DataFrame.join_asof incorrect result for by set to tuples

[hleumas]

Checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

import polars as pl

df1 = pl.DataFrame({
    'n': [10, 20, 30, 40, 50, 60],
    'id1': [0, 0, 3, 3, 5, 5],
    'id2': [1, 2, 1, 2, 1, 2],
    'x': ['a', 'b', 'c', 'd', 'e', 'f'],
}).sort(by='n')

df2 = pl.DataFrame({
    'n': [25, 8, 5, 23, 15, 35],
    'id1': [0, 0, 3, 3, 5, 5],
    'id2': [1, 2, 1, 2, 1, 2],
    'y': ['A', 'B', 'C', 'D', 'E', 'F'],
}).sort(by='n')

print('Incorrect join_asof')
print(df1.join_asof(df2, on='n', by=('id1', 'id2')))

print('Correct join_asof')
print(df1.join_asof(df2, on='n', by=['id1', 'id2']))

Issue description

Argument by of join_asof seems to be incorrectly ignored when tuple is supplied. Instead of expected result, we get this:

>>> df1.join_asof(df2, on='n', by=('id1', 'id2'))

shape: (6, 7)
┌─────┬─────┬─────┬─────┬───────────┬───────────┬─────┐
│ n   ┆ id1 ┆ id2 ┆ x   ┆ id1_right ┆ id2_right ┆ y   │
│ --- ┆ --- ┆ --- ┆ --- ┆ ---       ┆ ---       ┆ --- │
│ i64 ┆ i64 ┆ i64 ┆ str ┆ i64       ┆ i64       ┆ str │
╞═════╪═════╪═════╪═════╪═══════════╪═══════════╪═════╡
│ 10  ┆ 0   ┆ 1   ┆ a   ┆ 0         ┆ 2         ┆ B   │
│ 20  ┆ 0   ┆ 2   ┆ b   ┆ 5         ┆ 1         ┆ E   │
│ 30  ┆ 3   ┆ 1   ┆ c   ┆ 0         ┆ 1         ┆ A   │
│ 40  ┆ 3   ┆ 2   ┆ d   ┆ 5         ┆ 2         ┆ F   │
│ 50  ┆ 5   ┆ 1   ┆ e   ┆ 5         ┆ 2         ┆ F   │
│ 60  ┆ 5   ┆ 2   ┆ f   ┆ 5         ┆ 2         ┆ F   │
└─────┴─────┴─────┴─────┴───────────┴───────────┴─────┘

Expected behavior

When supplied a list, we get the correct result:

>>> df1.join_asof(df2, on='n', by=['id1', 'id2'])

shape: (6, 5)
┌─────┬─────┬─────┬─────┬──────┐
│ n   ┆ id1 ┆ id2 ┆ x   ┆ y    │
│ --- ┆ --- ┆ --- ┆ --- ┆ ---  │
│ i64 ┆ i64 ┆ i64 ┆ str ┆ str  │
╞═════╪═════╪═════╪═════╪══════╡
│ 10  ┆ 0   ┆ 1   ┆ a   ┆ null │
│ 20  ┆ 0   ┆ 2   ┆ b   ┆ B    │
│ 30  ┆ 3   ┆ 1   ┆ c   ┆ C    │
│ 40  ┆ 3   ┆ 2   ┆ d   ┆ D    │
│ 50  ┆ 5   ┆ 1   ┆ e   ┆ E    │
│ 60  ┆ 5   ┆ 2   ┆ f   ┆ F    │
└─────┴─────┴─────┴─────┴──────┘

Installed versions

--------Version info---------
Polars:              0.18.13
Index type:          UInt32
Platform:            macOS-13.4.1-arm64-arm-64bit
Python:              3.11.4 (main, Jun 20 2023, 17:23:00) [Clang 14.0.3 (clang-1403.0.22.14.1)]

----Optional dependencies----
adbc_driver_sqlite:  <not installed>
cloudpickle:         <not installed>
connectorx:          <not installed>
deltalake:           <not installed>
fsspec:              <not installed>
matplotlib:          3.7.2
numpy:               1.24.4
pandas:              2.0.3
pyarrow:             12.0.1
pydantic:            <not installed>
sqlalchemy:          <not installed>
xlsx2csv:            <not installed>
xlsxwriter:          <not installed>

[cmdlineluser]

Looks like it happens here:

polars/py-polars/polars/lazyframe/frame.py

Lines 3013 to 3018 in c4ac859

	if isinstance(by, str):
	by_left_ = [by]
	by_right_ = [by]
	elif isinstance(by, list):
	by_left_ = by
	by_right_ = by

There is an explicit check for str and list

[mcrumiller]

@cmdlineluser good find, the typing suggests we should take a sequence, which a tuple definitely is.

[zundertj]

Thank you for reporting this, I have raised a fix in #10447.

[hleumas]

Thanks a lot everybody! I must say wow, you're reacting super fast.