Replace HashMaps with a bit-vector for unique depth computation (#201)

Author: sampsyo

flatgfa/Cargo.lock

@@ -10,6 +10,7 @@ path = "src/cli/main.rs"
 [dependencies]
 argh = "0.1.12"
 atoi = "2.0.0"
+bit-vec = "0.8.0"
 bstr = "1.10.0"
 memchr = "2.7.4"
 memmap = "0.7.0"

flatgfa/Cargo.toml

@@ -204,7 +204,7 @@ pub fn depth(gfa: &flatgfa::FlatGFA) {
             "{}\t{}\t{}",
             name,
             depths[id.index()],
-            uniq_paths[id.index()].len()
+            uniq_paths[id.index()],
         );
     }
 }

flatgfa/src/cli/cmds.rs

@@ -1,22 +1,35 @@
 use crate::flatgfa;
-use std::collections::HashSet;
+use bit_vec::BitVec;
 
-pub fn depth(gfa: &flatgfa::FlatGFA) -> (Vec<usize>, Vec<HashSet<usize>>) {
-    // Initialize node depth
+/// Compute the *depth* of each segment in the variation graph.
+///
+/// The depth is defined to be the number of times that a path traverses a given
+/// segment. We return two values: the ordinary depth and the *unique* depth,
+/// which only counts each path that tarverses a given segment once.
+///
+/// Both outputs are depth values indexed by segment ID.
+pub fn depth(gfa: &flatgfa::FlatGFA) -> (Vec<usize>, Vec<usize>) {
+    // Our output vectors: the ordinary and unique depths of each segment.
     let mut depths = vec![0; gfa.segs.len()];
-    // Initialize uniq_paths
-    let mut uniq_paths = Vec::<HashSet<usize>>::new();
-    uniq_paths.resize(gfa.segs.len(), HashSet::new());
-    // do not assume that each handle in `gfa.steps()` is unique
-    for (idx, path) in gfa.paths.all().iter().enumerate() {
+    let mut uniq_depths = vec![0; gfa.segs.len()];
+
+    // This bit vector keeps track of whether the current path has already
+    // traversed a given segment, and therefore whether we should ignore
+    // subsequent traversals (for the purpose of counting unique depth).
+    let mut seen = BitVec::from_elem(gfa.segs.len(), false);
+
+    for path in gfa.paths.all().iter() {
+        seen.clear(); // All segments are unseen.
         for step in &gfa.steps[path.steps] {
             let seg_id = step.segment().index();
-            // Increment depths
             depths[seg_id] += 1;
-            // Update uniq_paths
-            uniq_paths[seg_id].insert(idx);
+            if seen[seg_id] {
+                // The first traversal of this path over this segment.
+                uniq_depths[seg_id] += 1;
+                seen.set(seg_id, true);
+            }
         }
     }
 
-    (depths, uniq_paths)
+    (depths, uniq_depths)
 }