Package: emacs;
Reported by: Okamsn <okamsn <at> protonmail.com>
Date: Sun, 16 Mar 2025 18:25:02 UTC
Severity: normal
Tags: patch
Done: Stefan Monnier <monnier <at> iro.umontreal.ca>
Bug is archived. No further changes may be made.
Message #17 received at 77058 <at> debbugs.gnu.org (full text, mbox):
From: Daniel Colascione <dancol <at> dancol.org> To: Eli Zaretskii <eliz <at> gnu.org> Cc: okamsn <at> protonmail.com, 77058 <at> debbugs.gnu.org, Stefan Monnier <monnier <at> iro.umontreal.ca> Subject: Re: bug#77058: [PATCH] Add cl-with-accessors Date: Sat, 29 Mar 2025 11:59:57 -0400
Eli Zaretskii <eliz <at> gnu.org> writes:
> Ping! Okamsn, would you like to submit an updated patch?
>
>> Cc: 77058 <at> debbugs.gnu.org
>> Date: Tue, 18 Mar 2025 12:19:56 -0400
>> From: Stefan Monnier via "Bug reports for GNU Emacs,
>> the Swiss army knife of text editors" <bug-gnu-emacs <at> gnu.org>
>>
>> > The attached patch adds the macro `cl-with-accessors'. This macro uses
>> > `cl-symbol-macrolet' to treat a symbol like a use of an accessor
>> > function, such as those made by `cl-defstruct' and `cl-defclass'. This
>> > can be helpful when using repeated calls to read and write with the
>> > functions.
>>
>> Thanks. See comments below.
>>
>> > Also, in the test `cl-lib-struct-accessors' in
>> > test/lisp/emacs-lisp/cl-lib-tests.el, it checks for the keyword
>> > `:readonly', but the documented keyword is `:read-only'. Is this a
>> > typo, or was this intentional?
>>
>> I'd go with "typo".
>>
>> > --- a/doc/misc/cl.texi
>> > +++ b/doc/misc/cl.texi
>> > @@ -4225,6 +4225,19 @@ Structures
>> >
>> > Other slot options are currently ignored.
>> >
>> > +The macro @code{cl-with-accessors} can simplify code that repeatedly
>> > +accesses slots.
>> > +
>> > +@example
>> > +(defun add-one-year (person)
>> > + (cl-with-accessors ((age person-age))
>> > + person
>> > + (if (null age)
>> > + (error "Age is missing")
>> > + (cl-incf age)))
>> > + person)
>> > +@end example
>>
>> It'd be worthwhile to mention that it also works with accessors of other
>> objects than structs, e.g. `car`, or `process-filter`.
>>
>> Of course, any EIEIO user would rewrite the above to:
>>
>> (defun add-one-year (person)
>> (with-slots (age) person
>> (if (null age)
>> (error "Age is missing")
>> (cl-incf age)))
>> person)
>>
>> so maybe it's worth mentioning `with-slots` (which arguably should be
>> moved to `cl-lib`).
>>
>> > --- a/etc/NEWS
>> > +++ b/etc/NEWS
>> > @@ -2496,6 +2496,13 @@ This function is like 'type-of' except that it sometimes returns
>> > a more precise type. For example, for nil and t it returns 'null'
>> > and 'boolean' respectively, instead of just 'symbol'.
>> >
>> > +** New macro 'cl-with-accessors'
>> > +This macro is similar to 'map-let', but works with structures and their
>> > +accessors functions. It is useful when slots' accessor functions are
>> > +use repeatedly, such as reading from a slot and then writing to that
>> > +slot. Symbol macros are created for the accessor functions using
>> > +'cl-symbol-macrolet', so that they can be used with 'setq' and 'setf'.
>>
>> I think this should be moved to a CL-Lib subsection in the
>> "package-specific changes" section.
>>
>> > +;;;###autoload
>> > +(defmacro cl-with-accessors (bindings instance &rest body)
>> > + "Generate code using symbols as accessing expressions for INSTANCE inside BODY.
>> > +
>> > +This macro helps when writing code that makes repeated use of the
>> > +accessor functions of a structure or object instance, such as those
>> > +created by `cl-defstruct' and `defclass'.
>> > +
>> > +Inside BODY, SYMBOL-NAME is treated as the function call (ACCESSOR-NAME
>> > +INSTANCE) using `cl-symbol-macrolet'. SYMBOL-NAME can be used with
>> > +`setf' and `setq' as a generalized variable.
>>
>> I'd use `VAR` or `SYMBOL` rather than `SYMBOL-NAME`.
>> Probably similarly use `ACCESSOR` rather than `ACCESSOR-NAME`.
>>
>> > +\(fn ((SYMBOL-NAME ACCESSOR-NAME) ...) INSTANCE &REST BODY)"
>>
>> Should be lower case `&rest` on this last line.
>>
>> > + (declare (debug [(&rest (symbolp place)) form body])
>> > + (indent 2))
>>
>> I think `place` is wrong here. `(person-age FOO)` is a "place" but
>> `person-age` is not (or, well, it is but not the right one: when taken
>> as a place it refers to the variable, not the accessor).
>> I think it should just be `symbolp`.
>>
>> > + (cond ((null body)
>> > + (macroexp-warn-and-return "No body given" nil 'empty-body))
>>
>> Please use the same format as other similar warnings we have, such as:
>>
>> (format-message "`unless' with empty body")
>>
>>
>> - Stefan
Cool patch!
I really hate repeating myself in code, so I've been working with a
variant of the idea. It looks like this:
(cl-defmacro jr2-with-slots
((prefix object &optional (class nil class-specified))
&rest body)
"Convenience macro for field access.
In (jr2-with-slots (OBJECT TYPE) BODY..)
Make OBJECT->[SLOT] available in BODY as a generic place referencing
SLOT in OBJECT, which must be a symbol and evaluate to an object of type
TYPE, which must be a `defclass'-ed class name.
The definition of TYPE must be visible to the compiler and so must be
evaluated at both compile time and run time. Use the `jr2-defclass'
macro to automatically make the class definition
available appropriately.
For example:
(jr2-defclass my-class () ((foo) (bar)))
(cl-defmethod initialize-instance :around
((this my-class) &optional _slots)
(jr2-with-slots (this my-class)
(+ this->foo this->bar)))
In this specific case, though, prefer writing (jr2-defmethod ...), which
automatically adds a `jr2-with-slots'.
Within the lexical scope of BODY, the special forms (boundp*
OBJECT->[SLOT]) and (makunbound* OBJECT->[SLOT]) refer to EIEIO
slot-boundedness of SLOT in OBJECT.
Instead of (OBJECT TYPE), you can write (PREFIX OBJECT TYPE), in which
case the slot access symbols are PREFIX[SLOT] and OBJECT need not be
a symbol. For example:
(jr2-with-slots (foo-> (get-foo) my-foo) foo->field)
\(fn (OBJECT TYPE) &rest BODY)"
(declare (indent 1)
(debug (jr2--with-slots-spec body)))
(unless class-specified
(cl-psetf
prefix (intern (format "%s->" (cl-the symbol prefix)))
object prefix
class object))
(unless (class-p class)
(signal 'wrong-type-argument
(list (format
(concat
"%s is not a class: did you forget to make it "
"available at compile time, e.g. by writing it with "
"`jr2-defclass' or wrapping it with `eval-and-compile'?")
class))))
(let* ((slot-info
(cl-loop
for slot in (eieio-class-slots class)
for slot-name = (eieio-slot-descriptor-name slot)
for accessor-name =
(intern (concat (symbol-name prefix) (symbol-name slot-name)))
collect `(,accessor-name ,slot-name)))
;; Generate helpful symbol name for cl-assert
(object-sym (gensym
(if (symbolp object)
(format "#%s#" object) "jr2--object"))))
`(cl-macrolet ((boundp* (form)
(jr2--rewrite-slot-access 'slot-boundp form))
(makunbound* (form)
(jr2--rewrite-slot-access 'slot-makeunbound form)))
(cl-symbol-macrolet
,(cl-loop
for (accessor-name slot-name) in slot-info
collect `(,accessor-name (slot-value ,object-sym ',slot-name)))
(let ((,object-sym ,object))
(cl-check-type ,object-sym ,class)
,@body)))))
(cl-defmacro jr2-with-slots* (specs &rest body)
"Like `jr2-with-slots' but with a list of specifications instead of
just one."
(declare (indent 1) (debug (([&rest jr2--with-slots-spec]) def-body)))
(if (null specs)
(macroexp-progn body)
`(jr2-with-slots ,(car specs)
(jr2-with-slots* ,(cdr specs) ,@body))))
(eval-and-compile
(defun jr2--extract-kw-arg-names (kw-args)
(cl-loop
for arg in kw-args
if (memq arg '(&allow-other-keys)) do (ignore arg)
else if (and (symbolp arg) (string-prefix-p "&" (symbol-name arg)))
do (error "unknown &-directive %S" arg)
else collect (intern (concat ":"
(symbol-name
(cl-the (and symbol (not null))
(or (car-safe arg) arg))))))))
I have various other macros that work with the system, e.g. a
jr2-defmethod. Overall effect is you can write code like this:
(jr2-defclass my-class () (a :type integer))
...
(jr2-defmethod my-method ((this my-class) b)
(+ this->a b))
Look ma, no repetition!
Now, I realize the arrow syntax isn't everyone's cup of tea, but it's
legible and concise --- moreso than oref IMHO. You could get away from
the need for compile-time visible of classes by switching from
cl-symbol-macrolet to a custom expander function (which is how
cl-symbol-macrolet itself is implemented), but I was too lazy.
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.