fix: Use binary(16) for UUID type to ensure Spark compatibility

pyiceberg/io/pyarrow.py

@@ -789,7 +789,11 @@ def visit_string(self, _: StringType) -> pa.DataType:
         return pa.large_string()
 
     def visit_uuid(self, _: UUIDType) -> pa.DataType:
-        return pa.uuid()
+        # TODO: Change to uuid when PyArrow implements filtering for UUID types
+        # Using binary(16) instead of pa.uuid() because filtering is not
+        # implemented for UUID types in PyArrow
+        # (context: https://github.com/apache/iceberg-python/issues/2372)
+        return pa.binary(16)
 
     def visit_unknown(self, _: UnknownType) -> pa.DataType:
         """Type `UnknownType` can be promoted to any primitive type in V3+ tables per the Iceberg spec."""

tests/integration/test_add_files.py

@@ -828,7 +828,7 @@ def test_add_files_with_valid_upcast(
                 pa.field("list", pa.list_(pa.int64()), nullable=False),
                 pa.field("map", pa.map_(pa.string(), pa.int64()), nullable=False),
                 pa.field("double", pa.float64(), nullable=True),
-                pa.field("uuid", pa.uuid(), nullable=True),
+                pa.field("uuid", pa.binary(16), nullable=True),
             )
         )
     )

tests/integration/test_reads.py

@@ -610,15 +610,15 @@ def test_partitioned_tables(catalog: Catalog) -> None:
 def test_unpartitioned_uuid_table(catalog: Catalog) -> None:
     unpartitioned_uuid = catalog.load_table("default.test_uuid_and_fixed_unpartitioned")
     arrow_table_eq = unpartitioned_uuid.scan(row_filter="uuid_col == '102cb62f-e6f8-4eb0-9973-d9b012ff0967'").to_arrow()
-    assert arrow_table_eq["uuid_col"].to_pylist() == [uuid.UUID("102cb62f-e6f8-4eb0-9973-d9b012ff0967")]
+    assert arrow_table_eq["uuid_col"].to_pylist() == [uuid.UUID("102cb62f-e6f8-4eb0-9973-d9b012ff0967").bytes]
 
     arrow_table_neq = unpartitioned_uuid.scan(
         row_filter="uuid_col != '102cb62f-e6f8-4eb0-9973-d9b012ff0967' and uuid_col != '639cccce-c9d2-494a-a78c-278ab234f024'"
     ).to_arrow()
     assert arrow_table_neq["uuid_col"].to_pylist() == [
-        uuid.UUID("ec33e4b2-a834-4cc3-8c4a-a1d3bfc2f226"),
-        uuid.UUID("c1b0d8e0-0b0e-4b1e-9b0a-0e0b0d0c0a0b"),
-        uuid.UUID("923dae77-83d6-47cd-b4b0-d383e64ee57e"),
+        uuid.UUID("ec33e4b2-a834-4cc3-8c4a-a1d3bfc2f226").bytes,
+        uuid.UUID("c1b0d8e0-0b0e-4b1e-9b0a-0e0b0d0c0a0b").bytes,
+        uuid.UUID("923dae77-83d6-47cd-b4b0-d383e64ee57e").bytes,
     ]
 
 

tests/integration/test_writes/test_writes.py

@@ -1500,7 +1500,7 @@ def test_table_write_schema_with_valid_upcast(
                 pa.field("list", pa.list_(pa.int64()), nullable=False),
                 pa.field("map", pa.map_(pa.string(), pa.int64()), nullable=False),
                 pa.field("double", pa.float64(), nullable=True),  # can support upcasting float to double
-                pa.field("uuid", pa.uuid(), nullable=True),
+                pa.field("uuid", pa.binary(16), nullable=True),
             )
         )
     )
@@ -2138,7 +2138,7 @@ def test_uuid_partitioning(session_catalog: Catalog, spark: SparkSession, transf
     tbl.append(arr_table)
 
     lhs = [r[0] for r in spark.table(identifier).collect()]
-    rhs = [str(u.as_py()) for u in tbl.scan().to_arrow()["uuid"].combine_chunks()]
+    rhs = [str(uuid.UUID(bytes=u.as_py())) for u in tbl.scan().to_arrow()["uuid"].combine_chunks()]
     assert lhs == rhs
 
 
@@ -2530,3 +2530,63 @@ def test_v3_write_and_read_row_lineage(spark: SparkSession, session_catalog: Cat
     assert tbl.metadata.next_row_id == initial_next_row_id + len(test_data), (
         "Expected next_row_id to be incremented by the number of added rows"
     )
+
+
+@pytest.mark.integration
+def test_write_uuid_in_pyiceberg_and_scan(session_catalog: Catalog, spark: SparkSession) -> None:
+    """Test UUID compatibility between PyIceberg and Spark.
+
+    UUIDs must be written as binary(16) for Spark compatibility since Java Arrow
+    metadata differs from Python Arrow metadata for UUID types.
+    """
+    identifier = "default.test_write_uuid_in_pyiceberg_and_scan"
+
+    catalog = load_catalog("default", type="in-memory")
+    catalog.create_namespace("ns")
+
+    schema = Schema(NestedField(field_id=1, name="uuid_col", field_type=UUIDType(), required=False))
+
+    test_data_with_null = {
+        "uuid_col": [
+            uuid.UUID("00000000-0000-0000-0000-000000000000").bytes,
+            None,
+            uuid.UUID("11111111-1111-1111-1111-111111111111").bytes,
+        ]
+    }
+
+    try:
+        session_catalog.drop_table(identifier=identifier)
+    except NoSuchTableError:
+        pass
+
+    table = _create_table(session_catalog, identifier, {"format-version": "2"}, schema=schema)
+
+    arrow_table = pa.table(test_data_with_null, schema=schema.as_arrow())
+
+    # Write with pyarrow
+    table.append(arrow_table)
+
+    # Write with pyspark
+    spark.sql(
+        f"""
+        INSERT INTO {identifier} VALUES ("22222222-2222-2222-2222-222222222222")
+        """
+    )
+    df = spark.table(identifier)
+
+    table.refresh()
+
+    assert df.count() == 4
+    assert len(table.scan().to_arrow()) == 4
+
+    result = df.where("uuid_col = '00000000-0000-0000-0000-000000000000'")
+    assert result.count() == 1
+
+    result = df.where("uuid_col = '22222222-2222-2222-2222-222222222222'")
+    assert result.count() == 1
+
+    result = table.scan(row_filter=EqualTo("uuid_col", uuid.UUID("00000000-0000-0000-0000-000000000000").bytes)).to_arrow()
+    assert len(result) == 1
+
+    result = table.scan(row_filter=EqualTo("uuid_col", uuid.UUID("22222222-2222-2222-2222-222222222222").bytes)).to_arrow()
+    assert len(result) == 1