Error when using unhashable key in '__setitem__'

[YarShev]

Describe the problem

I get TypeError: unhashable type: 'slice' when trying to set value(s) to Series using slice.

Source code / logs

import modin.pandas as pd
import pandas

ps = pandas.Series([1,2,3])
ms = pd.Series([1,2,3])

ps[1:3] = 0
ps
0    1
1    0
2    0
dtype: int64

ms[1:3] = 0
TypeError: unhashable type: 'slice'

[dchigarev]

It doesn't work for all cases where key is not a label-like value.

Example

import modin.pandas as pd

ms = pd.Series([1,2,3])
ms[1] = 0 # Works fine
ms[1:3] = 0 # TypeError
ms[[True, False, True]] = 0 # TypeError
ms[[0, 1]] = 0 # TypeError

Current Series.__setitem__ implementation:

modin/modin/pandas/series.py

Lines 287 to 292 in f14a3c1

	def __setitem__(self, key, value):
	if key not in self.keys():
	raise KeyError(key)
	self._create_or_update_from_compiler(
	self._query_compiler.setitem(1, key, value), inplace=True
	)

Problems:

1. We check that key is in self.keys dict assuming that key is always hashable, however it doesn't work for lists and slices.
2. Right after this check it calls query compilers setitem method, which also works only for label-like values, so adding a check that key is hashable to the first if-statement will not resolve the problem

[YarShev]

We probably can fix similar case for DF with the same way because the case for DF is currently defaulting to pandas. @dchigarev , will you look at this, please?

import modin.pandas as pd
df = pd.DataFrame([1,2,3])
ms[1:3] = 0
UserWarning: `DataFrame.setitem_without_string_columns` defaulting to pandas implementation.

[dchigarev]

sure