Proper backquoting and variable capture with labels+macrolet within defmacro - macros

I'm writing a small special-purpose language in Common Lisp using a defmacro. I cannot figure out the proper backquote procedure to have a variable defined in a top-level let statement, shadowed in a nested macrolet, and returned in a nested labels, all point to the same symbol.
Here's some small example code that shows the problem:
(defmacro with-keylang (&body body)
(let ((keyblank-size-x (gensym)))
`(let ((,keyblank-size-x 42))
(labels ((keyblank-size-x ()
,keyblank-size-x))
(macrolet ((with-keyblank-size-x (size-x &body body)
`(let ((,',keyblank-size-x ,size-x))
,#body)))
,#body)))))
CL-USER>
(with-keylang
(print (keyblank-size-x)))
42
42
All is well so far.
CL-USER>
(with-keylang
(with-keyblank-size-x 24
(print (keyblank-size-x))))
;Compiler warnings :
; In an anonymous lambda form: Unused lexical variable #:G123171
42
42
There's the problem. I want the symbol representing keyblank-size-x to be shadowed with the value 24, and this is not happening.
I have a feeling that the ,', backquote pattern isn't proper for this case, as this quotes the symbol representing keyblank-size-x, and is therefore not eq. But if I try ,,, it doesn't work, and I get this interesting compiler error:
While compiling WITH-KEYBLANK-SIZE-X :
Illegal reference to lexically defined variable #:G123192.
[Condition of type CCL::COMPILE-TIME-PROGRAM-ERROR]
EDIT:
The keyblank-size-x variable was lexically scoped, and I wanted dynamic scope for this particular case. So here's the rewrite declaring the keyblank-size-x variable to have dynamic scope:
(defmacro with-keylang (&body body)
(let ((keyblank-size-x (gensym)))
`(let ((,keyblank-size-x 42))
(declare (special ,keyblank-size-x))
(labels ((keyblank-size-x ()
,keyblank-size-x))
(macrolet ((with-keyblank-size-x (size-x &body body)
`(let ((,',keyblank-size-x ,size-x))
(declare (special ,',keyblank-size-x))
,#body)))
,#body)))))
And the test code:
CL-USER>
(with-keylang
(with-keyblank-size-x 25
(with-keyblank-size-x 21
(print (keyblank-size-x)))
(print (keyblank-size-x)))
(print (keyblank-size-x)))
21
25
42
42

If we fully expand the code (here using the Walk command in LispWorks):
(LET ((#:G19508 42))
(LABELS ((KEYBLANK-SIZE-X () #:G19508))
(MACROLET ((WITH-KEYBLANK-SIZE-X (SIZE-X &BODY BODY)
`(LET ((#:G19508 ,SIZE-X))
,#BODY)))
(LET ((#:G19508 24))
(PRINT (KEYBLANK-SIZE-X))))))
Rebinding #:G19508 has no effect and can't have. The function keyblank-size-x has a different lexical binding. This is just the usual effect of lexical binding.

Related

Extracting and executing a closure/lambda from a plist macro body

I'm trying to create a macro (bar) that should be used like this:
(let ((my-var "foo"))
(bar ("some")
:buzz (lambda () (format t "~a~%" my-var))))
The macro should basically just FUNCALL the lambda with taking MY-VAR into account.
What I've come up with is this:
(defmacro bar ((thing) &body body)
`(funcall (coerce (getf (list ,#body) :buzz) 'function)))
This works, it prints "foo".
But I'd like to know if this is how this is done or if there is a better way of doing this.
Well, for a start if you don't need the extra, unused argument, then this is just funcall, since (lambda ...) denotes a function.
(let ((my-var "foo"))
(funcall (lambda () (format t "~a~%" my-var))))
So even if you didn't want to call this funcall you could write it as a function, not a macro: it's not doing any code transformation. But presumably you actually do need this extra argument, and you are intending to use it for some purpose in the expansion, so what you need then is a macro which takes keyword arguments and just expands into a suitable funcall form:
(defmacro bar ((thing) &key (buzz '(lambda ())))
(declare (ignore thing))
`(funcall ,buzz))

How can macro variable capture happen with a gensym symbol?

I'm learning common lisp. I have written a version of the once-only macro, which suffers from an unusual variable capture problem.
My macro is this:
(defmacro my-once-only (names &body body)
(let ((syms (mapcar #'(lambda (x) (gensym))
names)))
``(let (,,#(mapcar #'(lambda (sym name) ``(,',sym ,,name))
syms names))
,(let (,#(mapcar #'(lambda (name sym) `(,name ',sym))
names syms))
,#body))))
The canonical version of only-once is this:
(defmacro once-only ((&rest names) &body body)
(let ((gensyms (loop for n in names collect (gensym))))
`(let (,#(loop for g in gensyms collect `(,g (gensym))))
`(let (,,#(loop for g in gensyms for n in names collect ``(,,g ,,n)))
,(let (,#(loop for n in names for g in gensyms collect `(,n ,g)))
,#body)))))
The difference, as far as I can tell, is that the canonical version generates new symbols for every expansion of the macro using only-once. For example:
CL-USER> (macroexpand-1 '(once-only (foo) foo))
(LET ((#:G824 (GENSYM)))
`(LET (,`(,#:G824 ,FOO))
,(LET ((FOO #:G824))
FOO)))
T
CL-USER> (macroexpand-1 '(my-once-only (foo) foo))
`(LET (,`(,'#:G825 ,FOO))
,(LET ((FOO '#:G825))
FOO))
T
The variable my macro uses to store the value of foo is the same for every expansion of this form, in this case it would be #:G825. This is akin to defining a macro like the following:
(defmacro identity-except-for-bar (foo)
`(let ((bar 2))
,foo))
This macro captures bar, and this capture manifests when bar is passed to it, like so:
CL-USER> (let ((bar 1))
(identity-except-for-bar bar))
2
However, I cannot think of any way to pass #:G825 to a macro that uses my-only-once so that it breaks like this, because the symbols gensym returns are unique, and I cannot create a second copy of it outside of the macro. I assume that capturing it is unwanted, otherwise the canonical version wouldn't bother adding the additional layer of gensym. How could capturing a symbol like #:G826 be a problem? Please provide an example where this capture manifests.
We can demonstrate a behavioral difference between my-once-only and once-only:
Let's store our test form in a variable.
(defvar *form* '(lexalias a 0 (lexalias b (1+ a) (list a b))))
This test form exercises a macro called lexalias, which we will define in two ways. First with once-only:
(defmacro lexalias (var value &body body)
(once-only (value)
`(symbol-macrolet ((,var ,value))
,#body)))
(eval *form*) -> (0 1)
Then with my-once-only:
(defmacro lexalias (var value &body body)
(my-once-only (value)
`(symbol-macrolet ((,var ,value))
,#body)))
(eval *form*) -> (1 1)
Oops! The problem is that under my-once-only, both a and b end up being symbol-macrolet aliases for exactly the same gensym; the returned expression (list a b) ends up being something like (list #:g0025 #:g0025).
If you're writing a macro-writing helper that implements once-only evaluation, you have no idea how the symbol is going to be used by the code which calls the macro, whose author uses your once-only tool. There are two big unknowns: the nature of the macro and of its use.
As you can see, if you don't make fresh gensyms, it will not work correctly in all conceivable scenarios.

How is sharp quote (#') different from symbol-function?

To me these operators seem to do the same thing. Both take a symbol and return the function associated with it. Is there any difference?
elisp evaluation returns the following:
(defun foo (x) (+ 1 x))
foo
(foo 3)
4
#'foo
Which I don't understand either.
Furthermore is there a difference between common lisp and elisp? I'm learning from resources on either.
Common Lisp:
SYMBOL-FUNCTION can't retrieve functions from lexically bound functions. FUNCTION references the lexically bound function by default. #'foo is just a shorter notation for (FUNCTION foo).
CL-USER 1 > (defun foo () 'foo)
FOO
CL-USER 2 > (flet ((foo () 'bar))
(list (funcall (symbol-function 'foo))
(funcall #'foo)
(funcall (function foo))
(eq (function foo) (symbol-function 'foo))))
(FOO BAR BAR NIL)
CL-USER 3 > (eq (function foo) (symbol-function 'foo))
T
Rainer's answer discusses the things that you can do with function that you can't do with symbol-function, namely retrieve the value of lexically scoped functions, but there are a few other differences too.
The special operator FUNCTION
The special operator FUNCTION provides a way to find the functional value of a name in a lexical environment. Its argument is either a function name or a lambda expression. That function can take a lambda expression means that you can write: (function (lambda (x) (list x x))). The term function name includes more than just symbols. It also includes lists of the form (setf name), which means that you can do things like (function (setf car)).
The accessor SYMBOL-FUNCTION
The accessor symbol-function, on the other hand, lets you retrieve and set the functional value of a symbol. Since it requires a symbol, you can't do (symbol-function (lambda …)) or (function (setf name)). Symbol-function also can't see lexical environments; it only works with global definitions. E.g.,
(flet ((foo () 'result))
(symbol-function 'foo))
;=> NIL
Since symbol-function is an accessor, you can change the value of a function's symbol with it. E.g.:
CL-USER> (setf (symbol-function 'foo) (lambda () 42))
#<FUNCTION (LAMBDA ()) {1005D29AAB}>
CL-USER> (foo)
42
CL-USER> (setf (symbol-function 'foo) (lambda () 26))
#<FUNCTION (LAMBDA ()) {1005D75B9B}>
CL-USER> (foo)
26
The accessor FDEFINITION
There's also the accessor fdefinition which is kind of like symbol-function in that is can only access globally defined functions, but kind of like function in that it can access of symbols and (setf name) lists. However, it does not retrieve the value of lambda functions.

Unquoting without eval

I store some macros in quoted form (because in fact they produce lambdas with tricky lexical environment and I prefer store and serialize them as lists). So now I'm trying:
(defun play (s)
(funcall (macroexpand s)))
Macroexpand evaluates quoted lambda, so funcall can't run it. How to unquote result of macroexpand without eval? Because in my case it would cause indefensible security hole.
MORE INFO:
What I get look like this (in simplest case):
FUNCALL: #1=#'(LAMBDA (#:G6008) (SYMBOL-MACROLET NIL T)) is not a function name; try using a symbol instead
and symbol-macrolet is what actually builds up "tricky lexical environment" inside lambda.
Macroexpand evaluates quoted lambda, so funcall can't run it. How to
unquote result of macroexpand without eval? Because in my case it
would cause indefensible security hole.
I think that Sylwester's comment about the XY problem is probably right here; it sounds like you're trying to do something that might be done better in a different way. That said, if you have a list that's a lambda expression, you can use coerce to get a function object instead of using eval. That is, you can do this:
CL-USER> (funcall '(lambda () 42))
; Error, like you've been having
CL-USER> (funcall (coerce '(lambda () 42) 'function))
42 ; turned the list (lambda () 42) into a function and called it
This is described in the documentation for coerce; when the "output" type is function, this is what happens with the object argument:
If the result-type is function, and object is any function name that
is fbound but that is globally defined neither as a macro name nor as
a special operator, then the result is the functional value of object.
If the result-type is function, and object is a lambda expression,
then the result is a closure of object in the null lexical
environment.
Thus, if you have a function that returns list of the form (lambda ...), you can use coerce and funcall with its result. This includes macroexpansions, although you may want to use macroexpand-1 rather than macroexpand, because lambda is already a macro, so if you expand too far, (lambda () ...) turns into (function (lambda () ...)).
CL-USER> (defmacro my-constantly (value)
`(lambda () ,value))
MY-CONSTANTLY
CL-USER> (macroexpand-1 '(my-constantly 36))
(LAMBDA () 36)
T
CL-USER> (funcall (coerce (macroexpand-1 '(my-constantly 36)) 'function))
36
If you try that with the plain macroexpand, though, there's a problem. Consider yourself warned:
CL-USER> (macroexpand '(my-constantly 36))
#'(LAMBDA () 36) ; not a list, but a function
T
CL-USER> (funcall (coerce (macroexpand '(my-constantly 36)) 'function))
; Error. :(
I think this is a case where you will find that the REPL is your friend. To get you started:
cl-user> (defmacro foo () #'(lambda () :hi))
foo
cl-user> (foo)
#<Compiled-function (:internal foo) (Non-Global) #x3020014F82FF>
cl-user> (funcall *)
:hi
cl-user> (macroexpand '(foo))
#<Compiled-function (:internal foo) (Non-Global) #x3020014F82FF>
t
cl-user> (funcall *)
:hi
cl-user>
I'll note in passing that the lambda form that appears in your example takes an argument, while your funcall doesn't provide one.
There are several issues with this. First I think this is a XY problem so if you show more of your problem I guess we might find a solution for it.
It has nothing with unquoting since when evaluation a quoted expression it's not longer quoted, but it's turn into data representation of the original quoted expression. You must eval data if you want to run it.
With eval you won't get the current lexical scope. So since you mention eval and lexical environment in the same post makes me thing you won't get what you want.
Now it's no problem making a list of functions, like this:
(list (lambda (x) (+ x x))
This is not quoted since I use list and teh lambda is evaluated to a closure. If it was assigned to variable x you could call it with (funcall (car x) 10)) ; ==> 20
EDIT
I actually made the same error message with this code:
(defmacro test () '#'(lambda (x) x))
(macroexpand '(test)) ; ==> #'(lambda (x) x) ; t
(funcall (macroexpand '(test)) 5) ; ==>
*** - funcall: #'(lambda (x) x) is not a function name; try using a symbol
instead
It doesn't work since you cannot call a lambda unevaluated. You need to call the closure (function) which is the result of the evaluation of the lambda form. If you would instead not quote it you'll have that evaluation in the macro:
(defmacro test () #'(lambda (x) x))
(macroexpand '(test)) ; ==> #<function :lambda (x) x> ;
(funcall (macroexpand '(test)) 5) ; ==> 5
Actually I cannot see why you would need to make this a macro. Lets make it a function instead.
(defun test () #'(lambda (x) x))
(funcall (test) 5 ) ; ==> 5
Even if this was more comples in most cases you would do with a closure.

Saving symbol-names of argument in macro

One of the exercises in Paul Grahams ANSI Common Lisp book is this: Define a macro that takes a list of variables and a body of code, and ensures that the variables revert to their original values after the body of code is evaluated.
The problem I'm having with this exercise is how to save the symbol-names of the input variables. Below I have a start where I only save the values that the symbols are bound to.
(defmacro save-run (varlist &body body)
`(let ((valuelist (list ,#varlist)))
(format t "valuelist: ~A" valuelist)))
(let ((a 5)(b 6))
(values '(a b))
(save-run (a b)
(setf a 7)
(setf b 8)))
[507]> valuelist: (5 6)
Edit: Here's a solution where the variables are saved and then restored (using tips from finnw below). But shadowing the variables as in Vatine's answer is probably more elegant.
(defmacro save-run (varlist &body body)
`(let ((valuelist (list ,#varlist)))
,#body
(multiple-value-setq ,varlist (values-list valuelist))))
To get the list of symbols, all you need to do is quote the contents of VARLIST:
(defmacro save-run (varlist &body body)
`(let ((namelist ',varlist)
(valuelist (list ,#varlist)))
(format t "namelist: ~A~%" namelist)
(format t "valuelist: ~A~%" valuelist)))
I suspect this will not be useful in the final defnition however. There is not much you can do with the list of symbols at run-time. Instead, look for a good place to insert the list in the macro expansion.
Also you might want to use a GENSYM instead of the hard-coded variable name VALUELIST:
(defmacro save-run (varlist &body body)
(let ((valuelist (gensym)))
`(let ((,valuelist (list ,#varlist)))
,#body
(setf (values ,#varlist) (values-list ,valuelist)))))
Personally, I'd introduce another binding layer for the variables we want to save.
You have the list of variables in varlist so something like this may work:
(defmacro save-run (varlist &body body)
`(let ,(loop for var in varlist
collect (list var var))
,#body))
I really liked Vatine's approach. Here is an implementation that expands to the same code, but uses mapcar instead of the loop macro:
(defmacro save-run (varlist &body body)
`(let ,(mapcar #'list varlist varlist)
,#body))