
>> did not bother to define ‘bar’, so Elisp assumes it would be a
>> function and complains at evaluation time. But I could further
>> define ‘bar’ as a macro: (defmacro bar (&rest body)` (baz ,@body))
>> 
>> In this case, evaluating the original form shows that Elisp expanded
>> both macros ‘foo’ and ‘bar’, and then tried to call the undefined
>> function ‘baz’:
>> 
>> (foo 'quux)
>> ⇒ Debugger entered--Lisp error: (void-function baz)
>> 
>> Meanwhile, ‘macroexpand’ still just expands a single level of macros:
>> 
>> (macroexpand '(foo 'quux))
>> ⇒ (bar 'quux)
>> 
>> and you can invoke it repeatedly until you get to the fixed point:
>> 
>> (macroexpand (macroexpand '(foo 'quux)))
>> ⇒ (baz 'quux)
>> 
>> (macroexpand (macroexpand (macroexpand '(foo 'quux))))
>> ⇒ (baz 'quux)
>
> Then macroexpand is useful for diagnostics to expand at one level only
> at a time.  Thusly, if I just want to get the expanded code produced by 
> a macro, I can just do pp-to-string upon the object made by a macro.
>
> (defmacro adder (mopi mopj)
>   `(+ ,(cl-second mopi) ,(cl-third mopj)))
>
> (princ (pp-to-string '(adder (* 3 5) (* 5 7)) ))
                       ^
                       don't do this

If you quote an expression, it won't be evaluated or macro-expanded any
further.  You can sort-of think of a macro like a kind of inline
function call.  The evaluation would go along these lines:

(princ (pp-to-string (adder (* 3 5) (* 5 7))))

will be transformed into this at macro-expansion time, and evaluation
would do this:

(princ (pp-to-string (+ (cl-second '(* 3 5)) (cl-third '(* 5 7)))))
(princ (pp-to-string (+ 3 7)))
(princ (pp-to-string 10))
(princ "10\n")
"10\n"

> I would not do
>
> (princ (pp-to-string (macroexpand '(adder (* 3 5) (* 5 7))) ))