[slime-devel] the inspector (warning allegro and lispworks users)
i just changed the swank inspector api. inspect-for-emacs (ex inspected-parts) now takes a second argument which is an inspector object. this means that we get implementation specific inspectors where available and default generic inspectors otherwise and clears up the load order issues i was having. i went and re-wrote all the implementation inspector methods to use the new return-value format (previously those methods were being ignored so the issue never came up). HOWEVER: I also made these changes for allegro and lispworks, even though i don't have neither an allegro nor a lispworks to test the changes on. i may very well have b0rken things. -- -Marco Ring the bells that still can ring. Forget your perfect offering. There is a crack in everything. That's how the light gets in. -Leonard Cohen
On Tue, 14 Sep 2004 18:17:33 +0200, marco <mb@bese.it> said:
Marco> i just changed the swank inspector api. inspect-for-emacs (ex Marco> inspected-parts) now takes a second argument which is an inspector Marco> object. this means that we get implementation specific inspectors Marco> where available and default generic inspectors otherwise and clears up Marco> the load order issues i was having. Marco> i went and re-wrote all the implementation inspector methods to use Marco> the new return-value format (previously those methods were being Marco> ignored so the issue never came up). HOWEVER: I also made these Marco> changes for allegro and lispworks, even though i don't have neither an Marco> allegro nor a lispworks to test the changes on. i may very well have Marco> b0rken things. No problems, it seems to be working in LispWorks. I'm not sure that the getter and setter functions should be returned though, because they are part of the inspector, not part of the object. BTW, is there an intentional change to the working practice for commit messages in slime? I see that you use the whole ChangeLog entry in every file, rather than just the paragraph that describes the changes in that file (without the prefix). __Martin
Martin Simmons <martin@xanalys.com> writes:
BTW, is there an intentional change to the working practice for commit messages in slime? I see that you use the whole ChangeLog entry in every file, rather than just the paragraph that describes the changes in that file (without the prefix).
i thought that was the working practice. i'll be happy to do whatever's expected as long as someone would point out what that is. -- -Marco Ring the bells that still can ring. Forget your perfect offering. There is a crack in everything. That's how the light gets in. -Leonard Cohen
marco <mb@bese.it> writes:
Martin Simmons <martin@xanalys.com> writes:
BTW, is there an intentional change to the working practice for commit messages in slime? I see that you use the whole ChangeLog entry in every file, rather than just the paragraph that describes the changes in that file (without the prefix).
i thought that was the working practice. i'll be happy to do whatever's expected as long as someone would point out what that is.
If you commit with Emacs then `C-c C-a' in the *cvs-commit* buffer should extract the relevant part of the changelog for the current file (and strip out unneeded filename info and whitespace). See e.g. 'cvs log slime.el' for examples. BTW, this inspector feels nice! I would like source code lines to break at 80 columns though :-) Cheerio, Luke P.S., sorry about all the spam that has been getting through on the slime-cvs list, I've changed it to only allow posts from subscribers or from @common-lisp.net addresses (as sent by cvs).
Was now again abel to compile slime with sbcl 0.8.14.24. For anyone with have encountered troubles I would suggest to try this version. Of course I can not guarantee that it'll work for you. Regards Friedrich
Am Tue, 14 Sep 2004 18:17:33 +0200 schrieb marco <mb@bese.it>:
i just changed the swank inspector api. inspect-for-emacs (ex inspected-parts) now takes a second argument which is an inspector object. this means that we get implementation specific inspectors where available and default generic inspectors otherwise and clears up the load order issues i was having.
i went and re-wrote all the implementation inspector methods to use the new return-value format (previously those methods were being ignored so the issue never came up). HOWEVER: I also made these changes for allegro and lispworks, even though i don't have neither an allegro nor a lispworks to test the changes on. i may very well have b0rken things.
Just some minor fixes. Works all fine now.
Inlined is a little patch to have method documentation displayed and have methods sorted by specifity. This resulted from a #lisp discussion and is pretty much inspired by the McCLIM lister. (I think we can steal even more from there ;) ) Hm, McCLIM is GPL. Does that interfere with Slime's license? Unfortunately I had to change the swank-mop, which will break Slime on every implementation other than acl62. We could avoid this by simply moving my changes to swank-allegro.lisp, but I think this also fixes a bug for all other implementations. Ok, to call it by name, the current implementation had problems with eql-specializers. So I added the eql-spezializer class to the mop. (In clisp we this should probably simply be an alias for cons.) At best the sorting should be customizable. Also it currently doesn't consider argument-precedence order and has problems with non-finalized classes (which I currently don't know how to solve). Oh, and finally some little typo fixes. ,--- Patches --- Index: swank-backend.lisp =================================================================== RCS file: /project/slime/cvsroot/slime/swank-backend.lisp,v retrieving revision 1.67 diff -u -r1.67 swank-backend.lisp --- swank-backend.lisp 14 Sep 2004 16:01:07 -0000 1.67 +++ swank-backend.lisp 16 Sep 2004 00:15:50 -0000 @@ -42,6 +42,7 @@ #:standard-slot-definition #:standard-method #:standard-class + #:eql-specializer ;; standard-class readers #:class-default-initargs #:class-direct-default-initargs Index: swank-allegro.lisp =================================================================== RCS file: /project/slime/cvsroot/slime/swank-allegro.lisp,v retrieving revision 1.57 diff -u -r1.57 swank-allegro.lisp --- swank-allegro.lisp 14 Sep 2004 21:24:58 -0000 1.57 +++ swank-allegro.lisp 16 Sep 2004 00:16:13 -0000 @@ -36,6 +36,7 @@ mop::standard-slot-definition cl:method cl:standard-class + mop:eql-specializer ;; standard-class readers mop:class-default-initargs mop:class-direct-default-initargs @@ -73,6 +74,10 @@ (defun swank-mop:slot-definition-documentation (slot) (documentation slot)) + +(defmethod swank-mop:class-name ((x swank-mop:eql-specializer)) + `(eql ,(mop:eql-specializer-object x))) + ;;;; TCP Server Index: swank.lisp =================================================================== RCS file: /project/slime/cvsroot/slime/swank.lisp,v retrieving revision 1.237 diff -u -r1.237 swank.lisp --- swank.lisp 15 Sep 2004 17:29:39 -0000 1.237 +++ swank.lisp 16 Sep 2004 00:21:25 -0000 @@ -2531,7 +2531,7 @@ (values "An array." `("Dimensions: " (:value ,(array-dimensions array)) (:newline) - "It's element type is: " (:value ,(array-element-type array)) + "Its element type is: " (:value ,(array-element-type array)) (:newline) "Total size: " (:value ,(array-total-size array)) (:newline) @@ -2583,7 +2583,7 @@ " " (:action ,(format nil "[remove name ~S (does not affect class object)]" symbol) (lambda () (setf (find-class symbol) nil))))))) (values "A symbol." - `("It's name is: " (:value ,(symbol-name symbol)) + `("Its name is: " (:value ,(symbol-name symbol)) (:newline) ;; check to see whether it is a global variable, a ;; constant, or a symbol macro. @@ -2665,18 +2665,61 @@ collect "#<unbound>" collect '(:newline))))) + +(defun sorted-methods (gf) + ;;FIXME: How to deal with argument-precedence-order? + (let* ((methods (copy-list (swank-mop:generic-function-methods gf))) + (lambda-list (swank-mop:generic-function-lambda-list gf))) + ;; sorter function (most specific is defined as smaller) + (flet ((method< (meth1 meth2) + ;; First ordering rule is by qualifiers, that is :before-methods + ;; com before :around methods, before primary methods, before + ;; :after methods, other qualifiers are treated like none at all + ;; (so like primary methods) + (let ((qualifier-order '(:before :around nil :after))) + (let ((q1 (or (position (first (swank-mop:method-qualifiers meth1)) qualifier-order) 2)) + (q2 (or (position (first (swank-mop:method-qualifiers meth2)) qualifier-order) 2))) + (cond ((< q1 q2) (return-from method< t)) + ((> q1 q2) (return-from method< nil))))) + ;; If qualifiers are equal, go by arguments + (loop for sp1 in (swank-mop:method-specializers meth1) + for sp2 in (swank-mop:method-specializers meth2) + do (cond + ((eq sp1 sp2)) ;; continue comparision + ;; an eql specializer is most specific + ((typep sp1 'swank-mop:eql-specializer) + (return-from method< t)) + ((typep sp2 'swank-mop:eql-specializer) + (return-from method< nil)) + ;; otherwise the longer the CPL the more specific + ;; the specializer is + (t (let ((l1 (if (swank-mop:class-finalized-p sp1) + (length (swank-mop:class-precedence-list sp1)) + 0)) + (l2 (if (swank-mop:class-finalized-p sp2) + (length (swank-mop:class-precedence-list sp2)) + 0))) + (cond + ((> l1 l2) + (return-from method< t)) + ((< l1 l2) + (return-from method< nil)))))) + finally (return nil)))) + (declare (dynamic-extent #'method<)) + (sort methods #'method<)))) + (defmethod inspect-for-emacs ((gf standard-generic-function) (inspector t)) (declare (ignore inspector)) (values "A generic function." `("Name: " (:value ,(swank-mop:generic-function-name gf)) (:newline) - "It's argument list is: " ,(princ-to-string (swank-mop:generic-function-lambda-list gf)) (:newline) + "Its argument list is: " ,(princ-to-string (swank-mop:generic-function-lambda-list gf)) (:newline) "Documentation: " (:newline) ,(princ-to-string (documentation gf t)) (:newline) - "It's method class is: " (:value ,(swank-mop:generic-function-method-class gf)) (:newline) + "Its method class is: " (:value ,(swank-mop:generic-function-method-class gf)) (:newline) "It uses " (:value ,(swank-mop:generic-function-method-combination gf)) " method combination." (:newline) "Methods: " (:newline) ,@(loop - for method in (swank-mop:generic-function-methods gf) + for method in (sorted-methods gf) collect `(:value ,method , (with-output-to-string (meth) (let ((specs (swank-mop:method-specializers method)) @@ -2690,6 +2733,13 @@ collect " " collect (let ((meth method)) `(:action "[remove method]" ,(lambda () (remove-method gf meth)))) + collect '(:newline) + collect " Documentation: " + ;; Display the first sentence or the first 50 characters of the docstring + ;; FIXME: Idea's for something less random to chop the docline? + collect (let ((doc (documentation method t))) + (or (subseq doc 0 (min (1+ (or (position #\. doc) 49)) 50)) + "None")) collect '(:newline))))) (defmethod inspect-for-emacs ((method standard-method) (inspector t))
"Thomas Schilling" <tjs_ng@yahoo.de> writes:
- "It's argument list is: " ,(princ-to-string (swank-mop:generic-function-lambda-list gf)) (:newline) + "Its argument list is: " ,(princ-to-string
damn it. i knew i was getting that wrong. -- -Marco Ring the bells that still can ring. Forget your perfect offering. There is a crack in everything. That's how the light gets in. -Leonard Cohen
participants (5)
-
Friedrich Dominicus -
Luke Gorrie -
marco -
Martin Simmons -
Thomas Schilling