feat(rust, python, cli): Add ignore_nulls for list.join

[reswqa]

This should be a breaking change as null value is no longer be treated as a string null.

[reswqa]

Rebased on master to resolve the conflict. And @stinodego fancy taking a look at this? The changes are relatively minor, but breaking compatibility.

[ritchie46]

I think we must wait until a breaking release then. :/

[reswqa]

I think we must wait until a breaking release then.

Make sense, we should merge this when 1.0 is about to be released. Are we close to it?

[stinodego]

Just to be clear, the existing behavior is:

import polars as pl

s = pl.Series([["a", None, "b"], ["c", None]])
result = s.list.join("")
print(result)

shape: (2,)
Series: '' [str]
[
        "anullb"
        "cnull"
]

That looks broken to me and I can't imagine anyone relies on this behavior. In my opinion, this PR can be merged as a fix.

[alexander-beedie]

@reswqa: looks like the latest polars-sql merges created a conflict; I can fix that when I get home tonight 👌

[reswqa]

That looks broken to me and I can't imagine anyone relies on this behavior. In my opinion, this PR can be merged as a fix.

Ohh, It occurs to me that our introduction of the ignore_nulls to str.concat is also treated as a bug-fix RP.

[alexander-beedie]

Resolved merge conflicts / rebased 👍

[alexander-beedie]

That looks broken to me and I can't imagine anyone relies on this behavior. In my opinion, this PR can be merged as a fix.

@stinodego: Yup; it also enables a definite fix on the SQL side (for slightly different reasons - concat and concat_ws both require use of the new ignore_nulls=True option to match standard SQL behaviour, although we might still need to hook that up - will check that now); I'd say go for it.

[alexander-beedie]

And one last rebase (following #13877).
Once CI succeeds, let's merge 👍