temp for part 2

p21/p21.org

@@ -87,6 +87,8 @@ go from p to q without getting out of the boundary
 #+RESULTS:
 : path-allowed-p
 
+The following of course only works for part 1; with part 2 the stack
+would blow in my face since we would have too many options to check.
 #+begin_src emacs-lisp
    (defun directions-for-string (s keypad-data)
     (--map (connecting-paths it keypad-data)
@@ -99,6 +101,7 @@ go from p to q without getting out of the boundary
     (-reduce (lambda (a b)
                (if (> (length a) (length b)) b a))
              l))
+
   (defun shortest-path (s)
    (let* ((first-directional-keypad
            (directions-for-string s numeric-keypad-data))
@@ -147,3 +150,17 @@ go from p to q without getting out of the boundary
 
 #+RESULTS:
 : 222670
+
+
+The new idea would be to run things bottom-down. For a given pair of
+chars at a given depth, what is the optimal way to put things
+together? It should be true that the optimal way to assemble things is
+the shortest concatenation of the optimal way at the previous depth
+
+So we start from all possible combinations of keypad pairs
+#+begin_src emacs-lisp
+  (setq direction-pairs  (let ((directions (advent/split-string-into-char-list "<>^vA")))
+                           (-mapcat (lambda (x) (--map (cons x it) directions)) directions)))
+
+  (--map (connecting-paths (char-pairs-to-coordinate-pairs it dir-keypad-data) dir-keypad-data) direction-pairs)
+#+end_src