## Knight Rider

### December 2, 2011

Today’s solution is provided by Jos Koot:

`; Given a chessboard, find a path of knight jumps visiting all squares of the`

; board. A normal chessboard has 8 rows and 8 columns. Generalize the

; knight rider such as to acccept a square board of arbitrary size.

;

; When forming a path, for each element of the path found so far, the list of

; moves yet to be tried must be remembered (movelists).

;

; Make the path finder such as to return either a complete path or

; #f when no path is found. Keep trying other moves as long as no path is found.

; If from a given starting square no path is found, then try another starting

; square.

```
```

`(define (knight-rider boardsize)`

(define max-index (sub1 boardsize))

(define nr-of-fields (sqr boardsize))

; Rows and columns are repesented by numbers 0 up to but not including

; the boardsize. At top level procedure main is called as (main)

; Squares are represented as (row . col).

; Jumps are represented as (row-difference . col-difference)

(define (main) ; Try all possible starting squares.

(let loop ((row 0) (col 0))

(and (< row boardsize) ; If all starting squares have been tried, return #f.

(if (= col boardsize) (loop (add1 row) 0) ; Try starting squares in next row.

(or (find-path row col) (loop row (add1 col)))))))

(define (find-path row col)

(let ((start (cons row col))) ; Squares are represented as (row . col).

(occupy! start) ; Mark the starting square occupied.

(let path-loop

((path (list start)) ; Notice that the path will be built in reversed order.

(move-lists (list (get-sorted-moves start)))

(n 1))

; n is the length of the path found so far.

; Move-lists has as many elements as the path.

; Each move-list shows which jumps yet to try if the last jump fails.

(cond

((zero? n) #f) ; Return #f when no path is found from current start.

((= n nr-of-fields) (reverse path)) ; Path has been found.

((null? (car move-lists)) ; End of path has no more moves to try,

(clear! (car path)) ; hence backtrack by removing the last jump.

(path-loop (cdr path) (cdr move-lists) (sub1 n)))

((let ((field (caar move-lists))) ; Try the next move.

(occupy! field)

(let

((result

(path-loop

(cons field path) ; Add one move to the path.

(cons

(get-sorted-moves field) ; Why sorting? See get-sorted-moves.

; Below: remove the move just made from its move-list.

(cons (cdar move-lists) (cdr move-lists)))

(add1 n))))

(or result (begin (clear! field) #f)))))

(else ; Previous move did not work, hence try next move.

(path-loop path (cons (cdar move-lists) (cdr move-lists)) n))))))

(define jumps '((1 . 2) (-1 . 2) (1 . -2) (-1 . -2)

(2 . 1) (-2 . 1) (2 . -1) (-2 . -1)))

; Procedure add-jump computes the squares reached when making a jump from a

; given square. It returns #f if the jump cannot be made from the square.

(define (add-jump field jump)

(let

((x (+ (car field) (car jump)))

(y (+ (cdr field) (cdr jump))))

(and (<= 0 x max-index) (<= 0 y max-index) (cons x y))))

; Procedure get-moves returns all non occupied squares we can jump to

; from a given square.

(define (get-moves field)

(filter

(lambda (move) (and move (board-ref move)))

(map (lambda (jump) (add-jump field jump)) jumps)))

; Procedure get-sorted-moves provides the following heuristic:

; The feasible moves from the last square of the path

; (id est (car path)) are sorted such as to try those moves first

; that allow as few remaining moves as possible.

; This heuristic speeds up very much. It works well for boardsize < 35.

; For some boardsizes greater than 34, the speedup may not work.

(define (get-sorted-moves p)

(sort

(lambda (p q) (< (length (get-moves p)) (length (get-moves q))))

(get-moves p)))

; The board is a boardsize x boardsize matrix.

; At all stages the board shows #f for occupied squares and #t for free squares.

(define board (make-matrix boardsize boardsize #t))

(define (occupy! field) (matrix-set! board (car field) (cdr field) #f))

(define (clear! field) (matrix-set! board (car field) (cdr field) #t))

(define (board-ref field) (matrix-ref board (car field) (cdr field)))

(main))

On an 8-by-8 chessboard, we get this solution:

`< (knight-rider 8)`

((0 . 0) (1 . 2) (0 . 4) (1 . 6) (3 . 7) (5 . 6) (7 . 7) (6 . 5)

(5 . 7) (7 . 6) (6 . 4) (7 . 2) (6 . 0) (4 . 1) (2 . 0) (0 . 1)

(1 . 3) (0 . 5) (1 . 7) (2 . 5) (0 . 6) (2 . 7) (4 . 6) (6 . 7)

(7 . 5) (6 . 3) (7 . 1) (5 . 0) (6 . 2) (7 . 0) (5 . 1) (3 . 0)

(1 . 1) (0 . 3) (1 . 5) (0 . 7) (2 . 6) (4 . 7) (6 . 6) (7 . 4)

(5 . 5) (3 . 6) (4 . 4) (3 . 2) (2 . 4) (4 . 5) (5 . 3) (3 . 4)

(2 . 2) (1 . 0) (0 . 2) (1 . 4) (3 . 5) (4 . 3) (3 . 1) (2 . 3)

(4 . 2) (5 . 4) (7 . 3) (6 . 1) (4 . 0) (5 . 2) (3 . 3) (2 . 1))

Jos used `when`

, `add1`

, `sub1`

, `sqr`

, `sort`

, `make-matrix`

, `matrix-ref`

, `matrix-set!`

, `any?`

and `filter`

from the Standard Prelude. You can run the program at http://programmingpraxis.codepad.org/XQEm0Kc6.

Pages: 1 2

You can find a somewhat more elegant version on http://www.telefonica.net/web2/koot/knight-rider-0.rkt. It requires DrRacket though. In this version the path-finder is abstracted from the actual problem. It could also be used for other computations requiring backtracking. It should not be too difficult to adapt the code to programming praxis.

Best wishes, Jos Koot