feat(#3): prioritize on filename coverage

This further improves the ranking algorithm by prioritizing a match by how much of the name it matched. As a concrete example, assume that ./nested/test.txt and ./my_test.txt exist. Previous to this commit, a search for "test" would rank my_test.txt over test.txt because the rank would be the same, so the shorter path would be shown. With this change, test.txt would have a better rank because the query "test" matches a greater portion of the filename. This prioritization is only applied to filename matching for now, but maybe it could be applied in general. Closes #3
natecraddock · Mar 4, 2022 · d31e705 · d31e705
1 parent ef63125
commit d31e705
Showing 1 changed file with 9 additions and 1 deletion.
diff --git a/src/filter.zig b/src/filter.zig
@@ -203,11 +203,19 @@ pub fn rankToken(
         } else break;
     }
 
-    // retry on the full string
     if (best_rank != null) {
         // was a filename match, give priority
         best_rank.? /= 2.0;
+
+        // how much of the token matched the filename?
+        if (token.len == name.?.len) {
+            best_rank.? /= 2.0;
+        } else {
+            const coverage = 1.0 - (@intToFloat(f64, token.len) / @intToFloat(f64, name.?.len));
+            best_rank.? *= coverage;
+        }
     } else {
+        // retry on the full string
         it = IndexIterator.init(str, token[0], smart_case);
         while (it.next()) |start_index| {
             if (scanToEnd(str, token[1..], start_index, smart_case)) |match| {