From 8e081c015417c5a8d2a99f9db6bbc9a2c438e477 Mon Sep 17 00:00:00 2001
From: GALI PREM SAGAR <sagarprem75@gmail.com>
Date: Mon, 18 Sep 2023 12:51:08 -0500
Subject: [PATCH] Add support for nested dict in `DataFrame` constructor
 (#14119)

Fixes: #14096

This PR enables nested dict initialization support in `DataFrame` constructor.

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - Matthew Roeschke (https://github.com/mroeschke)

URL: https://github.com/rapidsai/cudf/pull/14119
---
 python/cudf/cudf/core/dataframe.py       |  4 ++--
 python/cudf/cudf/tests/test_dataframe.py | 19 +++++++++++++++++++
 2 files changed, 21 insertions(+), 2 deletions(-)

diff --git a/python/cudf/cudf/core/dataframe.py b/python/cudf/cudf/core/dataframe.py
index 5a3d25a08a7..4fc175512a0 100644
--- a/python/cudf/cudf/core/dataframe.py
+++ b/python/cudf/cudf/core/dataframe.py
@@ -977,7 +977,7 @@ def _align_input_series_indices(data, index):
         input_series = [
             Series(val)
             for val in data.values()
-            if isinstance(val, (pd.Series, Series))
+            if isinstance(val, (pd.Series, Series, dict))
         ]
 
         if input_series:
@@ -994,7 +994,7 @@ def _align_input_series_indices(data, index):
                 index = aligned_input_series[0].index
 
             for name, val in data.items():
-                if isinstance(val, (pd.Series, Series)):
+                if isinstance(val, (pd.Series, Series, dict)):
                     data[name] = aligned_input_series.pop(0)
 
         return data, index
diff --git a/python/cudf/cudf/tests/test_dataframe.py b/python/cudf/cudf/tests/test_dataframe.py
index 61372bab3ad..652bdbbee45 100644
--- a/python/cudf/cudf/tests/test_dataframe.py
+++ b/python/cudf/cudf/tests/test_dataframe.py
@@ -10349,3 +10349,22 @@ def test_dataframe_round_builtin(digits):
     actual = round(gdf, digits)
 
     assert_eq(expected, actual)
+
+
+def test_dataframe_init_from_nested_dict():
+    ordered_dict = OrderedDict(
+        [
+            ("one", OrderedDict([("col_a", "foo1"), ("col_b", "bar1")])),
+            ("two", OrderedDict([("col_a", "foo2"), ("col_b", "bar2")])),
+            ("three", OrderedDict([("col_a", "foo3"), ("col_b", "bar3")])),
+        ]
+    )
+    pdf = pd.DataFrame(ordered_dict)
+    gdf = cudf.DataFrame(ordered_dict)
+
+    assert_eq(pdf, gdf)
+    regular_dict = {key: dict(value) for key, value in ordered_dict.items()}
+
+    pdf = pd.DataFrame(regular_dict)
+    gdf = cudf.DataFrame(regular_dict)
+    assert_eq(pdf, gdf)