FIX-#4456: fix loc in case when need reindex item

docs/release_notes/release_notes-0.15.0.rst

@@ -17,6 +17,7 @@ Key Features and Updates
   * FIX-#4373: Fix invalid file path when trying `read_csv_glob` with `usecols` parameter (#4405)
   * FIX-#4394: Fix issue with multiindex metadata desync (#4395)
   * FIX-#4425: Add parameters to groupby pct_change (#4429)
+  * FIX-#4457: Fix `loc` in case when need reindex item (#4457)
 * Performance enhancements
   * FEAT-#4320: Add connectorx as an alternative engine for read_sql (#4346)
 * Benchmarking enhancements

modin/pandas/indexing.py

@@ -424,23 +424,21 @@ def _broadcast_item(self, row_lookup, col_lookup, item, to_shape):
         # the target the user is trying to overwrite. This
         if isinstance(item, (pandas.Series, pandas.DataFrame, Series, DataFrame)):
             # convert indices in lookups to names, as Pandas reindex expects them to be so
+            kw = {}
             index_values = self.qc.index[row_lookup]
-            if not all(idx in item.index for idx in index_values):
-                raise ValueError(
-                    "Must have equal len keys and value when setting with "
-                    + "an iterable"
-                )
+            if len(index_values) < len(item.index) or not all(
+                idx in item.index for idx in index_values
+            ):
+                kw["index"] = index_values
             if hasattr(item, "columns"):
                 column_values = self.qc.columns[col_lookup]
-                if not all(col in item.columns for col in column_values):
-                    # TODO: think if it is needed to handle cases when columns have duplicate names
-                    raise ValueError(
-                        "Must have equal len keys and value when setting "
-                        + "with an iterable"
-                    )
-                item = item.reindex(index=index_values, columns=column_values)
-            else:
-                item = item.reindex(index=index_values)
+                if len(column_values) < len(item.columns) or not all(
+                    col in item.columns for col in column_values
+                ):
+                    kw["columns"] = column_values
+            # New value for columns/index make that reindex add NaN values
+            if kw:
+                item = item.reindex(**kw)
         try:
             item = np.array(item)
             if np.prod(to_shape) == np.prod(item.shape):

modin/pandas/test/dataframe/test_indexing.py

@@ -67,6 +67,24 @@ def eval_setitem(md_df, pd_df, value, col=None, loc=None):
     )
 
 
+def eval_loc(md_df, pd_df, value, key):
+    if isinstance(value, tuple):
+        assert len(value) == 2
+        # case when value for pandas different
+        md_value, pd_value = value
+    else:
+        md_value, pd_value = value, value
+
+    eval_general(
+        md_df,
+        pd_df,
+        lambda df: df.loc.__setitem__(
+            key, pd_value if isinstance(df, pandas.DataFrame) else md_value
+        ),
+        __inplace__=True,
+    )
+
+
 @pytest.mark.parametrize(
     "dates",
     [
@@ -384,6 +402,46 @@ def test_loc(data):
         modin_df.loc["NO_EXIST"]
 
 
+def test_loc_4456():
+    data = test_data["float_nan_data"]
+    modin_df, pandas_df = pd.DataFrame(data), pandas.DataFrame(data)
+
+    key = ["col1", "col2"]
+
+    # len(key) == df_value.columns; different columns
+    value = [[1, 2]] * pandas_df.shape[0]
+    df_value = pandas.DataFrame(value, columns=["value_col1", "value_col2"])
+    eval_loc(modin_df, pandas_df, df_value, (slice(None), key))
+
+    # len(key) == df_value.columns; different columns; modin DataFrame
+    mdf_value = pd.DataFrame(df_value)
+    eval_loc(modin_df, pandas_df, (mdf_value, df_value), (slice(None), key))
+
+    # len(key) == df_value.columns; different index
+    df_value = pandas.DataFrame(
+        value, columns=key, index=list(reversed(pandas_df.index))
+    )
+    eval_loc(modin_df, pandas_df, df_value, (slice(None), key))
+
+    # len(key) > df_value.columns
+    value = [[1]] * pandas_df.shape[0]
+    df_value = pandas.DataFrame(value, columns=key[:1])
+    eval_loc(modin_df, pandas_df, df_value, (slice(None), key))
+
+    # len(key) == df_value.columns; different columns; modin Series
+    # import pdb; pdb.set_trace()
+    md_series_value = mdf_value[mdf_value.columns[0]]
+    pd_series_value = df_value[df_value.columns[0]]
+    eval_loc(
+        modin_df, pandas_df, (md_series_value, pd_series_value), (slice(None), key)
+    )
+
+    # len(key) < df_value.columns
+    value = [[1, 2, 3]] * pandas_df.shape[0]
+    df_value = pandas.DataFrame(value, columns=key + ["col3"])
+    eval_loc(modin_df, pandas_df, df_value, (slice(None), key))
+
+
 # This tests the bug from https://github.com/modin-project/modin/issues/3736
 def test_loc_setting_single_categorical_column():
     modin_df = pd.DataFrame({"status": ["a", "b", "c"]}, dtype="category")