Pandas version checks
Reproducible Example
import pandas as pd
class MyDict(dict):
pass
df = pd.DataFrame(
{"A": [1, 2],
"B": ["x", "y"]}
)
out = df.to_dict("index", into=MyDict)
print(out)
print(type(out))
print(type(out[0]))
assert isinstance(out, MyDict)
assert isinstance(out[0], MyDict) # failsIssue Description
Running the code above raises an AssertionError like below.
{0: {'A': 1, 'B': 'x'}, 1: {'A': 2, 'B': 'y'}}
<class '__main__.MyDict'>
<class 'dict'>
---------------------------------------------------------------------------
AssertionError Traceback (most recent call last)
Cell In[19], line 22
18 print(type(out))
19 print(type(out[0]))
20
21 assert isinstance(out, MyDict)
---> 22 assert isinstance(out[0], MyDict) # fails
AssertionError:This suggests DataFrame.to_dict(..., into=...) applying into only to the outer mapping when orient="index".
In the example above, the outer returned object is MyDict, but each nested row mapping is a plain dict.
I checked whether this behavior is intended, but the docstrings of both functions
pandas.core.frame.DataFrame.to_dict()
pandas.core.methods.to_dict.to_dict()
describe the result as dict-like, but do not clarify whether nested mappings should also respect into:
Parameters
----------
orient : str {'dict', 'list', 'series', 'split', 'tight', 'records', 'index'}
Determines the type of the values of the dictionary.
...
- 'index' : dict like {index -> {column -> value}}
I also inspected the to_dict implementation in pandas.core.methods.to_dict.
I believe this behavior comes from the following three return paths in the elif orient == "index" branch:
- plain dict is used directly:
return into_c(
(t[0], dict(zip(df.columns, map(maybe_box_native, t[1:]), strict=True)))
for t in df.itertuples(name=None)
) return into_c(
(t[0], dict(zip(columns, t[1:], strict=True)))
for t in df.itertuples(name=None)
)
- a dict comprehension is used directly, bypassing
into_c for the nested mapping:
return into_c(
(
t[0],
{
column: maybe_box_native(v)
if i in object_dtype_indices_as_set
else v
for i, (column, v) in enumerate(
zip(columns, t[1:], strict=True)
)
},
)
for t in df.itertuples(name=None)
)This could be fixed by using the standardized mapping constructor into_c for the nested row mappings as well. For example, the dict-comprehension branch could be changed to:
return into_c(
(
t[0],
into_c(
(
column,
maybe_box_native(v)
if i in object_dtype_indices_as_set
else v,
)
for i, (column, v) in enumerate(
zip(columns, t[1:], strict=True)
)
),
)
for t in df.itertuples(name=None)
)With this change, the following code passes:
class MyDict(dict):
pass
df = pd.DataFrame(
{
"A": [1, 2],
"B": ["x", "y"],
}
)
out = to_dict(df, "index", into=MyDict)
print(out)
print(type(out))
print(type(out[0]))
assert isinstance(out, MyDict)
assert isinstance(out[0], MyDict) # success{0: {'A': 1, 'B': 'x'}, 1: {'A': 2, 'B': 'y'}}
<class '__main__.MyDict'>
<class '__main__.MyDict'>
I would be happy to open a pull request for this.
Expected Behavior
Expected behavior:
result = df.to_dict(orient="index", into=MyDict)
type(result)
# <class '__main__.MyDict'>
type(result[0])
# <class '__main__.MyDict'>
In other words, the result should be equivalent to:
MyDict(
{
0: MyDict({"A": 1, "B": "x"}),
1: MyDict({"A": 2, "B": "y"}),
}
)
Installed Versions
Details
INSTALLED VERSIONS
commit : 72f2fea
python : 3.14.4
python-bits : 64
OS : Windows
OS-release : 11
Version : 10.0.26200
machine : AMD64
processor : Intel64 Family 6 Model 198 Stepping 2, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : Korean_Korea.utf8
pandas : 3.0.3
numpy : 2.4.3
dateutil : 2.9.0.post0
pip : 26.0.1
Cython : None
sphinx : None
IPython : 9.12.0
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : 4.14.3
bottleneck : None
fastparquet : 2025.12.0
fsspec : 2026.4.0
html5lib : None
hypothesis : None
gcsfs : None
jinja2 : 3.1.6
lxml.etree : 6.0.2
matplotlib : 3.10.8
numba : 0.65.1
numexpr : None
odfpy : None
openpyxl : 3.1.5
psycopg2 : 2.9.11
pymysql : None
pyarrow : 21.0.0
pyiceberg : None
pyreadstat : None
pytest : None
python-calamine : None
pytz : 2026.1.post1
pyxlsb : None
s3fs : None
scipy : 1.17.1
sqlalchemy : 2.0.49
tables : None
tabulate : None
xarray : None
xlrd : None
xlsxwriter : 3.2.9
zstandard : 0.25.0
qtpy : None
pyqt5 : None
Pandas version checks
I have checked that this issue has not already been reported.
I have confirmed this bug exists on the latest version of pandas.
I have confirmed this bug exists on the main branch of pandas.
Reproducible Example
Issue Description
Running the code above raises an
AssertionErrorlike below.{0: {'A': 1, 'B': 'x'}, 1: {'A': 2, 'B': 'y'}} <class '__main__.MyDict'> <class 'dict'> --------------------------------------------------------------------------- AssertionError Traceback (most recent call last) Cell In[19], line 22 18 print(type(out)) 19 print(type(out[0])) 20 21 assert isinstance(out, MyDict) ---> 22 assert isinstance(out[0], MyDict) # fails AssertionError:This suggests
DataFrame.to_dict(..., into=...)applyingintoonly to the outer mapping whenorient="index".In the example above, the outer returned object is
MyDict, but each nested row mapping is a plaindict.I checked whether this behavior is intended, but the docstrings of both functions
describe the result as
dict-like, but do not clarify whether nested mappings should also respectinto:I also inspected the
to_dictimplementation inpandas.core.methods.to_dict.I believe this behavior comes from the following three return paths in the
elif orient == "index"branch:into_cfor the nested mapping:This could be fixed by using the standardized mapping constructor
into_cfor the nested row mappings as well. For example, the dict-comprehension branch could be changed to:With this change, the following code passes:
I would be happy to open a pull request for this.
Expected Behavior
Expected behavior:
In other words, the result should be equivalent to:
Installed Versions
Details
INSTALLED VERSIONS
commit : 72f2fea
python : 3.14.4
python-bits : 64
OS : Windows
OS-release : 11
Version : 10.0.26200
machine : AMD64
processor : Intel64 Family 6 Model 198 Stepping 2, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : Korean_Korea.utf8
pandas : 3.0.3
numpy : 2.4.3
dateutil : 2.9.0.post0
pip : 26.0.1
Cython : None
sphinx : None
IPython : 9.12.0
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : 4.14.3
bottleneck : None
fastparquet : 2025.12.0
fsspec : 2026.4.0
html5lib : None
hypothesis : None
gcsfs : None
jinja2 : 3.1.6
lxml.etree : 6.0.2
matplotlib : 3.10.8
numba : 0.65.1
numexpr : None
odfpy : None
openpyxl : 3.1.5
psycopg2 : 2.9.11
pymysql : None
pyarrow : 21.0.0
pyiceberg : None
pyreadstat : None
pytest : None
python-calamine : None
pytz : 2026.1.post1
pyxlsb : None
s3fs : None
scipy : 1.17.1
sqlalchemy : 2.0.49
tables : None
tabulate : None
xarray : None
xlrd : None
xlsxwriter : 3.2.9
zstandard : 0.25.0
qtpy : None
pyqt5 : None