Add unit tests for sparse bitmap keys and the Java key range limit

Author: moomindani

tests/table/test_puffin.py

@@ -120,6 +120,18 @@ def test_puffin_round_trip(tmp_path: Path) -> None:
     assert read_vectors[file_path].to_pylist() == sorted(set(deletions))
 
 
+def test_puffin_round_trip_with_sparse_bitmap_keys(tmp_path: Path) -> None:
+    # High bits 0 and 2 are present while 1 is absent; the writer must emit sorted keys
+    # and the reader pads the missing key with an empty bitmap.
+    positions = [3, (2 << 32) + 4]
+    writer, puffin_path = _new_writer(tmp_path)
+    writer.set_blob(positions=positions, referenced_data_file="file.parquet")
+    writer.finish()
+
+    vectors = PuffinFile(puffin_path.read_bytes()).to_vector()
+    assert vectors["file.parquet"].to_pylist() == positions
+
+
 def test_write_and_read_puffin_file(tmp_path: Path) -> None:
     writer, puffin_path = _new_writer(tmp_path)
     writer.set_blob(positions=[1, 2, 3], referenced_data_file="file1.parquet")
@@ -203,3 +215,9 @@ def test_set_blob_rejects_empty_positions(tmp_path: Path) -> None:
     writer, _ = _new_writer(tmp_path)
     with pytest.raises(ValueError, match="Deletion vector must contain at least one position"):
         writer.set_blob(positions=[], referenced_data_file="file.parquet")
+
+
+def test_set_blob_rejects_position_exceeding_java_key_range(tmp_path: Path) -> None:
+    writer, _ = _new_writer(tmp_path)
+    with pytest.raises(ValueError, match="Key 2147483648 is too large, max 2147483647"):
+        writer.set_blob(positions=[(2**31) << 32], referenced_data_file="file.parquet")