should fset 'symbol get an undefined function warning?

should fset 'symbol get an undefined function warning?

[Dan Nicolaescu]

The warning "the function `my-fset-test' is not known to be defined"
is generated when compiling:

(fset 'my-fset-test 'foo)
(my-fset-test)

Wouldn't it be better if the warning was not generated?

Re: should fset 'symbol get an undefined function warning?

[Stefan Monnier]

> The warning "the function `my-fset-test' is not known to be defined"
> is generated when compiling:

> (fset 'my-fset-test 'foo)
> (my-fset-test)

> Wouldn't it be better if the warning was not generated?

Could be.  But if you use `defalias' the warning will be gone.  Why do
you want to use `fset'?


        Stefan

Re: should fset 'symbol get an undefined function warning?

[Dan Nicolaescu]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

  > > The warning "the function `my-fset-test' is not known to be defined"
  > > is generated when compiling:
  > 
  > > (fset 'my-fset-test 'foo)
  > > (my-fset-test)
  > 
  > > Wouldn't it be better if the warning was not generated?
  > 
  > Could be.  But if you use `defalias' the warning will be gone.  Why do
  > you want to use `fset'?

"I" don't, but there are a some files in the tree that do that, one
that I remember of the top of my head is viper-util. But there are
some more.

RE: should fset 'symbol get an undefined function warning?

[Drew Adams]

>   > > The warning "the function `my-fset-test' is not known to be defined"
>   > > is generated when compiling:
>   > > (fset 'my-fset-test 'foo) (my-fset-test)
>   > > Wouldn't it be better if the warning was not generated?
>   >
>   > Could be.  But if you use `defalias' the warning will be gone.  Why do
>   > you want to use `fset'?
>
> "I" don't, but there are a some files in the tree that do that, one
> that I remember of the top of my head is viper-util. But there are
> some more.

AFAIK, `fset' is not deprecated or obsolete (and the Elisp manual is full of
`fset').

`defalias' might be better to use in this context, but it should still be
good to not issue that warning for `fset'. (I'm not saying anyone has
disagreed with that.)

Re: should fset 'symbol get an undefined function warning?

[Stefan Monnier]

> AFAIK, `fset' is not deprecated or obsolete (and the Elisp manual is full of
> `fset').

> `defalias' might be better to use in this context, but it should still be
> good to not issue that warning for `fset'. (I'm not saying anyone has
> disagreed with that.)

I disagree: `fset' is meant to *change* the function-slot
of a symbol, not to initialize it.  So it's use shouldn't be treated by
the byte-compiler as a declaration.  Insted, the byute-compiler should
complain about the use of `fset' on an unknown function similarly to the
use of `set(q)' on an unknown variable.


        Stefan

RE: should fset 'symbol get an undefined function warning?

[Drew Adams]

> > `defalias' might be better to use in this context, but it
> > should still be good to not issue that warning for `fset'.
> > (I'm not saying anyone has disagreed with that.)
>
> I disagree: `fset' is meant to *change* the function-slot
> of a symbol, not to initialize it.  So it's use shouldn't be treated by
> the byte-compiler as a declaration.  Insted, the byute-compiler should
> complain about the use of `fset' on an unknown function similarly to the
> use of `set(q)' on an unknown variable.

Good point. I'm convinced.

Was there an opposing consideration that led you to say "Could be" at first,
or was that before this point occurred to you?

Re: should fset 'symbol get an undefined function warning?

[Richard Stallman]

    The warning "the function `my-fset-test' is not known to be defined"
    is generated when compiling:

    (fset 'my-fset-test 'foo)
    (my-fset-test)

If you use `fset' it means that you are playing around with the
symbol's definition.  So it seems like a mistake to assume anything
about it.  Normally the first arg of `fset' is not a quoted constant
anyway.

If you want to make an alias that the compiler will know about, use
`defalias'.

Re: should fset 'symbol get an undefined function warning?

[Stefan Monnier]

>> > `defalias' might be better to use in this context, but it
>> > should still be good to not issue that warning for `fset'.
>> > (I'm not saying anyone has disagreed with that.)
>> 
>> I disagree: `fset' is meant to *change* the function-slot
>> of a symbol, not to initialize it.  So it's use shouldn't be treated by
>> the byte-compiler as a declaration.  Insted, the byute-compiler should
>> complain about the use of `fset' on an unknown function similarly to the
>> use of `set(q)' on an unknown variable.

> Good point. I'm convinced.

> Was there an opposing consideration that led you to say "Could be" at first,
> or was that before this point occurred to you?

No.  My area of expertise is functional programming where there's a *very*
clear difference between changing the value of a variable and initializing
that variable (e.g. in SML and Haskell you simply can't do the equivalent
of `setq' or `fset').


        Stefan