Improve Series.cut / Expr.cut: Add number of bins as an alternative

[Julian-J-S]

Description

Pandas cut and qcut both allow to specify the "number of bins" (e.g. 4) as alternative to the actual bins as list (e.g. [2, 5, 10])

Idea

cut with number of bins N:

• create N bins of equal width

qcut with number of bins N:

• create N bins of equal probability

Pandas

Pandas qcut

pd.qcut([1, 1, 2, 2, 4 ,4, 9, 9], q=4, retbins=True)
# ([(0.999, 1.75], (0.999, 1.75], (1.75, 3.0], (1.75, 3.0], (3.0, 5.25], (3.0, 5.25], (5.25, 9.0], (5.25, 9.0]]
#  Categories (4, interval[float64, right]): [(0.999, 1.75] < (1.75, 3.0] < (3.0, 5.25] < (5.25, 9.0]],
#  array([1.  , 1.75, 3.  , 5.25, 9.  ]))

Pandas cut

pd.cut([1, 1, 2, 2, 4 ,4, 9, 9], bins=4, retbins=True)
# ([(0.992, 3.0], (0.992, 3.0], (0.992, 3.0], (0.992, 3.0], (3.0, 5.0], (3.0, 5.0], (7.0, 9.0], (7.0, 9.0]]
#  Categories (4, interval[float64, right]): [(0.992, 3.0] < (3.0, 5.0] < (5.0, 7.0] < (7.0, 9.0]],
#  array([0.992, 3.   , 5.   , 7.   , 9.   ]))

Polars

Polars qcut (almost equal except that polars uses "inf" bounds at the edges)

pl.Series([1, 1, 2, 2, 4, 4, 9 ,9]).qcut(4, include_breaks=True).struct.unnest()
shape: (8, 2)
┌─────────────┬──────────────┐
│ break_point ┆ category     │
│ ---         ┆ ---          │
│ f64         ┆ cat          │
╞═════════════╪══════════════╡
│ 1.75        ┆ (-inf, 1.75] │
│ 1.75        ┆ (-inf, 1.75] │
│ 3.0         ┆ (1.75, 3]    │
│ 3.0         ┆ (1.75, 3]    │
│ 5.25        ┆ (3, 5.25]    │
│ 5.25        ┆ (3, 5.25]    │
│ inf         ┆ (5.25, inf]  │
│ inf         ┆ (5.25, inf]  │
└─────────────┴──────────────┘

Polars cut 💥

• Option to specify number of bins is missing

[stinodego]

Thanks for the input, but we are already planning to overhaul the API of cut and qcut. See #10468

I'll close this in favor of that one.