Re: Syntax object representation

[Taylan Ulrich Bayirli/Kammer]

Dmitry Bogatov <address@hidden> writes:

> Hello.
>
> How stable the fact, that
>     (equal? #'(foo bar) (list #'foo #'bar))
> ?

Not stable.  A syntax object for a list can also wrap the whole list, it
is not necessarily a list of syntax objects.

E.g. the following is a valid syntax object representing '(+ 1 2)':

#(syntax-object (+ 1 2) ((top)) (hygiene guile-user))

> Why? Combining two syntax objects in list can be done via
> #` and #, but it would be nice to know, than I am free
> to use all power of list functions.

A syntax object like the above example for '(+ 1 2)' could contain
source-location or other meta-data.  If we guarantee that all non-atomic
syntax objects are lists, we lose that ability.  It's better that they
are "opaque" and that 'syntax-case' or other special operators are the
only way to "destructure" them.

> Also, it seems there is no syntax-object? predicate. Is it by design?

Guile uses psyntax, so syntax objects are actually just vectors! :)

You can enter the following in the Guile REPL and see what happens:

#(syntax-object (+ 1 2) ((top)) (hygiene guile-user))

Note that it's not the same as:

'#(syntax-object (+ 1 2) ((top)) (hygiene guile-user))

(This one is quoted, so it's just a literal vector.)

> It seems, than syntax->datum is no-op on non-syntax objects.
> Can I rely on it?

I think that's also an implementation detail of psyntax.  I wouldn't
rely on it, but I'm not sure on this one.

> Thanks in advance.

Happy to help!

> --
> Best regards, Dmitry Bogatov <address@hidden>,
> Free Software supporter, esperantisto and netiquette guardian.
>       git://kaction.name/rc-files.git
>       GPG: 54B7F00D

Taylan