Revert: REF: do length-checks in boilerplate decorator

[charlesdong1991]

closes #34123

As discussed in #34123
let's revert this PR34081 first since it has potentially big impact on some cases in ops.

cc @jbrockmendel @jorisvandenbossche

[pep8speaks]

Hello @charlesdong1991! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2020-05-15 09:02:38 UTC

[jbrockmendel]

other -> other.dtype

[charlesdong1991]

done, thanks!

[jbrockmendel]

small request, otherwise LGTM

[jbrockmendel]

Can you add test(s) for the behavior that was broken in #34081

[charlesdong1991]

emm, something might have changed alongside, now after this revert pr,

>>> pd.Index([('a', 'b'), ('b', 'c'), ('c', 'a')]) == ('c', 'a')
array([False, False, False])

it returns all False.

@jbrockmendel any idea which PR recently merged could influence this?

[charlesdong1991]

emm, not sure how on earth i could run get array([False, False, True]) result that day, cannot replicate anymore. but seems since 1.0 or later this case returns array([False, False, False]) which i feel is wrong as well as a warning from numpy. i put some comments and todo there and pls let me know if it is appropriate

i could create a new issue to address this if you think is necessary.

[jorisvandenbossche]

@charlesdong1991 I was also quite sure that I saw the correct behaviour in one of my environments, but also don't know where anymore.

Now, looking at it on 1.0.3, the problem is that it is creating a MultiIndex, and not a plain index with tuples in it. Ensuring we get an Index, I actually get the correct behaviour:

In [14]: pd.Index([('a', 'b'), ('b', 'c'), ('c', 'a')], tupleize_cols=False) == ('c', 'a') 
Out[14]: array([False, False,  True])

(and this example also actually fails on master)

[charlesdong1991]

yeah, indeed, the issue i saw is that in such case, it automatically recognises as a MI, not Index.

@jorisvandenbossche thanks very much, I will add tupleize_cols=False in the test to get the correct result and I will also create an issue to report this case.

[jbrockmendel]

LGTM

[jreback]

thanks @charlesdong1991

can you open a new issue to re-implement this (e.g. essentially a cleanup to revert the revert with appropriate changes).

[charlesdong1991]

ref #34233

cc. @jreback