only call parsers once (#1898)

* only call parsers once Signed-off-by: cosmicBboy <[email protected]> * fix dask test error Signed-off-by: cosmicBboy <[email protected]> --------- Signed-off-by: cosmicBboy <[email protected]>
unionai-oss · Jan 22, 2025 · 33fe68b · 33fe68b
1 parent a832172
commit 33fe68b
Show file tree

Hide file tree

Showing 3 changed files with 35 additions and 13 deletions.
diff --git a/pandera/api/pandas/types.py b/pandera/api/pandas/types.py
@@ -108,6 +108,7 @@ def register_geopandas_backend():
     register_fn = {
         "pandas": register_pandas_backend,
         "dask_expr": register_dask_backend,
+        "dask": register_dask_backend,
         "modin": register_modin_backend,
         "pyspark": register_pyspark_backend,
         "geopandas": register_geopandas_backend,

diff --git a/pandera/backends/pandas/components.py b/pandera/backends/pandas/components.py
@@ -119,27 +119,29 @@ def validate_column(check_obj, column_name, return_check_obj=False):
                 except SchemaErrors as exc:
                     error_handler.collect_errors(exc.schema_errors)
 
-            if schema.parsers:
-                for parser_index, parser in enumerate(schema.parsers):
-                    check_obj[column_name] = self.run_parser(
-                        check_obj[column_name],
-                        parser,
-                        parser_index,
-                    ).parser_output
-
             if is_table(check_obj[column_name]):
                 for i in range(check_obj[column_name].shape[1]):
-                    validate_column(
-                        check_obj[column_name].iloc[:, [i]], column_name
+                    validated_column = validate_column(
+                        check_obj[column_name].iloc[:, [i]],
+                        column_name,
+                        return_check_obj=True,
                     )
+                    if schema.parsers:
+                        check_obj[column_name] = validated_column
             else:
                 if getattr(schema, "drop_invalid_rows", False):
-                    # replace the check_obj with the validated check_obj
+                    # replace the check_obj with the validated
                     check_obj = validate_column(
                         check_obj, column_name, return_check_obj=True
                     )
-                else:
-                    validate_column(check_obj, column_name)
+
+                validated_column = validate_column(
+                    check_obj,
+                    column_name,
+                    return_check_obj=True,
+                )
+                if schema.parsers:
+                    check_obj[column_name] = validated_column
 
         if lazy and error_handler.collected_errors:
             raise SchemaErrors(

diff --git a/tests/core/test_parsers.py b/tests/core/test_parsers.py
@@ -94,3 +94,22 @@ def int_column_lt_100(cls, series: pd.Series):
     )
     with pytest.raises(pa.errors.SchemaInitError, match=err_msg):
         Schema.to_schema()
+
+
+def test_parser_called_once():
+
+    data = pd.DataFrame({"col": [2.0, 4.0, 9.0]})
+    n_calls = 0
+
+    class DFModel(pa.DataFrameModel):
+        col: float
+
+        @pa.parser("col")
+        @classmethod
+        def negate(cls, series):
+            nonlocal n_calls
+            n_calls += 1
+            return series * -1
+
+    DFModel.validate(data)
+    assert n_calls == 1