Quoted function in `define-key'

Quoted function in `define-key'

[Narendra Joshi]

The following piece of code is confusing me. It is from `rinari-minor-mode'.

(dolist (el (append (mapcar
                     (lambda (el)
                       (cons (concat "f" (second el))
                             (read (format "'rinari-find-%S"
                                           (first el)))))
                     rinari-jump-schema)
                    rinari-minor-mode-keybindings))
  (eval `(define-key rinari-prefix-map ,(car el) ,(cdr el))))


If I replace the last expression with:

(define-key rinari-prefix-map (car el) (cdr el))

why does it fail?

(cdr el) is (quote function-name) which is exactly the same as
'function-name.

--
Narendra Joshi

Re: Quoted function in `define-key'

[Narendra Joshi]

Narendra Joshi <narendraj9@gmail.com> writes:

> The following piece of code is confusing me. It is from `rinari-minor-mode'.
>
> (dolist (el (append (mapcar
>                      (lambda (el)
>                        (cons (concat "f" (second el))
>                              (read (format "'rinari-find-%S"
>                                            (first el)))))
>                      rinari-jump-schema)
>                     rinari-minor-mode-keybindings))
>   (eval `(define-key rinari-prefix-map ,(car el) ,(cdr el))))
I have also tried changing this ^ to
(let ((key (car el))
      (func (cdr el)))
  (define-key rinari-prefix-map key func))

I thought maybe `define-key` is a macro and hence the calls to `car` and
`cdr` might not be evaluated. But that doesn't seem to be the case. This
still fails. The key is bound to a value that is equal to (quote some-function-name).

-- 
Narendra Joshi

Re: Quoted function in `define-key'

[Michael Heerdegen]

Narendra Joshi <narendraj9@gmail.com> writes:

> The following piece of code is confusing me. It is from `rinari-minor-mode'.
>
> (dolist (el (append (mapcar
>                      (lambda (el)
>                        (cons (concat "f" (second el))
>                              (read (format "'rinari-find-%S"
>                                            (first el)))))
>                      rinari-jump-schema)
>                     rinari-minor-mode-keybindings))
>   (eval `(define-key rinari-prefix-map ,(car el) ,(cdr el))))
>
>
> If I replace the last expression with:
>
> (define-key rinari-prefix-map (car el) (cdr el))
>
> why does it fail?
>
> (cdr el) is (quote function-name) which is exactly the same as
> 'function-name.

Sure, but this is then not evaluated again, and you want the DEF
argument of `define-key' to be function-name, not (quote function-name).

And yes, using `eval' in this context is indeed overkill (as well as
`read' which could be replaced with `intern').


Michael.

Re: Quoted function in `define-key'

[Narendra Joshi]

Michael Heerdegen <michael_heerdegen@web.de> writes:


> Sure, but this is then not evaluated again, and you want the DEF
> argument of `define-key' to be function-name, not (quote function-name).
Okay. Let me re-state this whole thing so that I can assure myself that
I have understood.

   (quote my-function) is exactly the same as 'my-function, i.e.
   (eq (quote my-function) 'my-function) evaluates to t.

   But here we are talking about the result of evaluating

   (quote my-function) which isn't a cons but a symbol. But if we have
   `(quote my-function) that isn't the same as 'my-function. 

   Side note: If anybody knows how to type inline code in gnus message,
   it would be greatly appreciated.

--
Narendra Joshi

Re: Quoted function in `define-key'

[Michael Heerdegen]

Narendra Joshi <narendraj9@gmail.com> writes:

>    (quote my-function) is exactly the same as 'my-function, i.e.

Yes, the second is an alternative read syntax for the first expression.

>    (eq (quote my-function) 'my-function) evaluates to t.
>
>    But here we are talking about the result of evaluating
>
>    (quote my-function) which isn't a cons but a symbol.

No, it's a list (and a cons), but the result of evaluation is a symbol.

>     But if we have `(quote my-function) that isn't the same as
>    'my-function.

`(quote my-function) and '(quote my-function) eval to the list
(quote my-function).  ''my-function is an alternative read syntax for
'(quote my-function).  But that all doesn't appear in the example.

It's so: you want to specify a symbol as third argument to `define-key'.
`define-key' is a function, so the argument positions are evaluated.
Thus you want to specify an expression that evaluates to the symbol you
want.  (quote my-function) or 'my-function evaluate to
the symbol you want.  In

   (define-key rinari-prefix-map (car el) (cdr el))

the expression at that position is (cdr el), so that expression should
eval to a symbol, not to a list like (quote symbol) - even when this
list would give you what you want when it would be evaluated.
Evaluation happens only once.

With other words: when you write something like

  (define-key my-map keys 'my-function)

you use the quote to prevent my-function from being evaluated, because
you want to specify that symbol (unlike its binding as a variable), but
_not_ because define-key would expect something quoted.

BTW, most people prefer

  (define-key my-map keys #'my-function)

however.  #'my-function is a read syntax for (function my-function),
where function is like quote but tells the byte compiler that the quoted
thing is a function expression.

>    Side note: If anybody knows how to type inline code in gnus message,
>    it would be greatly appreciated.

I use `message-mark-inserted-region' for multiline code, but I don't
know of any rule for smaller snippets.  Most of the time I don't care
too much (like here).  It's good style to quote a `symbol' like this,
but I sometimes avoid it when code snippets already involve lots of
quoting.


HTH,

Michael.

Re: Quoted function in `define-key'

[Stefan Monnier]

>>>>> "Narendra" == Narendra Joshi <narendraj9@gmail.com> writes:

> The following piece of code is confusing me. It is from `rinari-minor-mode'.

> (dolist (el (append (mapcar
>                      (lambda (el)
>                        (cons (concat "f" (second el))
>                              (read (format "'rinari-find-%S"
>                                            (first el)))))
>                      rinari-jump-schema)
>                     rinari-minor-mode-keybindings))
>   (eval `(define-key rinari-prefix-map ,(car el) ,(cdr el))))


> If I replace the last expression with:

> (define-key rinari-prefix-map (car el) (cdr el))

> why does it fail?

> (cdr el) is (quote function-name) which is exactly the same as
> 'function-name.

"Exactly the same"?  Well, clearly not, since it behaves differently:

    (eval (cdr el))         =>   (quote function-name)
    (eval 'function-name)   =>   function-name

A good first rule is that if you're using `eval` it means you're not really
doing it right:

    (dolist (el (append (mapcar
                         (lambda (el)
                           (cons (vconcat [?\f] (nth 1 el))
                                 (intern "rinari-find-%s" (nth 0 el))))
                         rinari-jump-schema)
                        rinari-minor-mode-keybindings))
      (define-key rinari-prefix-map (car el) (cdr el)))


-- Stefan

Re: Quoted function in `define-key'

[Narendra Joshi]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> Narendra Joshi <narendraj9@gmail.com> writes:
>
>>    (quote my-function) is exactly the same as 'my-function, i.e.
>
> Yes, the second is an alternative read syntax for the first expression.
>
>>    (eq (quote my-function) 'my-function) evaluates to t.
>>
>>    But here we are talking about the result of evaluating
>>
>>    (quote my-function) which isn't a cons but a symbol.
>
> No, it's a list (and a cons), but the result of evaluation is a symbol.
>
>>     But if we have `(quote my-function) that isn't the same as
>>    'my-function.
>
> `(quote my-function) and '(quote my-function) eval to the list
> (quote my-function).  ''my-function is an alternative read syntax for
> '(quote my-function).  But that all doesn't appear in the example.
>
> It's so: you want to specify a symbol as third argument to `define-key'.
> `define-key' is a function, so the argument positions are evaluated.
> Thus you want to specify an expression that evaluates to the symbol you
> want.  (quote my-function) or 'my-function evaluate to
> the symbol you want.  In
>
>    (define-key rinari-prefix-map (car el) (cdr el))
>
> the expression at that position is (cdr el), so that expression should
> eval to a symbol, not to a list like (quote symbol) - even when this
> list would give you what you want when it would be evaluated.
> Evaluation happens only once.
>
> With other words: when you write something like
>
>   (define-key my-map keys 'my-function)
This was really helpful. Thanks! :)

>
> you use the quote to prevent my-function from being evaluated, because
> you want to specify that symbol (unlike its binding as a variable), but
> _not_ because define-key would expect something quoted.
>
> BTW, most people prefer
>
>   (define-key my-map keys #'my-function)
> however.  #'my-function is a read syntax for (function my-function),
> where function is like quote but tells the byte compiler that the quoted
> thing is a function expression.

So, the symbol `my-function' can have the function definition in its
value slot? What does the byte compiler do with this information? I am
just curious about this. If this is recommended, I would also start
quoting my functions as #'my-function. 


>>    Side note: If anybody knows how to type inline code in gnus message,
>>    it would be greatly appreciated.
>
> I use `message-mark-inserted-region' for multiline code, but I don't
> know of any rule for smaller snippets.  Most of the time I don't care
> too much (like here).  It's good style to quote a `symbol' like this,
> but I sometimes avoid it when code snippets already involve lots of
> quoting.
Thanks! `message-mark-inserted-region' is good but I would probably change
the boundary that it inserts. :) 

-- 
Narendra Joshi

Re: Quoted function in `define-key'

[Michael Heerdegen]

Narendra Joshi <narendraj9@gmail.com> writes:

> So, the symbol `my-function' can have the function definition in its
> value slot?

There's a separate slot for function bindings.  You can refer to it
directly with passing the symbol to `symbol-function'.  So any symbol
can have two separate "bindings" - a binding as a variable, and one as a
function (unlike scheme, for example).  That's why Elisp is a Lisp-2.
Actually, there is one more cell for symbol property lists ("plists").
See

  (info "(elisp) Symbol Components")

for more details.


> What does the byte compiler do with this information? I am just
> curious about this.

When you sharp quote a lambda expression, the compiler knows that this
lambda list is supposed to be called as a function, and will byte
compile it.  The evaluator might turn it into a closure when lexical
binding is used.

In elisp, the `lambda' macro comes with implicit function quoting (see
its definition), so you can just omit it.

But OTOH, it's an error to `quote' lambdas like

  '(lambda () body...)

because it will prevent byte compilation or closure creation and make
Stefan angry.

Obviously for `function' quoted symbols there is not much to do, but the
byte compiler can determine whether that function is defined/ will be
defined at run time, so it can warn you about typos as an extra service.

For example, you get a compiler warning when compiling

  (define-key my-map [key] #'make-frame-cmomand)

but not for

  (define-key my-map [key] 'make-frame-cmomand)

> If this is recommended, I would also start quoting my functions as
> #'my-function.

Yes, it's a good idea.


Michael.

Re: Quoted function in `define-key'

[Narendra Joshi]

Michael Heerdegen <michael_heerdegen@web.de> writes:

>   (info "(elisp) Symbol Components")
How did you insert this link to info?

>
>
>> What does the byte compiler do with this information? I am just
>> curious about this.
>
> When you sharp quote a lambda expression, the compiler knows that this
> lambda list is supposed to be called as a function, and will byte
> compile it.  The evaluator might turn it into a closure when lexical
> binding is used.
>
> In elisp, the `lambda' macro comes with implicit function quoting (see
> its definition), so you can just omit it.
>
> But OTOH, it's an error to `quote' lambdas like
>
>   '(lambda () body...)
>
> because it will prevent byte compilation or closure creation and make
> Stefan angry.
:)


> Obviously for `function' quoted symbols there is not much to do, but the
> byte compiler can determine whether that function is defined/ will be
> defined at run time, so it can warn you about typos as an extra service.
>
> For example, you get a compiler warning when compiling
>
>   (define-key my-map [key] #'make-frame-cmomand)
>
> but not for
>
>   (define-key my-map [key] 'make-frame-cmomand)
>
>> If this is recommended, I would also start quoting my functions as
>> #'my-function.
>
> Yes, it's a good idea.
Thanks! This is very informative.


-- 
Narendra Joshi

Re: Quoted function in `define-key'

[Michael Heerdegen]

Narendra Joshi <narendraj9@gmail.com> writes:

> >   (info "(elisp) Symbol Components")
> How did you insert this link to info?

Simply M-0 w from *info* (i.e. just w with a zero prefix arg) and you
have a call to visit the current info page in the kill ring.


Michael.