Merging on int64 and object columns, pandas 0.23.4

[YanTianqi]

Problem description

Hi, I was trying to merge 2 data frame based on a column named ID, which consists of integer.

In the case the ID column is of type numpy.int64 in one df, and of python native int in the other df. In the latest version of 0.23.4, I received this error

ValueError: Your are tryign to merge on int64 and object columns, If you wish to proceed you should use pd.concat

Attached the screenshot of the problem

I have never encountered this situation in earlier panda version such as 0.22.0. The same code ran without an issue.

May I know what is the purpose of adding this constraint in the upgraded version?

[TomAugspurger]

I think #9780 is the relevant issue.

IIRC, trying to merge between object-dtype columns and more specialized types was causing issues.

[YanTianqi]

@TomAugspurger I think the difference between #9780 is that in the previous version, it was not failure without raising error, but rather merging successfully instead.

[TomAugspurger]

Gotcha, I think this will be closed by #21681, but upcasting to object.