Fix code to work with Fugue 0.8.7 (#245)

* Fix code to work with Fugue 0.8.7

* update

* update

* update

* Update pyproject.toml

Co-authored-by: Faisal <[email protected]>

---------

Co-authored-by: Faisal <[email protected]>

datacompy/__init__.py

@@ -13,7 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-__version__ = "0.10.4"
+__version__ = "0.10.5"
 
 from datacompy.core import *
 from datacompy.fugue import (

datacompy/fugue.py

@@ -541,6 +541,7 @@ def _distributed_compare(
 
     def _serialize(dfs: Iterable[pd.DataFrame], left: bool) -> Iterable[Dict[str, Any]]:
         for df in dfs:
+            df = df.convert_dtypes()
             cols = {}
             for name in df.columns:
                 col = df[name]
@@ -577,8 +578,28 @@ def _deserialize(
         arr = [pickle.loads(r["data"]) for r in df if r["left"] == left]
         if len(arr) > 0:
             return pd.concat(arr).sort_values(schema.names).reset_index(drop=True)
-        return pd.DataFrame(
-            {k: pd.Series(dtype=v) for k, v in schema.pandas_dtype.items()}
+        # The following is how to construct an empty pandas dataframe with
+        # the correct schema, it avoids pandas schema inference which is wrong.
+        # This is not needed when upgrading to Fugue >= 0.8.7
+        sample_row: List[Any] = []
+        for field in schema.fields:
+            if pa.types.is_string(field.type):
+                sample_row.append("x")
+            elif pa.types.is_integer(field.type):
+                sample_row.append(1)
+            elif pa.types.is_floating(field.type):
+                sample_row.append(1.1)
+            elif pa.types.is_boolean(field.type):
+                sample_row.append(True)
+            elif pa.types.is_timestamp(field.type):
+                sample_row.append(pd.NaT)
+            else:
+                sample_row.append(None)
+        return (
+            pd.DataFrame([sample_row], columns=schema.names)
+            .astype(schema.pandas_dtype)
+            .convert_dtypes()
+            .head(0)
         )
 
     def _comp(df: List[Dict[str, Any]]) -> List[List[Any]]:

pyproject.toml

@@ -15,7 +15,7 @@ dependencies = [
     "pandas<=2.0.2,>=0.25.0",
     "numpy<=1.26.0,>=1.22.0",
     "ordered-set<=4.1.0,>=4.0.2",
-    "fugue<=0.9.0,>=0.8.6",
+    "fugue<=0.8.7,>=0.8.7",
 ]
 requires-python = ">=3.8.0"
 classifiers = [

tests/test_fugue.py

@@ -80,37 +80,39 @@ def upper_col_df(shuffle_df):
 
 @pytest.fixture
 def simple_diff_df1():
-    return pd.DataFrame(dict(aa=[0, 1, 0], bb=[2.1, 3.1, 4.1]))
+    return pd.DataFrame(dict(aa=[0, 1, 0], bb=[2.1, 3.1, 4.1])).convert_dtypes()
 
 
 @pytest.fixture
 def simple_diff_df2():
-    return pd.DataFrame(dict(aa=[1, 0, 1], bb=[3.1, 4.1, 5.1], cc=["a", "b", "c"]))
+    return pd.DataFrame(
+        dict(aa=[1, 0, 1], bb=[3.1, 4.1, 5.1], cc=["a", "b", "c"])
+    ).convert_dtypes()
 
 
 @pytest.fixture
 def no_intersection_diff_df1():
     np.random.seed(0)
-    return pd.DataFrame(dict(x=["a"], y=[0.1]))
+    return pd.DataFrame(dict(x=["a"], y=[0.1])).convert_dtypes()
 
 
 @pytest.fixture
 def no_intersection_diff_df2():
-    return pd.DataFrame(dict(x=["b"], y=[1.1]))
+    return pd.DataFrame(dict(x=["b"], y=[1.1])).convert_dtypes()
 
 
 @pytest.fixture
 def large_diff_df1():
     np.random.seed(0)
     data = np.random.randint(0, 7, size=10000)
-    return pd.DataFrame({"x": data, "y": np.array([9] * 10000)})
+    return pd.DataFrame({"x": data, "y": np.array([9] * 10000)}).convert_dtypes()
 
 
 @pytest.fixture
 def large_diff_df2():
     np.random.seed(0)
     data = np.random.randint(6, 11, size=10000)
-    return pd.DataFrame({"x": data, "y": np.array([9] * 10000)})
+    return pd.DataFrame({"x": data, "y": np.array([9] * 10000)}).convert_dtypes()
 
 
 def test_is_match_native(