Description

Now that Expr.index_of is a thing, the next step is adding something similar to lists.

Potential APIs

Thing is, there are actually multiple different multiple APIs people might want:

Option 1: For each list in the Series/Column, find the index of the value in that list:

>>> pl.Series([[1, 3], [2, 4], [5, 1]).list.index_of_or_maybe_some_other_name(1)
shape: (3,)
Series: '' [i64]
[
        0
        null
        1
]

This is what @thobai requested when they filed #5503.

Option 1B: Like option 1, but supports a expression-oriented version

df.select(pl.col("list_of_floats").index_of_or_maybe_some_other_name(pl.col("floats"))

and search each list in the list_of_floats column for the corresponding value taken from the floats column. I think this is just a superset of option 1, and is what a G-Research user would like.

Option 2: Find the index of the first list that contains the value:

>>> pl.Series([[1, 3], [2, 4], [5, 1]).list.index_of_or_maybe_some_other_name(1)
0

Option 2 can be implemented on top of option 1, albeit inefficiently, so providing it as an exposed feature is an optimization, not a necessity. Perhaps this API if ever implemented should be called .list.first_index_of().

Initial questions for maintainers

1. Which option should be implemented? I'm inclined towards 1B since that is the most general purpose, and meets requests of multiple users.
2. What should the API be called? I'm inclined towards calling option 1/1B .list.index_of(), and option 2 .list.first_index_of().
3. For implementation of option 1/1B, amortized_iter() seems like the easy approach?