Support comparison operations for list types

[jdcla]

Description

Support comparison operations for list types. This feature would circumvent the subsequent use of explode() and agg(), which is memory intensive.

    df = df = pl.DataFrame({
        'x1': [[1,2],[3,4]],
        'x2': [1,4]
    })

    df.with_columns(
        out = pl.col('x1').list.ge(pl.col('x2'))
    )
# OR
    df.with_columns(
        out = pl.col('x1') >= pl.col('x2')
    )

    # Desired output: 
    shape: (2, 3)
    ┌───────────┬─────┬───────────────┐
    │ x1        ┆ x2  ┆ out           │
    │ ---       ┆ --- ┆ ---           │
    │ list[i64] ┆ i64 ┆ list[f64]     │
    ╞═══════════╪═════╪═══════════════╡
    │ [1, 2]    ┆ 1   ┆ [True, True]  │
    │ [3, 4]    ┆ 4   ┆ [False, True] │
    └───────────┴─────┴───────────────┘

Trying to run pl.col('x1') >= pl.col('x2') or pl.col('x1').ge(pl.col('x2')) throws a SchemaError, referencing the difference in types for x1 and x2.

[thomasfrederikhoeck]

I guess this is very similar to #10918

[cmdlineluser]

List -> List arithmetic was part of today's 1.8 release thanks to the work in #17823

df.with_columns(pl.col.x1 - pl.col.x2.repeat_by(pl.col.x1.list.len()))
# shape: (2, 2)
# ┌───────────┬─────┐
# │ x1        ┆ x2  │
# │ ---       ┆ --- │
# │ list[i64] ┆ i64 │
# ╞═══════════╪═════╡
# │ [0, 1]    ┆ 1   │
# │ [-1, 0]   ┆ 4   │
# └───────────┴─────┘

Comparisons panic though - not sure if there have been discussed yet?

df.with_columns(pl.col.x1 > pl.col.x2.repeat_by(pl.col.x1.list.len()))
# thread '<unnamed>' panicked at crates/polars-core/src/chunked_array/comparison/mod.rs:690:9:
# not implemented
# PanicException: not implemented

I think all the different behaviours have popped up across a few different previous feature requests.

The latest updates I have seen are here:

• Simple arithmetic operations on the "list" type columns  #8006 (comment)