at sign as symbol name

at sign as symbol name

[Seth Alves]

In r7rs-small, I see this:

     All implementations of Scheme must support the following
     extended identifier characters:
     ! $ % & * + - . / : < = > ? @ ^ _ ~

and other schemes seem okay with it.  Should kawa accept '@ as a 
symbol?  I notice that

     (string->symbol "@")

works.

     -seth

Re: at sign as symbol name

[Per Bothner]

On 12/28/2014 01:14 PM, Seth Alves wrote:
> In r7rs-small, I see this:
>
>      All implementations of Scheme must support the following
>      extended identifier characters:
>      ! $ % & * + - . / : < = > ? @ ^ _ ~
>
> and other schemes seem okay with it.  Should kawa accept '@ as a symbol?

It does if you specify --r7rs on the command line.

The problem is that @ is used for the splice syntax, which I think is is
more useful than identifiers that start with @.

We could hack '@ so it works, but that doesn't help for unquoted
identifiers.  We could also do something similar that Kawa does for colon:
If @ is followed by a possibly-empty identifier NAME (i.e. @NAME)
we could disambiguate depending on syntactic context (e.g. in define)
and which of NAME or @NAME has an in-scope lexical binding.  That
would be somewhat hackish and not 100% compatible - just like the
colon handling.  I figure if you really want an identifier that
starts with a @ use the --r7rs flag.

> I notice that
>
>      (string->symbol "@")
>
> works.

Right, but note that (write (string->symbol "@"))
writes "|@|" (with escapes).

(However, (write (string->symbol ":")) just
writes ":" which is a bug ...)
-- 
	--Per Bothner
per@bothner.com   http://per.bothner.com/

Re: at sign as symbol name

[Seth Alves]

Oh, okay.  That's fine -- I keep forgetting to include --r7rs.

     -seth

On 12/28/2014 02:29 PM, Per Bothner wrote:
>
>
> On 12/28/2014 01:14 PM, Seth Alves wrote:
>> In r7rs-small, I see this:
>>
>>      All implementations of Scheme must support the following
>>      extended identifier characters:
>>      ! $ % & * + - . / : < = > ? @ ^ _ ~
>>
>> and other schemes seem okay with it.  Should kawa accept '@ as a symbol?
>
> It does if you specify --r7rs on the command line.
>
> The problem is that @ is used for the splice syntax, which I think is is
> more useful than identifiers that start with @.
>
> We could hack '@ so it works, but that doesn't help for unquoted
> identifiers.  We could also do something similar that Kawa does for 
> colon:
> If @ is followed by a possibly-empty identifier NAME (i.e. @NAME)
> we could disambiguate depending on syntactic context (e.g. in define)
> and which of NAME or @NAME has an in-scope lexical binding.  That
> would be somewhat hackish and not 100% compatible - just like the
> colon handling.  I figure if you really want an identifier that
> starts with a @ use the --r7rs flag.
>
>> I notice that
>>
>>      (string->symbol "@")
>>
>> works.
>
> Right, but note that (write (string->symbol "@"))
> writes "|@|" (with escapes).
>
> (However, (write (string->symbol ":")) just
> writes ":" which is a bug ...)