feat: Make pyo3_arrow::table::Table.append_column support Array

[surister]

Fixes #413

Additionally, as you can see in the unit test, ArrayReader is also supported, because it already had everything implemented in AnyArray.

I haven't touched the change-log since it looks automated to me.

Question: To have a coherent API, add_column should receive the same treatment, wtf?

[surister]

Hmm it passes locally, maybe related to #431?

[kylebarron]

This looks great! Thank you! Just a couple comments

[kylebarron]

And yes ideally add_column would also support an array

[kylebarron]

And yes you can change the Cargo.toml here to point arro3 to the local path pyo3-arrow

[surister]

And yes ideally add_column would also support an array

Soon™ PR.

And yes you can change the Cargo.toml here to point arro3 to the local path pyo3-arrow

So if I get this right, I change the local path to pyo3-arrow in this PR, and when you release do you release pyo3-arrow and then in arro3 turn it back on as it is right now? Do we have to do this back and forth between development and releases just for CI to pick up the changes? Or do you manually squash the changes and cherry pick out the path change

[kylebarron]

I tried to explain this here: https://github.com/kylebarron/arro3/blob/main/DEVELOP.md#updating-pyo3-version

During development we can and should use a path dependency.

Then any time we update with a new pyo3 version we have to do a dance where we temporarily set the pyo3-arrow version to the git hash.

For arro3 releases it's fine to use path dependencies, because they're not getting pushed to crates.io; it's only when we need to ensure all dependencies are using the same pyo3 version.

[kylebarron]

That new test looks great

Just one more comment

[kylebarron]

Perfect thank you!