1 / 36

66-2210-01 Programming in Lisp

66-2210-01 Programming in Lisp. Lecture 6 - Structure; Case Study: Blocks World. Structure. Lisp's use of pointers Let you put any value anywhere Details are taken care of by the Lisp interpreter What goes on "under the hood"? It's useful to know this sometimes

nenet
Télécharger la présentation

66-2210-01 Programming in Lisp

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. 66-2210-01 Programming in Lisp Lecture 6 - Structure; Case Study: Blocks World Alok Mehta - Programming in Lisp - Lecture 6

  2. Structure • Lisp's use of pointers • Let you put any value anywhere • Details are taken care of by the Lisp interpreter • What goes on "under the hood"? • It's useful to know this sometimes • Some functions let you get down to low level details of pointers Alok Mehta - Programming in Lisp - Lecture 6

  3. Shared Structure • Lists can share conses in common • > (setf part (list 'b 'c)) • (B C) • > (setf whole (cons 'a part)) • (A B C) • > (tailp part whole) • T PART WHOLE A B C Alok Mehta - Programming in Lisp - Lecture 6

  4. Tailp • Definition of tailp • (tailp object list) • returns T if object is a tail of list • Example implementation • > (defun our-tailp (x y) • (or (eql x y) • (and (consp y) • (our-tailp x (cdr y))))) • An object is a tail of itself (base case) • NIL is a tail of every proper list • Recursively take cdr's of the list, until you test for (eql NIL NIL) Alok Mehta - Programming in Lisp - Lecture 6

  5. Shared Structure • Lists can share structure without either being a tail of the other • > (setf part (list 'b 'c) • whole1 (cons 1 part) • whole2 (cons 2 part)) PART WHOLE1 1 B C WHOLE2 2 Alok Mehta - Programming in Lisp - Lecture 6

  6. Top-Level List Structure • Sometimes want to treat data structure as a list • Include only the conses that make up the list • Don't recurse into conses that make up elements • This is the top-level list structure • Other times, want to treat data structure as a tree • All conses are important and treated uniformly A D B C Alok Mehta - Programming in Lisp - Lecture 6

  7. Copy-list vs. Copy-tree Original A D B C Copy-list Copy-tree A D B C Alok Mehta - Programming in Lisp - Lecture 6

  8. Copy-list vs. Copy-tree (code) • Example implementation of copy-list • > (defun our-copy-list (lst) • (if (null lst) • nil • (cons (car lst) (our-copy-list (cdr lst))))) • Example implementation of copy-tree • > (defun our-copy-tree (lst) • (if (atom tr) • tr • (cons (our-copy-tree (car tr)) • (our-copy-tree (cdr tr))))) • Lisp handles pointers automatically • Then… why do we care? • Because some Lisp functions can modify structures Alok Mehta - Programming in Lisp - Lecture 6

  9. Setf revisited • > (setf whole (list 'a 'b 'c) • tail (cdr whole)) • (B C) • > (setf (second tail) 'e) • E • > tail • (B E) • > whole • (A B E) TAIL WHOLE A B C Alok Mehta - Programming in Lisp - Lecture 6

  10. Operators • Avoid modifying lists • Operators like setf, pop, rplaca, … • If you need to (or want to) modify a list • Make sure it's not shared • Or, make sure the shared update works as expected • Or, make changes to a copy of the list • > (setf whole (list 'a 'b 'c) • tail (cdr whole)) • (B C) • > (setf tail (cons (first tail) • (cons 'e (rest (rest tail))))) • (B C) • > tail • (B C) • > whole • (A B E) Alok Mehta - Programming in Lisp - Lecture 6

  11. Parameter passing • Parameters are passed by value • The value is copied into the function • If the value is a list… • The entire list is not copied • The reference to the list is copied • A function can permanently alter a list passed in as a parameter! • Similar to parameter passing in Java • Can cause unintentional errors • But, can also be useful Alok Mehta - Programming in Lisp - Lecture 6

  12. Queues • Queue data structure • First in, First out • Compare to stacks (last in, first out) • Stacks are easy in Lisp • Insert (push) / Retrieve (pop) happen at the same end of the list • Queues are harder • Insert (enqueue) / Retrieve (dequeue) happen at opposite ends Q1 A B C Alok Mehta - Programming in Lisp - Lecture 6

  13. Sample implementation • Implementation of Queue (Inefficient) • > (defmacro dequeue (q) `(pop ,q)) • > (defmacro enqueue (o q) • `(setf ,q (append ,q (cons ,o nil)))) • Implementation of Queue using CLOS (Inefficient) • > (defclass queue () • ((front :accessor front :initform nil))) • > (defmethod dequeue ((q queue)) • (pop (front q))) • > (defmethod enqueue (o (q queue) &aux (node (cons o nil))) • (setf (front q) (append (front q) node))) • Example usage • > (setf a (make-instance 'queue)) • > (enqueue 'a a) • > (enqueue 'b a) • > (dequeue a) Alok Mehta - Programming in Lisp - Lecture 6

  14. Efficient Queue Implementation • Efficient Implementation of Queue using CLOS • > (defclass queue () • ((front :accessor front :initform nil) • (back :accessor back :initform nil))) • > (defmethod enqueue (o (q queue) &aux (node (cons o nil))) • (if (eql (front q) nil) ; first one? • (setf (front q) node (back q) node) • (progn ; not first time • (setf (cdr (back q)) node) • (setf (back q) (cdr (back q)))) • )) • > (defmethod dequeue ((q queue)) • (if (eql (cdr (front q)) nil) (setf back nil)) ;last one • (pop (front q))) Alok Mehta - Programming in Lisp - Lecture 6

  15. Destructive Functions • Several functions update the lists passed to them • Examples • Delete (destructive version of remove) • > (setf a '(a b a d a)) • (A B A D A) • > (remove 'a a) ; Doesn't change A • (B D) • > (delete 'a A) ; Changes A • (B D) • Nconc (destructive version of append) • > (defun our-nconc (x y) • (if (consp x) • (progn • (setf (cdr (last x)) y) • x) • y)) Alok Mehta - Programming in Lisp - Lecture 6

  16. Case Study: The Blocks World • Rules • There are three kinds of movable objects: bricks, wedges, balls. • Robot has one hand. It can grasp any movable block that has nothing on top of it. • Every block is either held by the hand or supported by exactly one brick or the table. No block can overhang from its support. • Although a movable block can be moved to the top of a wedge or a ball, neither wedges nor balls can support anything. • Supporting bricks can support more than one block, as long as there is room. • The table is wide enough for all of the blocks to fit on it at once. Alok Mehta - Programming in Lisp - Lecture 6

  17. Case Study: The Blocks World • Several moves are required to put block B1 on block B2 • Sample path • Move W7; Move B4; Move B1 onto B2 P <=== Robotic Hand W5 W7 B4 B3 B1 B2 B6 L8 Table Alok Mehta - Programming in Lisp - Lecture 6

  18. Class Hierarchy Basic-block Hand Load-bearing-block Movable-block Table Brick Wedge Ball Alok Mehta - Programming in Lisp - Lecture 6

  19. Class Definitions • Basic-Block • > (defclass basic-block() • ((name :accessor block-name :initarg :name) • (width :accessor block-width :initarg :width) • (height :accessor block-height :initarg :height) • (position :accessor block-position :initarg :position) • (supported-by :accessor block-supported-by • :initform nil))) • Supported-by => What the block is supported by (what's underneath it) • Movable-block • > (defclass movable-block (basic-block) ()) • Load-bearing-block • Has new field • > (defclass load-bearing-block (basic-block) • ((support-for :accessor block-support-for • :initform nil))) • Support-for => What the block is a support for (what's on top of it) Alok Mehta - Programming in Lisp - Lecture 6

  20. Class Definitions (cont.) • Brick, wedge, ball, table • > (defclass brick (movable-block load-bearing-block) ()) • > (defclass wedge (movable-block) ()) • > (defclass ball (movable-block) ()) • Table • > (defclass table (load-bearing-block) ()) • Robotic Hand • > (defclass hand () • ((name :accessor hand-name :initarg :name) • (position :accessor hand-position :initarg :position) • (grasping :accessor hand-grasping :initform nil))) Alok Mehta - Programming in Lisp - Lecture 6

  21. Creating Blocks in Blocks World (defvar *blocks* (list (make-instance 'table :name 'table :width 20 :height 0 :position '(0 0)) (make-instance 'brick :name 'b1 :width 2 :height 2 :position '(0 0)) (make-instance 'brick :name 'b2 :width 2 :height 2 :position '(2 0)) (make-instance 'brick :name 'b3 :width 4 :height 4 :position '(4 0)) (make-instance 'brick :name 'b4 :width 2 :height 2 :position '(8 0)) (make-instance 'wedge :name 'w5 :width 2 :height 4 :position '(10 0)) (make-instance 'brick :name 'b6 :width 4 :height 2 :position '(12 0)) (make-instance 'wedge :name 'w7 :width 2 :height 2 :position '(16 0)) (make-instance 'ball :name 'L8 :width 2 :height 2 :position '(18 0)) )) (defvar *hand* (make-instance 'hand :name '*hand* :position '(0 6))) P B3 W5 W7 B1 B2 B4 B6 L8 Table Alok Mehta - Programming in Lisp - Lecture 6

  22. Initializing Blocks World • Create other global variables for convenience • > (dolist (l *blocks*) (set (block-name l) l)) • This sets the global variables TABLE, B1, B2, … W7, L8 to their respective block • All blocks rest on the table initially • > (dolist (l (rest *blocks*)) ; For each, except table • (setf (block-supported-by l) table) • (push l (block-support-for table))) • Load-bearing-block has a method block-support-for • This was automatically generated • Create a dummy stub for this method in basic-block • > (defmethod block-support-for ((object basic-block)) • nil) • By default, basic-blocks do not have anything on top of them. Only Load-bearing-blocks may have something on top of them. Alok Mehta - Programming in Lisp - Lecture 6

  23. Block-Support-For Returns Nil Returns value of Slot Support-For Basic-block Block-Support-For Hand Load-bearing-block Block-Support-For Movable-block Table Brick Wedge Ball Alok Mehta - Programming in Lisp - Lecture 6

  24. Put-On • Want a method, PUT-ON, that places one object on another • Get space for the object (may have to move things around) • Grasp object (may have to remove things on top of the object) • Move object • Ungrasp object • Basic Prototype • > (defmethod put-on ((object movable-block) • (support basic-block)) • … • ) Alok Mehta - Programming in Lisp - Lecture 6

  25. Put-On • Implementation • > (defmethod put-on ((object movable-block) • (support basic-block)) • (if (get-space object support) • (and (grasp object) • (move object support) • (ungrasp object)) • (format t "~&Couldn't get space to put ~a on ~a." • (block-name object) (block-name support)))) • get-space, grasp, move, ungrasp have yet to be defined • Get-space either finds space or makes space • > (defmethod get-space ((object movable-block) • (support basic-block)) • (or (find-space object support) • (make-space object support))) • find-space, make-space have yet to be defined (do this later) Alok Mehta - Programming in Lisp - Lecture 6

  26. Grasp • Grasp • Puts desired object in robot's hand • > (defmethod grasp ((object movable-block)) • (unless (eq (hand-grasping *hand*) object) ; already holding? • ; Make sure nothing else is on top of object • (when (block-support-for object) (clear-top object)) • ; Make sure robot isn't holding anything else • (when (hand-grasping *hand*) • (get-rid-of (hand-grasping *hand*))) • (format t "~&Move hand to pick up ~a at location ~a." • (block-name object) (top-location object)) • (setf (hand-position *hand*) (top-location object)) • (format t "~&Grasp ~a." (block-name object)) • (setf (hand-grasping *hand*) object)) • t) • Need to define: Clear-top, top-location • Function returns T if successful Alok Mehta - Programming in Lisp - Lecture 6

  27. Ungrasp • Ungrasp • Releases object, if object is being supported by something else • > (defmethod ungrasp ((object movable-block)) • (when (block-supported-by object) • (format t "~&Ungrasping ~a" (block-name object)) • (setf (hand-grasping *hand*) nil) • T)) • Get-rid-of • puts object on the table (out of the way) • > (defmethod get-rid-of ((object movable-block)) • (put-on object table)) Alok Mehta - Programming in Lisp - Lecture 6

  28. Make-space • Clear-top • Gets rid of everything on top of an object • > (defmethod clear-top ((support load-bearing-block)) • (dolist (obstacle (block-support-for support) t) • (get-rid-of obstacle))) • Make-space • Clears away just enough room to put block • Algorithm keeps getting rid of things on top of block • Until the find-space function returns that there is enough space available • > (defmethod make-space ((object movable-block) • (support basic-block)) • (dolist (obstruction (block-support-for support)) • (get-rid-of obstruction) • (let ((space (find-space object support))) • (when space (return space))))) Alok Mehta - Programming in Lisp - Lecture 6

  29. Move • Moves an object on top of a support • > (defmethod move ((object movable-block) • (support basic-block)) • (remove-support object) • (let ((newplace (get-space object support))) • (format t "~&Move ~a to top of ~a at location ~a." • (block-name object) (block-name support) newplace) • (setf (block-position object) newplace) • (setf (hand-position *hand*) (top-location object))) • (add-support object support) • t) • Calls remove-support, add-support • These are methods for bookkeeping • They maintain the bi-directional links (support-for, supported-by) • Remove-support removes the bi-directional links of an object • Add-support adds bi-directional links for an object that is to be placed on top of a support Alok Mehta - Programming in Lisp - Lecture 6

  30. Remove-support • Remove-support • This is a bookkeeping method that removes bi-directional links • > (defmethod remove-support ((object movable-block)) • (let ((support (block-supported-by object))) • (when support • (setf (block-support-for support) • (remove object (block-support-for support))) • (setf (block-supported-by object) nil) • t))) Alok Mehta - Programming in Lisp - Lecture 6

  31. Add-support • Adding support, for general case • This is just a stub (does no operation) • Can't really put an object on top of any basic-block • > (defmethod add-support ((object movable-block) • (support basic-block)) • t) • Adding support, for specific cases • For load-bearing blocks, we can put an object on top • Bookkeeping needed for this is to update bi-directional pointers • > (defmethod add-support ((object movable-block) • (support load-bearing-block)) • (push object (block-support-for support)) • (setf (block-supported-by object support))) Alok Mehta - Programming in Lisp - Lecture 6

  32. Top-location • Returns the location of the top (center) of a block • Example Usage • > (block-position b3) • (4 0) <-- Lower left position of block • > (block-width b3) • 4 • > (block-height b3) • 4 • > (top-location b3) • (6 4) <-- X = Pos.X + (width/2.0); Y = Pos.Y • Implementation • > (defmethod top-location ((object basic-block)) • (list (+ (first (block-position object)) • (/ (block-width object) 2)) • (+ (second (block-position object)) • (block-height object)))) Alok Mehta - Programming in Lisp - Lecture 6

  33. Find-space B4 B4 • Find-space • Basic algorithm • For each possible location on top of support that the object can be placed, • Is the position occupied? If not, we've found the space; else continue • > (defmethod find-space ((object basic-block) • (support basic-block)) • (dotimes (offset (+ 1 (- (block-width support) • (block-width object)))) • (unless (intersections-p object offset • (first (block-position support)) • (block-support-for support)) • (return (list (+ offset • (first (block-position support))) (+ (second (block-position support)) • (block-height support))))))) • Intersections-p is to be defined next B6 B6 Alok Mehta - Programming in Lisp - Lecture 6

  34. Intersections-p • Intersections-p • Checks for intersections (only checks the X dimension) • > (defun intersections-p (object offset base obstacles) • (dolist (obstacle obstacles) • (let* ((ls_proposed (+ offset base)) • (rs_proposed • (+ ls_proposed (block-width object))) • (ls_obstacle (first (block-position obstacle))) • (rs_obstacle • (+ ls_obstacle (block-width obstacle)))) • (unless (or (>= ls_proposed rs_obstacle) • (<= rs_Proposed ls_obstacle)) • (return t))))) Proposed ObjectLocation Obstacle Alok Mehta - Programming in Lisp - Lecture 6

  35. Blocks World Usage • Sample usage of blocks world • > (put-on b4 b1) • > (put-on w7 b2) • > (put-on b1 b2) Alok Mehta - Programming in Lisp - Lecture 6

  36. Final Exam • Next time • Open book, open notes • No calculators, computers, etc. • No lisp interpreter! • Mostly programming type questions • Write a program to … • Exam may contain material from • Chapters 1-12, plus case studies: Expert Systems, Blocks World • Anything from lecture notes Alok Mehta - Programming in Lisp - Lecture 6

More Related