Divide each element in a list by an int

[Liyixin95]

Description

I have a dataframe with a list column and an int column, then I want to divide these two columns like this:

import polars as pl
from polars import col

lf = pl.DataFrame({"a": [[1, 2, 3], [1, 3], [2, 4, 6]], "b": [4, 5, 6]}).lazy()

df = lf.with_columns(c=(col("a") / col("b")).alias("c")).collect()

print(df)

then I get an error:

Traceback (most recent call last):
  File "/root/miniconda3/envs/factor/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/root/miniconda3/envs/factor/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/root/python/ring_next/test/test.py", line 6, in <module>
    df = lf.with_columns(c=(col("a") / col("b")).alias("c")).collect()
  File "/root/miniconda3/envs/factor/lib/python3.10/site-packages/polars/lazyframe/frame.py", line 1937, in collect
    return wrap_df(ldf.collect())
polars.exceptions.ComputeError: cannot cast List type (inner: 'Int64', to: 'Float64')

[cmdlineluser]

To date, I have found:

It's allowed on structs, but you lose the list:

df.with_columns(
   pl.col("a").list.to_struct("max_width") / pl.struct("b")
)

# shape: (3, 2)
# ┌─────────────────────────┬─────┐
# │ a                       ┆ b   │
# │ ---                     ┆ --- │
# │ struct[3]               ┆ i64 │
# ╞═════════════════════════╪═════╡
# │ {0.25,0.5,0.75}         ┆ 4   │
# │ {0.2,0.6,null}          ┆ 5   │
# │ {0.333333,0.666667,1.0} ┆ 6   │
# └─────────────────────────┴─────┘

"Manual" broadcasting:

df.with_columns(
   (pl.col("a").flatten() / pl.col("b")).implode().over(pl.int_range(pl.len()))
)

# shape: (3, 2)
# ┌───────────────────────────┬─────┐
# │ a                         ┆ b   │
# │ ---                       ┆ --- │
# │ list[f64]                 ┆ i64 │
# ╞═══════════════════════════╪═════╡
# │ [0.25, 0.5, 0.75]         ┆ 4   │
# │ [0.2, 0.6]                ┆ 5   │
# │ [0.333333, 0.666667, 1.0] ┆ 6   │
# └───────────────────────────┴─────┘

Does anybody know if this is planned for list or maybe array types?

[mcrumiller]

Related is #14541. I definitely think simple arithmetic should be allowed on lists/arrays, and inter-series arithmetic on arrays of the same dtype.

[itamarst]

I will start on this once #17823 is merged.

[deanm0000]

This is a narrow case of #17496

[cmdlineluser]

This now works #19162

df = pl.DataFrame({"a": [[1, 2, 3], [1, 3], [2, 4, 6]], "b": [4, 5, 6]})

df.with_columns(c=(pl.col.a / pl.col.b))

# shape: (3, 3)
# ┌───────────┬─────┬───────────────────────────┐
# │ a         ┆ b   ┆ c                         │
# │ ---       ┆ --- ┆ ---                       │
# │ list[i64] ┆ i64 ┆ list[f64]                 │
# ╞═══════════╪═════╪═══════════════════════════╡
# │ [1, 2, 3] ┆ 4   ┆ [0.25, 0.5, 0.75]         │
# │ [1, 3]    ┆ 5   ┆ [0.2, 0.6]                │
# │ [2, 4, 6] ┆ 6   ┆ [0.333333, 0.666667, 1.0] │
# └───────────┴─────┴───────────────────────────┘