CLN: .values -> ._values #32778
CLN: .values -> ._values #32778
Conversation
There was a problem hiding this comment.
Similar general question as in #32781: what logic do we we use to decide when to use which property?
Just always _values when possible (to ensure we always are using the correct representation, ndarray or EA)? Or do we still prefer to use the public attribute where it's not different from _values?
Also, in the latter case, that's typically because we are either sure the result is always numpy array (in which case np.asarray() or to_numpy might be more explicit) or always an ExtensionArray (in which case .array might be more idiomatic).
The default should be ._values, as .values may be either lossy (dt64tz), involve an object-dtype cast (period, interval), or be surprising (until recently I forgot that Series[Sparse].values gives a SparseArray, not ndarray). Both this and #32781 are aimed at usages where the two are equivalent. This should make it easier to focus attention on usages where they are not equivalent, and double-check whether those are intentional or improveable. |
|
comments here: #32781 (comment) |
|
Travis yellow is wrong; travis is green |
|
kk, no test covers this? |
not sure what you're referring to |
No description provided.