.astype(SparseDtype(float)) on empty dataframe leads to "ValueError: No objects to concatenate"

[choucavalier]

Code Sample, a copy-pastable example if possible

import pandas as pd
pd.DataFrame().astype(pd.SparseDtype(float))

Problem description

Converting an empty dataframe to a sparse representation leads to a ValueError.

ValueError: No objects to concatenate

Expected Output

I expected this to work even though the dataframe is empty.

Output of pd.show_versions()

>>> pandas.show_versions()

INSTALLED VERSIONS
------------------
commit           : None
python           : 3.7.4.final.0
python-bits      : 64
OS               : Darwin
OS-release       : 19.0.0
machine          : x86_64
processor        : i386
byteorder        : little
LC_ALL           : en_US.UTF-8
LANG             : en_US.UTF-8
LOCALE           : en_US.UTF-8

pandas           : 1.0.3
numpy            : 1.18.2
pytz             : 2019.3
dateutil         : 2.8.1
pip              : 19.3.1
setuptools       : 45.0.0
Cython           : None
pytest           : 5.2.2
hypothesis       : None
sphinx           : None
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : 4.4.2
html5lib         : None
pymysql          : None
psycopg2         : None
jinja2           : 2.10.3
IPython          : 7.11.1
pandas_datareader: None
bs4              : None
bottleneck       : None
fastparquet      : None
gcsfs            : None
lxml.etree       : 4.4.2
matplotlib       : 3.1.2
numexpr          : 2.7.0
odfpy            : None
openpyxl         : None
pandas_gbq       : None
pyarrow          : None
pytables         : None
pytest           : 5.2.2
pyxlsb           : None
s3fs             : None
scipy            : 1.4.1
sqlalchemy       : None
tables           : 3.6.0
tabulate         : None
xarray           : None
xlrd             : 1.2.0
xlwt             : None
xlsxwriter       : None
numba            : 0.46.0

[simonjayhawkins]

Thanks @tgy for the report. The same exception is raised for Int64Dtype, StringDtype and BooleanDtype and maybe more.

[choucavalier]

@simonjayhawkins yep, i used float as an example but i think it fails for all the types because the code clearly does the same thing for all these types. see #33118

[KardoPaska]

My motivation is to make one line of code robust to "blank" inputs. And I encountered the same kind of error for slightly different input...

summarys = [{'testid': 'abc', 'avg': 1.2, 'count': 5, 'ok': True}, # Pass
            {'testid': 'xyz',             'count': 0, 'ok': False},# Pass
                                                           dict()] # Fail (empty)
summary_dtypes = {'testid': str, 'avg': float, 'count': pd.Int64Dtype(), 'ok': bool}
actual_cols_set = {kk for x in summarys for kk in x.keys()}
actual_dtypes = {k: v for k,v in summary_dtypes.items() if k in actual_cols_set}

This is all good (aside from "missing" bool being True)...

>>> pd.DataFrame(summarys).astype(dtype=actual_dtypes)

  testid  avg  count     ok
0    abc  1.2      5   True
1    xyz  NaN      0  False
2    NaN  NaN   <NA>   True

this is also fine...

>>> pd.DataFrame([dict(), dict()])

Empty DataFrame
Columns: []
Index: [0, 1]

... but... for a list of empty dict(), my actual_dtypes will also be an empty dict()...

>>> pd.DataFrame([dict(), dict()]).astype(dtype=dict())

ValueError: No objects to concatenate

So perhaps a short circuit for empty dataframes? (instead of raising error?)

[choucavalier]

@KardoPaska My small change on #33118 seems to fix your issue as well:

>>> import pandas as pd
>>> pd.DataFrame([dict(), dict()]).astype(dtype=dict())
Empty DataFrame
Columns: []
Index: [0, 1]

Waiting for @jreback to approve the change on the PR

[choucavalier]

Shouldn't the code be pd.DataFrame([dict(), dict()]).astype(dtype=dict) instead? (dict instead of dict() in the dtype)? That works on master.