Is #S(quux ...) a literal?

[garethw]

Hi all,

Given some structure, say:

Code: Select all

(defstruct quux
  name)

is the following a "literal"?

Code: Select all

#S(quux :name "Guy")

I ask because I was playing with OPTIMA, and wanted to use FARE-QUASIQUOTE for specifying patterns with backquotes, but FARE-QUASIQUOTE is not pleased:

Code: Select all

`(#S(quux :name "Guy"))
=>
"unrecognized object in quasiquote"

and in poking around, the expansion code includes a call to FARE-UTILS:LITERALP, which is:

Code: Select all

(defun literalp (x)
  "predicate that tells whether X is the source form for a literal expression."
  (typep x '(or boolean number character array keyword)))

Structures are conspicuously missing, and that surprised me. SBCL's new(ish) backquoting does what I expect. Is this a grey area? Not something one should rely on?

[Goheeca]

Can you describe what SBCL is doing? I have right now installed pretty old version 1.1.12.
Nevertheless, sharpsign syntax denotes a reader macro and that means it is by a standard way evaluated before the quasiquotation takes a place or in other words the quasiquotation doesn't stop evaluating reader macros. Does that FARE-QUASIQUOTE works with complex numbers?

[pjstirling]

Yes, for compilation purposes #s(quux :name "Guy") creates a literal object, complain to fare

[garethw]

Can you describe what SBCL is doing? I have right now installed pretty old version 1.1.12.

It returns =>

Code: Select all

(#S(QUUX :NAME "Guy"))

Nevertheless, sharpsign syntax denotes a reader macro and that means it is by a standard way evaluated before the quasiquotation takes a place or in other words the quasiquotation doesn't stop evaluating reader macros. Does that FARE-QUASIQUOTE works with complex numbers?

Understood. The problem is not in the reader, it's in Fare's expander, which relies on that function to determine if something is a literal, concludes it is not, and falls through the rest of the COND that calls it to that error condition.

F-Q handles complex numbers gracefully as expected:

Code: Select all

`(#C(1 2))
=>
(#C(1 2))

Yes, for compilation purposes #s(quux :name "Guy") creates a literal object, complain to fare

Done, for (hopefully) humble values of COMPLAIN. Thanks for the confidence boost to do so.

[garethw]

FWIW, Fare agreed this is likely a bug.

Thanks for your help, Goheeca and pjstirling.

[marcoxa]

You should play with CL-UNIFICATION (yep; this is a shameless plug).

Cheers
--
MA