Fix PandasOptionError import.

[huandzh]

pandas.core.config is no longer available after pandas v0.25

See also:

• Move core.config to _config.config pandas-dev/pandas#25793
• missing pandas.core.config in v0.25 pandas-dev/pandas#27471

[CLAassistant]

All committers have signed the CLA.

[huandzh]

odps/tests/test_config.py tested and passed with pandas 0.25.0.

[qinxuye]

Great job, but I think we can use try except other than version comparison.

[qinxuye]

We could try import pandas.core.config, if failed, try to import pandas._config.config. Documents can be added about pandas version.

[huandzh]

Yes, we can do that now.

But it seems not right to handle compatibility issues with exceptions (exceptions are most likely to be something go unexpected).

OptionError is private in '0.25.0' , but may be published and move to another place afterwards. We can't handle updates directly with ImportError then.

Any idea?

[qinxuye]

I still think try except is a better way. How about just make a modification?

We cannot predict the behavior of pandas in the future, let's fix the current issue first.

[huandzh]

Could you given some more reasoning? I really can't understand why try except is preferred in this use case. Importing from pandas.core.config will sure fail in current and future versions.

And code already committed here did fix the issue.

See also:

pandas-dev/pandas#27553
pandas-dev/pandas#27656

[qinxuye]

Could you please resolve the problem by try except way? thanks.

[qinxuye]

I still think try except is a better way. How about just make a modification?

We cannot predict the behavior of pandas in the future, let's fix the current issue first.

[qinxuye]

Sorry to tell you that we have solved the compatible issue in a way that pandas version does not matter. Refer to

aliyun-odps-python-sdk/odps/config.py

Lines 83 to 91 in ae65e3f

	def getvalue(self, silent=False):
	if self._use_pd:
	import pandas as pd
	try:
	return pd.get_option('.'.join(self._items))
	except (KeyError, LookupError, AttributeError):
	self._use_pd = False
	else:
	return self._val

Thank you all the same for your contribution, I will close this PR anyway, welcome to send other PRs.