REF: rename _data-->_mgr

[jbrockmendel]

At the sprint there was discussion of trimming down the exposed surface of internals. Among the things that makes this difficult is identifying exactly what that surface is. This makes it feasiable to grep for all accesses by using the unambiguous ._mgr instead of the overloaded ._data.

[jreback]

what is the upside of changing this?

[jreback]

is this actually hit anywhere? (maybe in pyarrow)?

[jbrockmendel]

It's not hit in our tests, but since pyarrow/fastparquet still reaches into core.internals, I figured we needed the alias for compat

[jbrockmendel]

what is the upside of changing this?

Discussed in the OP: it is hard to tell right now where BlockManager is accessed across the code-base (most of which we want to remove). With this, you can find it immediately by grepping.

[jreback]

what is the upside of changing this?

Discussed in the OP: it is hard to tell right now where BlockManager is accessed across the code-base (most of which we want to remove). With this, you can find it immediately by grepping.

I am not sure this is a positive return given the churn on other packages.