fix(python): dont show wrong inefficient apply warning when taking numpy func of constant

[MarcoGorelli]

I'd missed this when reviewing, but a "nonsense" warning is currently shown in some cases:

df = pl.DataFrame({"a": [1, 2, 3]})
df.select(pl.col("a").apply(lambda x: np.sin(3) + x))

shows

-  pl.col("a").apply(lambda x: ...)
+  3.sin() + pl.col("a")

I think we need to be extra-defensive here, and match as explicitly as we can - I've added param_name_index to the function so we can check that parameter name is exactly in the position we expect it to be

---

with this PR, the code above wouldn't display any warning, as np.sin(3) isn't a simple-enough constant to be considered (yet!)

[MarcoGorelli]

#10085 (comment) 😉

[MarcoGorelli]

moving these here from test_inefficient_apply, we should probably be testing the "no-op" cases across all supported python versions too

[alexander-beedie]

Hmm, turns out it only takes two/three lines to support this properly; as the result will resolve to a const it can be integrated more or less 'as is' - see what you think? 😅 Commit: f44b1ed.

Tried it out with lambda x: numpy.cos(3) + x - abs(-1).

(Unrelated change: removed len for now as it needs to be mapped to str.lengths).

[MarcoGorelli]

well you're a bit good, aren't you? 🙌

[MarcoGorelli]

closing in favour of #10104

[alexander-beedie]

well you're a bit good, aren't you?

I think you mean "a bit OCD" ... 🤣