BUG: _validate_setitem_value fails to raise for PandasArray

[jbrockmendel]

arr = pd.Series(range(5)).array

arr._validate_setitem_value("foo")  # <- fails to raise
arr._validate_setitem_value(1.5)  # <- fails to raise

AFAICT this doesn't affect anything bc we always extract PandasArray before doing anything.

Note that if this did raise on 1.5, it might cause a problem with arr.searchsorted(floats) raising when we wouldn't want it to.

[topper-123]

Just delete the Pandasarray._validate_setitem_value method, so if something calls this, then an AttributeError is raised?

[jbrockmendel]

it should raise ValueError/TypeError (not sure which off the top of my head), should match DTA/TDA etc behavior

[topper-123]

ah yes, it's set on a parent class as-is, so will have to raise here, of course.

[Jacsep]

take

[happyvignesh]

Hi @topper-123, I am a beginner contributor.
May I know what change I need to add for this issue.

[topper-123]

Hey @happyvignesh. Sounds great you want to contribute.

What is needed here is adding a method called _validate_setitem_value to PandasArray. The method should have the same signature as BaseMaskedArray._validate_setitem_value, but should raise a TypeError if the validated value cannot be inserted without changing the dtype.

[happyvignesh]

Hey @topper-123,
Thanks for your response.
if I understand correctly, I need to add the below snippet in class PandasArray.

`def _validate_setitem_value(self, value):
"""
Check if we have a scalar that we can cast losslessly.

    Raises
    ------
    TypeError
    """
    kind = self.dtype.kind
    # TODO: get this all from np_can_hold_element?
    if kind == "b":
        if lib.is_bool(value):
            return value

    elif kind == "f":
        if lib.is_integer(value) or lib.is_float(value):
            return value

    else:
        if lib.is_integer(value) or (lib.is_float(value) and value.is_integer()):
            return value
        # TODO: unsigned checks

    # Note: without the "str" here, the f-string rendering raises in
    #  py38 builds.
    raise TypeError(f"value '{str(value)}' cannot be inserted without changing the dtype to {self.dtype}")`

Could you please let me know if its fine.
Thanks.

[topper-123]

PandasArray can take many numpy dtypes, e.g. pd.array(["a"], dtype=object) can work, also pd.array(["a", 1]) will have dtype str672.

If you can find a way to ensure the new scalar is compatible with whatever numpy dtype is used, I think that is required also. I'm not sure ATM how'd you do that, but maybe you can find a way.

[theyashl]

Hi @topper-123 , instead of checking kind of data type, can we simply compare the data type itself?

for example:

PandasDtype(type(value)) == self.dtype

As this will compare data type itself of current class and provide value's class.

[dbalagula]

take

[hvsesha]

if isinstance(value, str):
raise ValueError("String values are not allowed in an integer array.")

This also works .Kindly let me know is this looks good .I made change in num_.py

[hvsesha]

Suppose you are setting integer array and if you are calling below statement
import pandas as pd
arr = pd.Series(range(5)).array

arr.setitem(arr,"foo") # <- fails to raise
print(arr)
arr.setitem(arr,"foo") This is raise exception

I tried and it worked

[kvnwng11]

take

[flaviaouyang]

@kvnwng11 are you still working on this issue?

[kvnwng11]

@flaviaouyang Thanks for the ping! I am not still working, so I will unassign myself

[flaviaouyang]

take

[yiorgosynkl]

is this actively being worked on? @flaviaouyang

[flaviaouyang]

@yiorgosynkl no, go ahead

[yiorgosynkl]

take

[jahn96]

@yiorgosynkl are you currently working on this?

[maushumee]

@yiorgosynkl is this being worked on?

[maushumee]

take