Toplevel closures in Medley Common Lisp

[pamoroso]

Consider this typical (at least in textbooks) idiom for creating closures in Common Lisp (there are better uses of closures but bear with me):

(LET ((MESSAGE "Hello world")) (DEFUN HELLO () (PRINT MESSAGE)))

When I evaluate this expression at the top level of a XCL Exec I get an error such as:

Invalid argument: #<Interpreted closure @ 166,117750>

Does Medley support toplevel closures like this? Does creating these closures require a different syntax?

[wmannis]

I also ran into this during some casual playing around with XCL recently.

Lexical closures first appeared in scheme, but I forget if they existed in CLTL1-era Common Lisp, or if they came later. I have a vague memory they came later to CL, but I can't find a cite for that.

[pamoroso]

This alternate definition works as expected:

2/5> (DEFVAR HELLO (LET ((MESSAGE "Hello world")) #'(LAMBDA () (PRINT MESSAGE))))
HELLO
2/7> HELLO
#<Interpreted closure @ 166,120750>
2/8> (FUNCALL HELLO)

"Hello world" 
"Hello world"

[masinter]

THe problem with DEFUNs inside lexical scope and not at top level is there is no correct way to cause the file manager (aka file package) to save the definition or LOADFNS a definition. It's a feature of Common LIsp that isn't consistent with the "residential? style, which presumes a source file contains a sequence of definitions and expressions to evaluate. The Medley Common Lisp compiler (cl:compile-file) will compile it if you make a (text) file.

This requires some discussion .... are lexical closures a common style or promoted by any Common LIsp implementation or book?? What are the advantages?

[wmannis]

There's an entire book called Let Over Lambda, by Doug Hoyte[1], about the different uses of this sort of lexical closure in Common Lisp. It is regularly discussed as one way to do state encapsulation (normally with more than one DEFUN under the LET).

[1] Chapter 2 is available online, and is a good example of the practice.

See also the Common Lisp Cookbook on closures.

[MattHeffron]

Lexical closures around a DEFUN seem kind of esoteric. But, in general, lexical closures around lambda, flet forms, etc seem perfectly reasonable to me.

…

On Tue, Feb 25, 2025, 4:32 PM Larry Masinter ***@***.***> wrote: THe problem with DEFUNs inside lexical scope and not at top level is there is no correct way to cause the file manager (aka file package) to save the definition or LOADFNS a definition. It's a feature of Common LIsp that isn't consistent with the "residential? style, which presumes a source file contains a sequence of definitions and expressions to evaluate. The Medley Common Lisp compiler (cl:compile-file) will compile it if you make a (text) file. This requires some discussion .... are lexical closures a common style or promoted by any Common LIsp implementation or book?? What are the advantages? — Reply to this email directly, view it on GitHub <#2016 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB7BB4SMYQYFKG3TMYM6KUD2RUDTPAVCNFSM6AAAAABWR4MQTCVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTEMZRHEZTMMY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[masinter]

issue #2051 to document the limitations and fix the specific case of DEFUN.

[pamoroso]

Concerning the possible uses of toplevel closures, the maintainers of the Common Lisp Coockbook are considering covering some use cases.