Implement Huffman encoding

src/huffman.rs

@@ -1,5 +1,69 @@
+use std::cmp::max;
+use std::cmp::min;
 
+#[derive(Debug)]
 pub struct HuffmanAlphabet {
-	pub length: u32,
+    pub length: usize,
-	pub frequencies: Vec<u32>,
+    pub frequencies: Vec<u64>,
-}
+}
+
+#[derive(Debug)]
+struct HuffmanTreeNode {
+    frequency: u64,
+    left: Option<Box<HuffmanTreeNode>>,
+    right: Option<Box<HuffmanTreeNode>>,
+}
+
+fn max_depth(tree: &HuffmanTreeNode) -> usize {
+    let mut depth_left = 0;
+    if let Some(node_left) = &tree.left {
+        depth_left = 1 + max_depth(&node_left);
+    }
+
+    let mut depth_right = 0;
+    if let Some(node_right) = &tree.right {
+        depth_right = 1 + max_depth(&node_right);
+    }
+
+    max(depth_left, depth_right)
+}
+
+fn min_depth(tree: &HuffmanTreeNode) -> usize {
+    let mut depth_left = 0;
+    if let Some(node_left) = &tree.left {
+        depth_left = 1 + min_depth(&node_left);
+    }
+
+    let mut depth_right = 0;
+    if let Some(node_right) = &tree.right {
+        depth_right = 1 + min_depth(&node_right);
+    }
+
+    min(depth_left, depth_right)
+}
+
+pub fn build_huffman_tree(h: &HuffmanAlphabet) -> (usize, usize) {
+    let mut nodes: Vec<HuffmanTreeNode> = Vec::new();
+    for f in &h.frequencies {
+        let n = HuffmanTreeNode {
+            frequency: *f,
+            left: None,
+            right: None,
+        };
+        nodes.push(n);
+    }
+    nodes.sort_by(|a, b| b.frequency.cmp(&a.frequency));
+    while nodes.len() > 1 {
+        let a = nodes.pop().unwrap();
+        let b = nodes.pop().unwrap();
+        let n = HuffmanTreeNode {
+            frequency: a.frequency + b.frequency,
+            left: Some(Box::new(a)),
+            right: Some(Box::new(b)),
+        };
+        nodes.push(n);
+        nodes.sort_by(|a, b| b.frequency.cmp(&a.frequency));
+    }
+    let tree = nodes.pop().unwrap();
+    (min_depth(&tree), max_depth(&tree))
+}

src/k_clustering_big.rs

@@ -4,22 +4,21 @@ pub type ImpliciteGraph = Vec<u32>;
 fn neighbors(a: u32, n: usize) -> Vec<u32> {
     let mut r = vec![a];
     for _ in 0..n {
-    	let mut r_new = r.clone();
+        let mut r_new = r.clone();
-    	for a in r {
+        for a in r {
-		    for i in 0..24 {
+            for i in 0..24 {
-		        let m = 0x1 << i;
+                let m = 0x1 << i;
-		        let neighbor = a ^ m;
+                let neighbor = a ^ m;
-		        r_new.push(neighbor);
+                r_new.push(neighbor);
-		    }
+            }
-    	}
+        }
-    	r_new.sort();
+        r_new.sort();
-    	r_new.dedup();
+        r_new.dedup();
-    	r = r_new;
+        r = r_new;
     }
     r
 }
 
-
 pub fn k_clustering_big(g: &ImpliciteGraph) -> usize {
     let mut node_id_to_cluster_id: Vec<usize> = (0..g.len()).collect();
     let mut clusters: Vec<Vec<usize>> = (0..g.len()).map(|x| vec![x]).collect();
@@ -35,13 +34,13 @@ pub fn k_clustering_big(g: &ImpliciteGraph) -> usize {
     }
 
     for node_a_id in 0..g.len() {
-	    // Iterate over all nodes in the graph. Then, for each node compute all
+        // Iterate over all nodes in the graph. Then, for each node compute all
-	    // neighbors that are two or less bits away.
+        // neighbors that are two or less bits away.
         for node_b_value in neighbors(g[node_a_id], 2) {
-        	// See if there exist nodes that match the neighbor. If such nodes
+            // See if there exist nodes that match the neighbor. If such nodes
-        	// exist iterate over them and merge the clusters if they are not
+            // exist iterate over them and merge the clusters if they are not
-        	// already the same. The key insight is that we have to cluster all
+            // already the same. The key insight is that we have to cluster all
-        	// nodes that are two or less (that includes zero) bits apart.
+            // nodes that are two or less (that includes zero) bits apart.
             if let Some(node_b_ids) = node_map.get(&node_b_value) {
                 for node_b_id in node_b_ids {
                     let cluster_id_a = node_id_to_cluster_id[node_a_id];

src/main.rs

@@ -1,5 +1,6 @@
 mod dijkstra;
 mod heap;
+mod huffman;
 mod jobs;
 mod k_clustering;
 mod k_clustering_big;
@@ -117,7 +118,9 @@ fn c3a2() {
 
 #[allow(dead_code)]
 fn c3a3() {
-	println!("continue here");
+    let h = util::read_huffman_alphabet("data/c3a3_huffman.txt").unwrap();
+    let r = huffman::build_huffman_tree(&h);
+    println!("r1 = {} r2 = {}", r.1, r.0);
 }
 
 fn main() {

src/util.rs

@@ -1,3 +1,4 @@
+use crate::huffman;
 use crate::jobs;
 use crate::k_clustering;
 use crate::k_clustering_big;
@@ -239,3 +240,23 @@ pub fn read_k_cluster_big(path: &str) -> Result<k_clustering_big::ImpliciteGraph
 
     Ok(g)
 }
+
+pub fn read_huffman_alphabet(path: &str) -> Result<huffman::HuffmanAlphabet, io::Error> {
+    let file = File::open(path)?;
+    let mut lines = BufReader::new(file).lines();
+    let line = lines.next().unwrap().unwrap();
+    let length = line.parse().unwrap();
+
+    let mut h = huffman::HuffmanAlphabet {
+        length: length,
+        frequencies: Vec::new(),
+    };
+
+    for line in lines {
+        let line = line?;
+        let frequency = line.parse().unwrap();
+        h.frequencies.push(frequency);
+    }
+    assert!(length == h.frequencies.len());
+    Ok(h)
+}