[FEA] Add developer/private cudf.pandas API to check for proxy objects about cudf HOT 5 CLOSED

@Matt711's check should be sufficient. By definition every scenario that we're thinking of is one where the caller knows the wrapped type (either slow or fast) that they expected to get.

from cudf.

When evaluating use cases for this feature we should also be thinking about if they would be better served by improving how well the proxy object masquerades for the original. We can't make isinstance usable for the differentiation, but ideally calling codes wouldn't have to differentiate, so any time that they do probably indicates an area where the proxy isn't a drop-in replacement to the extent that we'd like it to be.

from cudf.

I'm not sure if we want to go this route, but every proxy type is an instance of the same metaclass. So we could check for that.

s = pd.Series([1])
assert type(type(s)) == cudf.pandas.fast_slow_proxy._FastSlowProxyMeta

from cudf.

I think a developer API that checks for _FastSlowProxyMeta would be good initial check.

I'm not sure if it's worth differentiating specific object types, i.e. pandas proxy objects vs numpy proxy objects, or make it fully generic by checking that metaclass. My initial feeling is checking _FastSlowProxyMeta would be preferable

from cudf.

I'll add the change for it

from cudf.