From 600694c8124c249747f33c1f42608bab9af999a8 Mon Sep 17 00:00:00 2001
From: "google-labs-jules[bot]"
 <161369871+google-labs-jules[bot]@users.noreply.github.com>
Date: Mon, 25 May 2026 18:27:16 +0000
Subject: [PATCH] perf: optimize _classify_entry membership check

Replaced the $O(N)$ `any()` generator check with $O(1)$ logic using
dictionary length. Since the background class "0" is checked
explicitly, the presence of other classes is equivalent to
len(counts) > 1 (if "0" is present) or len(counts) > 0 (if "0" is absent).

This provides a measurable speed boost in dataset splitting, especially
for manifests with many patches and multiple classes.

Co-authored-by: tahamukhtar20 <91777330+tahamukhtar20@users.noreply.github.com>
---
 src/mapcv/splitter.py | 10 +++-------
 1 file changed, 3 insertions(+), 7 deletions(-)

diff --git a/src/mapcv/splitter.py b/src/mapcv/splitter.py
index b187d3d..3f51d30 100644
--- a/src/mapcv/splitter.py
+++ b/src/mapcv/splitter.py
@@ -31,13 +31,9 @@ def _classify_entry(entry: ManifestEntry) -> int:
     """
     counts = entry["per_class_pixel_counts"]
     if counts:
-        has_bg = "0" in counts
-        has_labeled = any(k != "0" for k in counts)
-        if has_bg and not has_labeled:
-            return 0
-        if has_labeled and not has_bg:
-            return 1
-        return 2
+        if "0" in counts:
+            return 0 if len(counts) == 1 else 2
+        return 1
     if entry["empty_ratio"] >= 1.0:
         return 0
     if entry["empty_ratio"] <= 0.0: