BUG: to_hdf and HDFStore raise KeyError for DataFrame subclasses

[sytham]

• [x ] I have checked that this issue has not already been reported.

• [x ] I have confirmed this bug exists on the latest version of pandas.

---

Code Sample, a copy-pastable example

import pandas as pd
class SubDataFrame(pd.DataFrame):
    @property
    def _constructor(self):
        return SubDataFrame

# fails with KeyError
sdf = SubDataFrame({'a':[1,2], 'b':[3,4]})
sdf.to_hdf('test.h5', 'test')

with pd.HDFStore('test.h5') as store:
    store.put('test', sdf)

Problem description

to_hdf() and HDFStore.put() fail for DataFrame subclasses.

This happens because in pandas/io/pytables.py in _create_storer line 1578 (or thereabouts), the _TYPE_MAP is accessed by type(), whereas upon _create_storer entry, the check is done using isinstance().

Expected Output

The check upon entry of _create_storer should at least be consistent with the way _TYPE_MAP is accessed. So if the choice is not to support writing DataFrame subclasses to HDF, instead of a KeyError, a TypeError("value must be None, Series, or DataFrame") should be raised.

But ideally, storing subclasses of DataFrame to HDF should be supported.

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None
python : 3.6.10.final.0
python-bits : 64
OS : Linux
OS-release : 4.14.111-1.el7.centos.x86_64
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.0.3
numpy : 1.18.1
pytz : 2019.3
dateutil : 2.8.1
pip : 20.0.2
setuptools : 46.1.3.post20200325
Cython : None
pytest : 5.4.1
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : 2.8.4 (dt dec pq3 ext lo64)
jinja2 : 2.10.3
IPython : 7.10.2
pandas_datareader: None
bs4 : 4.8.2
bottleneck : None
fastparquet : 0.3.2
gcsfs : None
lxml.etree : None
matplotlib : 3.1.1
numexpr : 2.7.1
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pytest : 5.4.1
pyxlsb : None
s3fs : 0.4.0
scipy : 1.2.1
sqlalchemy : 1.3.11
tables : 3.6.1
tabulate : None
xarray : None
xlrd : None
xlwt : None
xlsxwriter : None
numba : 0.46.0

[jreback]

this is possible but would need a community PR
to support

[sytham]

Well, I'd be happy to change line 1528 of _create_storer from
if value is not None and not isinstance(value, (Series, DataFrame)):
to
if value is not None and type(value) not in (Series, DataFrame):
to correctly catch that subclasses of Series or DataFrame are currently not supported (this is why I initially labeled it a bug -- so maybe this ticket could be split in two then, one for this correction, one for implementing the actual support?).

[yangyxt]

I have run into a similar error by just appending pandas chunk dfs to an HDF5 file.
I have confirmed that the first chunk dtypes and the following chunk dtypes are exactly the same for all the columns. BTW, I have 130 columns and all of them are converted to object dtype before writing to hdf5 file.
Still, I got this error message:

`Traceback (most recent call last):
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/io/pytables.py", line 3881, in create_axes
b, b_items = by_items.pop(items)
KeyError: ('1000g2015aug_all', 'AAChange.refGene', 'Alt', 'Alt.1', 'CADD_phred', 'CADD_raw', 'CADD_raw_rankscore', 'CLNALLELEID', 'CLNDISDB', 'CLNDN', 'CLNREVSTAT', 'CLNSIG', 'Chr', 'DANN_rankscore', 'DANN_score', 'Eigen-PC-raw', 'Eigen-r

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/paedwy/disk1/yangyxt/ngs_scripts/Reformatting_hg19_multianno_txt.py", line 243, in reformat
if len(new_chunk) > 0: new_chunk.to_hdf(tmp_hdf5_path, key='table', format='table', append=True)
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/core/generic.py", line 2530, in to_hdf
pytables.to_hdf(path_or_buf, key, self, **kwargs)
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/io/pytables.py", line 278, in to_hdf
f(store)
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/io/pytables.py", line 269, in
f = lambda store: store.append(key, value, **kwargs)
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/io/pytables.py", line 1059, in append
self._write_to_group(key, value, append=append, dropna=dropna, **kwargs)
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/io/pytables.py", line 1525, in _write_to_group
s.write(obj=value, append=append, complib=complib, **kwargs)
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/io/pytables.py", line 4615, in write
return super().write(obj=obj, data_columns=data_columns, **kwargs)
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/io/pytables.py", line 4194, in write
axes=axes, obj=obj, validate=append, min_itemsize=min_itemsize, **kwargs
File "/home/yangyxt/anaconda3/lib/python3.6/site-packages/pandas/io/pytables.py", line 3888, in create_axes
items=(",".join(pprint_thing(item) for item in items))
ValueError: cannot match existing table structure for [1000g2015aug_all,AAChange.refGene,Alt,Alt.1,CADD_phred,CADD_raw,CADD_raw_rankscore,CLNALLELEID,CLNDISDB,CLNDN,CLNREVSTAT,CLNSIG,Chr,DANN_rankscore,DANN_score,Eigen-PC-raw,Eigen-raw,`

I didn't find any solutions in the issues. This issue post is the most relevant one. Pls help debug it. Thanks!

[sytham]

That doesn't look like it's the same issue. This thread is specifically about storing subclasses of DataFrame to HDF.