Trickery to operate on dataframe's columns without lambdas

fluent_pandas.py

import pandas as pd
import operator as op
from functools import partial
from datetime import datetime


def eval(node, df):
    if not isinstance(node, tuple | str):
        return node
    if isinstance(node, str):
        return df[node]
    operand = node[0]
    if operand == "lit":
        return node[1]
    elif operand == "call":
        fn, args, kwargs = node[1:]
        return eval(fn, df)(
            *[eval(arg.ast, df) if isinstance(arg, col) else arg for arg in args],
            **{k: eval(v, df) if isinstance(v, col) else v for k, v in kwargs.items()})
    else:
        return operand(*map(partial(eval, df=df), node[1:]))


class col:
    @classmethod
    def new(cls, fn, *args, **kwargs):
        return cls(("call", ("lit", fn), args, kwargs))

    def __init__(self, ast):
        self.ast = ast

    def call(self, *args, **kwargs):
        return col(("call", self.ast, args, kwargs))

    def __call__(self, df):
        return eval(self.ast, df)

    def __sub__(self, rhs):
        return col((op.sub, self.ast, rhs.ast if isinstance(rhs, col) else rhs))

    def __add__(self, rhs):
        return col((op.add, self.ast, rhs.ast if isinstance(rhs, col) else rhs))

    def __mul__(self, rhs):
        return col((op.mul, self.ast, rhs.ast if isinstance(rhs, col) else rhs))

    def __pow__(self, rhs):
        return col((op.pow, self.ast, rhs.ast if isinstance(rhs, col) else rhs))

    def __truediv__(self, rhs):
        return col((op.truediv, self.ast, rhs.ast if isinstance(rhs, col) else rhs))

    def __neg__(self):
        return col((op.neg, self.ast))

    def __getattr__(self, name):
        return col((getattr, self.ast, ("lit", name)))

data = [
    dict(name="Ian", height=1.82, weight=79, birth=datetime(1988, 3, 1)),
    dict(name="Giu", height=1.70, weight=58, birth=datetime(1987, 2, 28)),
    dict(name="Nana", height=1.90, weight=120, birth=datetime(1970, 1, 1)),
]

print(pd.DataFrame(data)
      .assign(bmi=col("weight") / col("height") ** 2,
              name=col("name").str.lower.call(),
              year=col("birth").dt.year,
              category=col.new(pd.cut,
                               col("bmi"),
                               bins=[0, 18.4, 24.9, 100],
                               labels=["underweight", "normal", "overweight"])
              ))