Retire DataColumn and replace usages

[rcaudy]

No description provided.

[jmao-denver]

The Python changes LGTM. Two questions:

1. what's the performance implication of this change?
2. w/o reading the Java changes, can ColumnVectors be sliced (without flattening)?

[rcaudy]

The Python changes LGTM. Two questions:

1. what's the performance implication of this change?
2. w/o reading the Java changes, can ColumnVectors be sliced (without flattening)?

1. The Python code I changed will allocate less and be slightly faster when handling very small data sets. It should be nearly identical for large data sets.
2. Vectors are positional; they never expose row keys as part of their interface. They can be sliced positionally without needing to flatten the source table, assuming I understood the question correctly.

[jmao-denver]

The Python changes LGTM. Two questions:

1. what's the performance implication of this change?
2. w/o reading the Java changes, can ColumnVectors be sliced (without flattening)?

1. The Python code I changed will allocate less and be slightly faster when handling very small data sets. It should be nearly identical for large data sets.
2. Vectors are positional; they never expose row keys as part of their interface. They can be sliced positionally without needing to flatten the source table, assuming I understood the question correctly.

Yes，and I was thinking of using it as a way to give Python users the ability of reading the column data (single cell or range/slice) without needing to go through to_numpy which requires snapshotting the table first. Should we do that?

[rcaudy]

The Python changes LGTM. Two questions:

1. what's the performance implication of this change?
2. w/o reading the Java changes, can ColumnVectors be sliced (without flattening)?

1. The Python code I changed will allocate less and be slightly faster when handling very small data sets. It should be nearly identical for large data sets.
2. Vectors are positional; they never expose row keys as part of their interface. They can be sliced positionally without needing to flatten the source table, assuming I understood the question correctly.

Yes，and I was thinking of using it as a way to give Python users the ability of reading the column data (single cell or range/slice) without needing to go through to_numpy which requires snapshotting the table first. Should we do that?

I'm unsure. You're still dealing with a JNI access per cell if you just use get. Maybe you want a to_numpy on a slice that is aware of when it needs to use snapshots?

[jmao-denver]

This is what I had in mind very roughly:

class Table:
    def read_column(col: str, start: int,  end: Optional[int] = None) -> Union[int, float, bool, str, np.ndarray, jpy.JType]:
        ...

[chipkent]

Plotting and py LGTM. Let me know if I need to review more.

[lbooker42]

Skipped the python files, trusting @jmao-denver. Looks good, couple of comments for clarification.

[lbooker42]

You mentioned we might make this work chunky.

[chipkent]

Python and plotting LGTM

[deephaven-internal]

Labels indicate documentation is required. Issues for documentation have been opened:

Community: deephaven/deephaven-docs-community#227