Skip to main content

C'est la Z

Working code is better than clever code (AOC 2021 day 4)

I always tell my students that the cleverest program is worthless if it doesn't actually work.

There are always some kids in class that all too often try to write the fanciest solutions.

They're the ones that write

int l(char *s){return !*s?0:(l(++s)+1);}

instead of something like:

int string_length(char *s){
  int i = 0;
  while (s[i] != 0){
    i=i+1;
    }
  return i;
}

to calculate the length of a string.

They try to code in a way they envision a master programmer coding which invariably leads them to hours of heartache and misery as they try to fix a poorly and complexly designed system. When I was a kid I did this as well but learned better early on. One thing I was pretty proud of as a young professional programmer were the number of comments I'd get from peers and superiors on the readability of my work.

Of course, my view isn't unique. Much more famously, Donald Knuth said the same when he quipped that "premature optimization is the root of all evil."

So while today's advent of code problem presented a number of design choices, I decided to go lazy and simple. Sure, my solution's probably slower than others but it's easy and understandable.

Today's problem had you playing bingo against a giant squid (full problem here). The core was that after a number of numbers were drawn, each time you crossing that number off your boards, you had to figure out if a board was a winner.

First the input. It was easy enough to read in the file. The first line were the bingo balls in the order they were to be drawn and then you had a bunch of 5 line blocks, each line with 5 numbers representing a 5x5 board. Each board was separated by a space.

My strategy was to read in all the boards into a list, then as I draw each bingo ball, replace all occurrences of the number drawn in on the boards with an X. Then, check to see if we have 5 X's in a row either vertically or horizontally.

Reading was easy but there were some decisions to be made. I was going to have a list of boards but what should a board look like? I didn't want to use a 2D array since that's not Clojure's strong suit but I could easily use a list of lists (or vector of vectors).

((22 13 17 11  0)
 ( 8  2 23  4 24)
 (21  9 14 16  7)
 (6 10  3 18  5)
 (1 12 20 15 19))

If I were coding in Python, I could do this as well using a list of lists.

This representation makes it very easy to check to see if we have a winner going across but checking vertically is a little annoying and doing the substitutions is also annoying.

On the other hand, I could represent the board as a list of 25 items and use a mapping function to get to a specific row,col location such as index = board[row*5+col] . This representation makes checking the board for winners a little onerous but marking squares with an X becomes trivial - just replace all occurrences of the number in question with an X. This can be done in Clojure using map:

(defn add-mark [board mark]
  (map #(if (= % mark) "X" %) board))

Thinking about it, it was really easy to switch from the list of 25 items to the list of list representations. We can do this with the Clojure partition function.

    (partition 5 '(0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 24 24))

    ;; results in:
    ;;((0 1 2 3 4)
    ;; (5 6 7 8 9 )
    ;; (10 11 12 13 14)
    ;; (15 16 17 18 19)
    ;; (20 21 22 23 24))

So, we can use the 25 element list representation to mark the boxes and then switch to the list of lists to check for a winner. What's better, we can transpose the list of lists, that is switch the rows and columns so as to make the vertical check into a horizontal one.

Here's my code to do all the checking:

(defn has-five? [board]
  (>  (->> board
           (map #(filter (fn [x] (= x "X"))%)  ) ;; keep only the Xs
           (map count) ;; how many are there in each row
           (filter #(= 5 % )) ;; keep the winners (5 in a row)
           count) 0)) ;; return true if we have any winners



(defn check-board [board]
  (let [horiz (partition 5 board) ;; convert 1D board to a 2D one
        vert  (apply mapv vector horiz)] ;; transpose 
    (or  (has-five? horiz)
         (has-five? vert))))

Now, solving part 1 was easy:

  1. take the next ball
  2. mark all the locations with the ball's number
  3. see if there's a winner
  4. If not, go back to step 1.

Sure, we're doing more transformations on the data than needed but it's clean and easy and at least from my point of view, made the whole exercise much less error prone.

On to part 2.

Turns out that was also pretty simple. Instead of stopping at the first winning board we had to find the last winning board.

Pretty easy. We just had to modify the above procedure to the following:

  1. take the next ball
  2. mark all the locations wiht the ball's number
  3. remove any board that's a winner
  4. If we're down to one board remaining we've found the board
  5. If not, back to step 1

I did this but first got a wrong answer. After looking at my results I realized that I stopped when I got down to 1 board but it hadn't necessarily won yet. I had to then continue drawing balls and marking squares until that final board was solved.

That was it.

Upon later reflection, I realized my solution wasn't perfect. It would fail if there were some boards that were never solvable. This though is an easy fix.

So, maybe not a super clever super fast solution but a clean one that works and is easy to understand and modify. To me that's the definition of a great solution.

comments powered by Disqus