Fixes issues related to gr.Dataframe receiving an empty dataframe (#…

…7257)

* fixed

* tested and formatted

* add changeset

* add changeset

* fixes

* revert

* dataframe

* clean

---------

Co-authored-by: gradio-pr-bot <gradio-pr-bot@users.noreply.github.com>
Co-authored-by: pngwn <hello@pngwn.io>
Co-authored-by: Abubakar Abid <abubakar@huggingface.co>

.changeset/two-ghosts-sneeze.md

@@ -0,0 +1,5 @@
+---
+"gradio": patch
+---
+
+fix:Fixes issues related to `gr.Dataframe` receiving an empty dataframe

gradio/components/dataframe.py

@@ -126,7 +126,6 @@ def __init__(
         self.col_count = self.__process_counts(
             col_count, len(headers) if headers else 3
         )
-
         self.__validate_headers(headers, self.col_count[0])
 
         self.headers = (
@@ -199,13 +198,18 @@ def preprocess(
         """
         if self.type == "pandas":
             if payload.headers is not None:
-                return pd.DataFrame(payload.data, columns=payload.headers)
+                return pd.DataFrame(
+                    [] if payload.data == [[]] else payload.data,
+                    columns=payload.headers,
+                )
             else:
                 return pd.DataFrame(payload.data)
         if self.type == "polars":
             polars = _import_polars()
             if payload.headers is not None:
-                return polars.DataFrame(payload.data, schema=payload.headers)
+                return polars.DataFrame(
+                    [] if payload.data == [[]] else payload.data, schema=payload.headers
+                )
             else:
                 return polars.DataFrame(payload.data)
         if self.type == "numpy":
@@ -240,12 +244,19 @@ def postprocess(
         if value is None:
             return self.postprocess(self.empty_input)
         if isinstance(value, dict):
+            if len(value) == 0:
+                return DataframeData(headers=self.headers, data=[[]])
             return DataframeData(
                 headers=value.get("headers", []), data=value.get("data", [[]])
             )
         if isinstance(value, (str, pd.DataFrame)):
             if isinstance(value, str):
                 value = pd.read_csv(value)  # type: ignore
+            if len(value) == 0:
+                return DataframeData(
+                    headers=list(value.columns),  # type: ignore
+                    data=[[]],  # type: ignore
+                )
             return DataframeData(
                 headers=list(value.columns),  # type: ignore
                 data=value.to_dict(orient="split")["data"],  # type: ignore
@@ -262,25 +273,27 @@ def postprocess(
                     "Cannot display Styler object in interactive mode. Will display as a regular pandas dataframe instead."
                 )
             df: pd.DataFrame = value.data  # type: ignore
+            if len(df) == 0:
+                return DataframeData(
+                    headers=list(df.columns),
+                    data=[[]],
+                    metadata=self.__extract_metadata(value),  # type: ignore
+                )
             return DataframeData(
                 headers=list(df.columns),
                 data=df.to_dict(orient="split")["data"],  # type: ignore
                 metadata=self.__extract_metadata(value),  # type: ignore
             )
-        elif isinstance(value, (str, pd.DataFrame)):
-            df = pd.read_csv(value) if isinstance(value, str) else value  # type: ignore
-            return DataframeData(
-                headers=list(df.columns),
-                data=df.to_dict(orient="split")["data"],  # type: ignore
-            )
         elif _is_polars_available() and isinstance(value, _import_polars().DataFrame):
+            if len(value) == 0:
+                return DataframeData(headers=list(value.to_dict().keys()), data=[[]])
             df_dict = value.to_dict()
             headers = list(df_dict.keys())
             data = list(zip(*df_dict.values()))
             return DataframeData(headers=headers, data=data)
         elif isinstance(value, (np.ndarray, list)):
             if len(value) == 0:
-                return self.postprocess([[]])
+                return DataframeData(headers=self.headers, data=[[]])
             if isinstance(value, np.ndarray):
                 value = value.tolist()
             if not isinstance(value, list):

test/test_components.py

@@ -1187,7 +1187,7 @@ def test_postprocess(self):
         """
         dataframe_output = gr.Dataframe()
         output = dataframe_output.postprocess([]).model_dump()
-        assert output == {"data": [[]], "headers": [], "metadata": None}
+        assert output == {"data": [[]], "headers": ["1", "2", "3"], "metadata": None}
         output = dataframe_output.postprocess(np.zeros((2, 2))).model_dump()
         assert output == {
             "data": [[0, 0], [0, 0]],