create dedicated Rating type to replace awkward (bool,i32) pair

fish/src/bot.rs

@@ -14,9 +14,8 @@ mod trans;
 
 use crate::bot::node::{Node, RawNodePtr};
 use crate::bot::trans::TransTable;
-use crate::eval::{features, Features, Weights};
-
-use bumpalo::Bump as Arena;
+use crate::eval::{features, Features, Rating, Weights};
+use crate::Arena;
 
 /// Encompasses an instance of the algorithm.
 pub struct Bot {
@@ -35,7 +34,7 @@ pub struct Metrics {
     pub start_heuristic: i32,
     pub end_features: Features,
     pub end_heuristic: i32,
-    pub end_rating: (bool, i32),
+    pub end_rating: Rating,
     pub end_iteration: u32,
     // TODO(?) memory usage metrics
     // TODO(?) transposition table metrics
@@ -127,7 +126,7 @@ impl Bot {
 
 struct Evaluator {
     weights: Weights,
-    root_score: i32,
+    root_heuristic: i32,
     root_queue_len: usize,
 }
 
@@ -135,30 +134,26 @@ impl Evaluator {
     fn new(weights: &Weights, root: &Node) -> Self {
         Self {
             weights: *weights,
-            root_score: features(root.matrix(), 0).evaluate(weights),
+            root_heuristic: features(root.matrix(), 0).evaluate(weights),
             root_queue_len: root.queue().len(),
         }
     }
 
-    fn evaluate(&self, mat: &Mat, queue: Queue<'_>, cleared: &Range<i16>) -> (bool, i32) {
-        let pcnt = self.root_queue_len.saturating_sub(queue.len());
+    fn evaluate(&self, mat: &Mat, queue: Queue<'_>, cleared: &Range<i16>) -> Rating {
+        debug_assert!(queue.len() < self.root_queue_len);
+        let pcnt = self.root_queue_len - queue.len();
 
-        if self.greed() && cleared.contains(&0) {
-            // cleared the bottom row of the matrix, which must be the last line of cheese
-            // in the race. piece count is negated so that less pieces is better (larger
-            // value).
-            return (true, -(pcnt as i32));
+        // check if we cleared the bottom row of the matrix, which is assumed to be the
+        // last line of cheese in the race. if so, then consider this node to be a solve
+        // and use the piece count as its rating.
+        //
+        // TODO: make this condition configurable since its a bit of a hack
+        if cleared.contains(&0) {
+            return Rating::Solve(pcnt as u32);
         }
 
-        let score = features(mat, pcnt).evaluate(&self.weights);
-
-        // larger (further below the root score) is better
-        (false, self.root_score - score)
-    }
-
-    fn greed(&self) -> bool {
-        // TODO: make this parameter configurable on `Bot` initialization
-        true
+        let heuristic = features(mat, pcnt).evaluate(&self.weights);
+        Rating::Score(self.root_heuristic - heuristic)
     }
 }
 
@@ -269,8 +264,7 @@ impl SegmentedAStar {
         let cand = open_set.pop().ok_or(None)?;
         let cand = unsafe { cand.0.as_node() };
 
-        if cand.queue().is_empty() || cand.rating().0 {
-            // terminal node; end search and back up its rating
+        if cand.is_terminal() {
             return Err(Some(cand));
         }
 

fish/src/bot/node.rs

@@ -6,8 +6,9 @@ use mino::matrix::{Mat, MatBuf};
 use mino::srs::{Piece, PieceType, Queue};
 
 use crate::bot::trans::TransTable;
-use crate::bot::Arena;
+use crate::eval::Rating;
 use crate::find::find_locations;
+use crate::Arena;
 
 /// Represents a node in the search tree. A node basically just consists of a board state
 /// (incl. queue) and some extra metadata relating it to previous nodes in the tree.
@@ -15,7 +16,7 @@ pub struct Node {
     matrix: *const Mat,
     queue: RawQueue,
     edge: Option<Edge>,
-    rating: (bool, i32),
+    rating: Rating,
     // currently there is no need to store a node's children, but maybe this could change
     // in the future.
 }
@@ -40,8 +41,7 @@ impl Node {
     pub fn alloc_root<'a>(arena: &'a Arena, matrix: &Mat, queue: Queue<'_>) -> &'a Self {
         let matrix = copy_matrix(arena, matrix);
         let queue = copy_queue(arena, queue);
-        let rating = (false, i32::MIN);
-        Node::alloc(arena, matrix, queue, rating, None)
+        Node::alloc(arena, matrix, queue, Rating::default(), None)
     }
 
     // `matrix` and `queue` must be allocated inside `arena`
@@ -49,7 +49,7 @@ impl Node {
         arena: &'a Arena,
         matrix: &'a Mat,
         queue: Queue<'a>,
-        rating: (bool, i32),
+        rating: Rating,
         edge: Option<Edge>,
     ) -> &'a Self {
         let matrix = matrix as *const Mat;
@@ -70,10 +70,14 @@ impl Node {
         unsafe { self.queue.as_queue() }
     }
 
-    pub fn rating(&self) -> (bool, i32) {
+    pub fn rating(&self) -> Rating {
         self.rating
     }
 
+    pub fn is_terminal(&self) -> bool {
+        matches!(self.rating, Rating::Solve(_)) || self.queue().is_empty()
+    }
+
     /// Get the initial placement made after the root node which eventually arrives at
     /// this node.
     pub fn root_placement(&self) -> Option<Piece> {
@@ -97,7 +101,7 @@ impl Node {
         mut evaluate: E,
     ) -> impl Iterator<Item = &'a Node> + 'a
     where
-        E: FnMut(&Mat, Queue<'_>, &Range<i16>) -> (bool, i32) + 'a,
+        E: FnMut(&Mat, Queue<'_>, &Range<i16>) -> Rating + 'a,
     {
         let mut matrix = MatBuf::new();
 
@@ -147,9 +151,9 @@ impl core::fmt::Debug for Node {
     fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
         write!(f, "Node {{ ")?;
         match self.rating {
-            (false, h) => write!(f, "rating: {}", h),
-            (true, n) => write!(f, "solution: {}", -n),
-        }?;
+            Rating::Score(v) => write!(f, "score: {}", v)?,
+            Rating::Solve(n) => write!(f, "solve: {}", n)?,
+        }
         if let Some(pc) = self.root_placement() {
             write!(f, ", root_placement: {:?}", pc)?;
         }

fish/src/eval.rs

@@ -65,3 +65,41 @@ impl Features {
         score
     }
 }
+
+/// Solutions can have two types of ratings depending on if they are considered to be a
+/// "solve", which means they are rated by how few pieces are required, or they are not a
+/// solve, so they are rated by their "score", which is measured as the difference between
+/// the evaluation of a terminal node and the evaluation of the root.
+#[derive(Debug, Copy, Clone, Eq, PartialEq)]
+pub enum Rating {
+    Score(i32),
+    Solve(u32),
+}
+
+impl core::cmp::Ord for Rating {
+    fn cmp(&self, other: &Self) -> core::cmp::Ordering {
+        use self::Rating::*;
+        use core::cmp::Ordering::*;
+
+        match (self, other) {
+            (Score(x), Score(y)) => x.cmp(y), // greater (difference from root) is better
+            (Solve(x), Solve(y)) => y.cmp(x), // less (piece count) is better
+            (Solve(_), _) => Greater,         // solve always better than non-solve
+            (_, Solve(_)) => Less,
+        }
+    }
+}
+
+impl core::cmp::PartialOrd for Rating {
+    fn partial_cmp(&self, other: &Self) -> Option<core::cmp::Ordering> {
+        Some(self.cmp(other))
+    }
+}
+
+impl core::default::Default for Rating {
+    /// The default rating is the worst possible rating which is ranked worse than any
+    /// other rating besides itself.
+    fn default() -> Self {
+        Rating::Score(i32::MIN)
+    }
+}

fish/src/lib.rs

@@ -9,3 +9,4 @@ pub mod find;
 type HashMap<K, V> = hashbrown::HashMap<K, V, HashBuilder>;
 type HashSet<T> = hashbrown::HashSet<T, HashBuilder>;
 type HashBuilder = core::hash::BuildHasherDefault<ahash::AHasher>;
+type Arena = bumpalo::Bump;

tidepool/src/output.rs

@@ -1,4 +1,5 @@
 use fish::bot::Metrics;
+use fish::eval::Rating;
 use mino::srs::Piece;
 use serde::Serialize;
 
@@ -184,8 +185,9 @@ mod ser {
         #[serde(untagged)]
         enum RatingVariant {
             NotApplicable,
+            // confusing: 'rating' is called 'score' actually
             Rating { rating: i32 },
-            Solution { solution: i32 },
+            Solution { solution: u32 },
         }
 
         metrics
@@ -202,8 +204,8 @@ mod ser {
                     features: m.end_features.1.to_vec(),
                     heuristic: m.end_heuristic,
                     rating: match m.end_rating {
-                        (false, v) => RatingVariant::Rating { rating: v },
-                        (true, v) => RatingVariant::Solution { solution: -v },
+                        Rating::Score(v) => RatingVariant::Rating { rating: v },
+                        Rating::Solve(n) => RatingVariant::Solution { solution: n },
                     },
                     iteration: Some(m.end_iteration),
                 },