fix: type schema not checking for empty columns (#1679)

alexbarros · azory-ydata · web-flow · commit f3bc959a5a9b · 2024-12-06T12:30:47.000-03:00
* fix: type schema not checking for empty columns

* fix: remove alerts unused parameters

* fix: indicate user defined type on empty columns

* fix(linting): code formatting

---------

Co-authored-by: Azory YData Bot &lt;azory@ydata.ai&gt;
diff --git a/src/ydata_profiling/model/alerts.py b/src/ydata_profiling/model/alerts.py
@@ -634,7 +634,7 @@ def supported_alerts(summary: dict) -> List[Alert]:
     return alerts
 
 
-def unsupported_alerts(summary: Dict[str, Any]) -> List[Alert]:
+def unsupported_alerts() -> List[Alert]:
     alerts: List[Alert] = [
         UnsupportedAlert(),
         RejectedAlert(),
@@ -657,7 +657,7 @@ def check_variable_alerts(config: Settings, col: str, description: dict) -> List
     alerts += generic_alerts(description)
 
     if description["type"] == "Unsupported":
-        alerts += unsupported_alerts(description)
+        alerts += unsupported_alerts()
     else:
         alerts += supported_alerts(description)
 
diff --git a/src/ydata_profiling/model/pandas/summary_pandas.py b/src/ydata_profiling/model/pandas/summary_pandas.py
@@ -16,6 +16,14 @@
 from ydata_profiling.utils.dataframe import sort_column_names
 
 
+def _is_cast_type_defined(typeset: VisionsTypeset, series: str) -> bool:
+    return (
+        isinstance(typeset, ProfilingTypeSet)
+        and typeset.type_schema
+        and series in typeset.type_schema
+    )
+
+
 @describe_1d.register
 def pandas_describe_1d(
     config: Settings,
@@ -38,11 +46,10 @@ def pandas_describe_1d(
     # Make sure pd.NA is not in the series
     series = series.fillna(np.nan)
 
-    if (
-        isinstance(typeset, ProfilingTypeSet)
-        and typeset.type_schema
-        and series.name in typeset.type_schema
-    ):
+    has_cast_type = _is_cast_type_defined(typeset, series.name)
+    cast_type = str(typeset.type_schema[series.name]) if has_cast_type else None
+
+    if has_cast_type and not series.isna().all():
         vtype = typeset.type_schema[series.name]
 
     elif config.infer_dtypes:
@@ -55,7 +62,12 @@ def pandas_describe_1d(
         vtype = typeset.detect_type(series)
 
     typeset.type_schema[series.name] = vtype
-    return summarizer.summarize(config, series, dtype=vtype)
+    summary = summarizer.summarize(config, series, dtype=vtype)
+    # Cast type is only used on unsupported columns rendering pipeline
+    # to indicate the correct variable type when inference is not possible
+    summary["cast_type"] = cast_type
+
+    return summary
 
 
 @get_series_descriptions.register
diff --git a/src/ydata_profiling/report/structure/variables/render_generic.py b/src/ydata_profiling/report/structure/variables/render_generic.py
@@ -12,7 +12,7 @@ def render_generic(config: Settings, summary: dict) -> dict:
     info = VariableInfo(
         anchor_id=summary["varid"],
         alerts=summary["alerts"],
-        var_type="Unsupported",
+        var_type=summary["cast_type"] or "Unsupported",
         var_name=summary["varname"],
         description=summary["description"],
         style=config.html.style,
diff --git a/tests/unit/test_typeset_default.py b/tests/unit/test_typeset_default.py
@@ -475,3 +475,14 @@ def test_type_schema(dataframe: pd.DataFrame, column: str, type_schema: dict):
     assert prof.typeset.type_schema[column] == prof.typeset._get_type(
         type_schema[column]
     )
+
+
+def test_type_schema_with_null_column():
+    df = pd.DataFrame({"null_col": [None] * 100})
+    prof = ProfileReport(df, type_schema={"null_col": "datetime"})
+    description = prof.description_set
+    assert description.variables["null_col"]["type"] == "Unsupported"
+
+    prof = ProfileReport(df, type_schema={"null_col": "numeric"})
+    description = prof.description_set
+    assert description.variables["null_col"]["type"] == "Unsupported"
