[Bug]: DataFrame API Series.value_counts(bins=...) always fails with TypeError #40503

What happened?

DeferredSeries.value_counts has a non-parallel path for bins= (its docstring says it needs to collect the whole dataset), but it can't be used at all. The argument list [self._expr] ended up inside the lambda, so creating the expression fails right away:

import apache_beam as beam
import pandas as pd
from apache_beam.dataframe.convert import to_dataframe, to_pcollection

values = [1.0, 2.0, 2.0, 3.0, 5.0, 8.0, 8.0, 9.0]
print(pd.Series(values, name='x').value_counts(bins=3))

with beam.Pipeline() as p:
  df = to_dataframe(p | beam.Create([beam.Row(x=v) for v in values]))
  with beam.dataframe.allow_non_parallel_operations():
    counts = df.x.value_counts(bins=3)
  _ = (
      to_pcollection(counts, include_indexes=True, yield_elements='pandas')
      | beam.Map(print))

pandas:

(0.991, 3.667]    4
(6.333, 9.0]      3
(3.667, 6.333]    1
Name: count, dtype: int64

Beam:

TypeError: ComputedExpression.__init__() missing 1 required positional argument: 'args'

Fixing the argument list alone isn't enough, because the proxy is computed from an empty Series and pandas can't compute bins for empty data (ValueError: Cannot cut empty array).

This is different from #20903, which asks for a distributed implementation.

Seen on Beam 2.77.0 and on master (7d93ef8), pandas 2.3.3, Python 3.12.

Issue Priority

Priority: 2 (default / most bugs should be filed as P2)

Issue Components

• Component: Python SDK