Hey @mroeschke, can you please take a look at my if the direction is right for you (i.e. you are OK with an additional argument) before I will fix the failing tests, linting errors and adjust the documentation. Ty in advance!

merge already has a quite complex signature, and what you are trying to solve here ca be easily done in pandas with:

pd.merge(df1.add_suffix("_left"), df2.add_suffix("_right"))

Let me know if I'm missing something, but while seems some people would appreciate that, doesn't seem any core dev is excited about this, ans I understand why, since it makes a tricky method even trickier.

@TomAugspurger in the issue seems like you were a bit more positive than others about adding this when discussed some time ago. Would you move on with this PR? Otherwise let's close it.

merge already has a quite complex signature, and what you are trying to solve here ca be easily done in pandas with:

pd.merge(df1.add_suffix("_left"), df2.add_suffix("_right"))

Let me know if I'm missing something, but while seems some people would appreciate that, doesn't seem any core dev is excited about this, ans I understand why, since it makes a tricky method even trickier.

Thanks for your feedback!

Let me motivate the additional flag approach. I agree with you, using add_suffix is a valid approach, which however adds complexity on the code of the uses, forcing them to pass additional arguments like left_on="uuid_left", right_on="id_right", which makes it even more complicated.

Using force_suffixes would make the joins with wide data frames in exploratory settings very easy, because people cannot remember 10 different column names for each of the participants.

However I also agree on the additional complexity of the upcoming implementation. The internal logic of renaming and returning columns is quite complex already, which is not easy to grasp, maintain and test.

I will wait upon your decision.

Thanks for the clarification. I didn't realize the suffix would be added to the columns to join and it'd make things more complex than just adding the add_suffix, which otherwise feels like a quite clean approach. There is clearly a trade-off here, I'm personally fine with the change even if adds some complexity, since there seems to be many users who would appreciate that.

@pandas-dev/pandas-core any opinion on adding a flag to pandas.merge to make the suffixes be added also to non-conflicting names?

@pandas-dev/pandas-core any opinion on adding a flag to pandas.merge to make the suffixes be added also to non-conflicting names?

I think this is a nice idea. If we default it to False, then the current behavior is preserved.

This is a situation I've run into occasionally. It's a few lines of user code, and yes, you need to track what you're joining on. I don't think it's unreasonable for the onus to be on users here, but no objection to adding a flag.