fix: Unify output of str(DataType) for different pythons (#2635) (#2638)

See also: #2633
pr: #2635

Signed-off-by: yangxuan <[email protected]>

Author: XuanYang-cn

pymilvus/client/types.py

@@ -7,7 +7,7 @@
     ExceptionsMessage,
     InvalidConsistencyLevel,
 )
-from pymilvus.grpc_gen import common_pb2, rg_pb2
+from pymilvus.grpc_gen import common_pb2, rg_pb2, schema_pb2
 from pymilvus.grpc_gen import milvus_pb2 as milvus_types
 
 Status = TypeVar("Status")
@@ -84,29 +84,36 @@ def OK(self):
 
 
 class DataType(IntEnum):
-    NONE = 0
-    BOOL = 1
-    INT8 = 2
-    INT16 = 3
-    INT32 = 4
-    INT64 = 5
-
-    FLOAT = 10
-    DOUBLE = 11
-
-    STRING = 20
-    VARCHAR = 21
-    ARRAY = 22
-    JSON = 23
-
-    BINARY_VECTOR = 100
-    FLOAT_VECTOR = 101
-    FLOAT16_VECTOR = 102
-    BFLOAT16_VECTOR = 103
-    SPARSE_FLOAT_VECTOR = 104
+    """
+    String of DataType is str of its value, e.g.: str(DataType.BOOL) == "1"
+    """
+
+    NONE = 0  # schema_pb2.None, this is an invalid representation in python
+    BOOL = schema_pb2.Bool
+    INT8 = schema_pb2.Int8
+    INT16 = schema_pb2.Int16
+    INT32 = schema_pb2.Int32
+    INT64 = schema_pb2.Int64
+
+    FLOAT = schema_pb2.Float
+    DOUBLE = schema_pb2.Double
+
+    STRING = schema_pb2.String
+    VARCHAR = schema_pb2.VarChar
+    ARRAY = schema_pb2.Array
+    JSON = schema_pb2.JSON
+
+    BINARY_VECTOR = schema_pb2.BinaryVector
+    FLOAT_VECTOR = schema_pb2.FloatVector
+    FLOAT16_VECTOR = schema_pb2.Float16Vector
+    BFLOAT16_VECTOR = schema_pb2.BFloat16Vector
+    SPARSE_FLOAT_VECTOR = schema_pb2.SparseFloatVector
 
     UNKNOWN = 999
 
+    def __str__(self) -> str:
+        return str(self.value)
+
 
 class RangeType(IntEnum):
     LT = 0  # less than

tests/test_types.py

@@ -31,28 +31,37 @@
 #  from ml_dtypes import bfloat16
 
 
-@pytest.mark.skip("please fix me")
 class TestTypes:
-    @pytest.mark.parametrize("input_expect", [
-        ([1], DataType.FLOAT_VECTOR),
-        ([True], DataType.UNKNOWN),
-        ([1.0, 2.0], DataType.FLOAT_VECTOR),
-        (["abc"], DataType.UNKNOWN),
-        (bytes("abc", encoding='ascii'), DataType.BINARY_VECTOR),
-        (1, DataType.INT64),
-        (True, DataType.BOOL),
-        ("abc", DataType.VARCHAR),
-        (np.int8(1), DataType.INT8),
-        (np.int16(1), DataType.INT16),
-        ([np.int8(1)], DataType.FLOAT_VECTOR),
-        ([np.float16(1.0)], DataType.FLOAT16_VECTOR),
-        #  ([np.array([1, 1], dtype=bfloat16)], DataType.BFLOAT16_VECTOR),
-    ])
+    @pytest.mark.skip("please fix me")
+    @pytest.mark.parametrize(
+        "input_expect",
+        [
+            ([1], DataType.FLOAT_VECTOR),
+            ([True], DataType.UNKNOWN),
+            ([1.0, 2.0], DataType.FLOAT_VECTOR),
+            (["abc"], DataType.UNKNOWN),
+            (bytes("abc", encoding="ascii"), DataType.BINARY_VECTOR),
+            (1, DataType.INT64),
+            (True, DataType.BOOL),
+            ("abc", DataType.VARCHAR),
+            (np.int8(1), DataType.INT8),
+            (np.int16(1), DataType.INT16),
+            ([np.int8(1)], DataType.FLOAT_VECTOR),
+            ([np.float16(1.0)], DataType.FLOAT16_VECTOR),
+            #  ([np.array([1, 1], dtype=bfloat16)], DataType.BFLOAT16_VECTOR),
+        ],
+    )
     def test_infer_dtype_bydata(self, input_expect):
         data, expect = input_expect
         got = infer_dtype_bydata(data)
         assert got == expect
 
+    def test_str_of_data_type(self):
+        for v in DataType:
+            assert isinstance(v, DataType)
+            assert str(v) == str(v.value)
+            assert str(v) != v.name
+
 
 class TestConsistencyLevel:
     def test_consistency_level_int(self):
@@ -91,6 +100,8 @@ def test_shard(self):
     def test_shard_dup_nodeIDs(self):
         s = Shard("channel-1", (1, 1, 1), 1)
         assert s.channel_name == "channel-1"
-        assert s.shard_nodes == {1,}
+        assert s.shard_nodes == {
+            1,
+        }
         assert s.shard_leader == 1
         print(s)