Day 7: Camel Cards
Picat: Part 1 (01:12:56, rank 8121), Part 2 (01:34:54, rank 6663)
This was a fun little puzzle. After deciding how to process and represent the input, I ran into a small issue in the Picat stdlib: there doesn't seem to be a sort_by function that takes a comparator function -- there are options for sorting based on a "key", but I couldn't think of a sensible way to convert each hand to an absolute value that's fully orderable. Instead, I implemented the last-ditch panic sort we all know and love: bubble sort.
I also had great difficulty with the tie_break predicate. For whatever reason, it only works when I use Horn clauses (head :- body) in Picat; the usual predicate notation (head => body) produces the wrong result. Maybe if I read those early chapters of the Picat manual I'll understand why...
A cool twist on the puzzle. I decided to take the opportunity to implement an equivalent to length <$> (group . sort) in Haskell (this is the kind of situation where Haskell is really handy for manipulating lists). There's no group function in Picat, so I used a foreach to store a count of each element in a map, then return the down-sorted counts.
This was also an opportunity to replace the exploding pattern matches of part 2's get_type function, transforming from cases like this:
get_type([A,A,A,_,_]) = 4.
get_type([_,A,A,A,_]) = 4.
get_type([_,_,A,A,A]) = 4.
...to this:
get_type([X|_],Js) = R, X+Js == 3 => R = 4.
The cases with two pairs, or one pair and one three, were a bit uglier, but still nicer.
2.866s
6.162s