DEPR: .as_blocks and .blocks

[jreback]

These shouldn't be public methods. Deprecate and make private (lead with _) for now.

[jorisvandenbossche]

+ 1

(as long as we keep the private access like df._data.blocks stable, but I don't think we have any plans or incentives to change that in 1.x)

Is it even necessary to make them private? Eg internally we have exact one usage of as_blocks, which I suppose can easily be changes to go through _data

cc @pandas-dev/pandas-core

[jbrockmendel]

+1

Speaking of _data, is there (or could there be) an implicit convention that foo._data should always (with the possible exception of Index._data) point at a BlockManager?

[jreback]

Speaking of _data, is there (or could there be) an implicit convention that foo._data should always

its just the internal impl. For DataFrame/Series its a BlockManager. For Index its an ndarry (or another array-like type, e.g. Categorical).