fix: slice expr can be taken in cse

[reswqa]

This fixes #11594.

I'm not 100% sure, but I feel that slice can be reused by cse.

---

Besides:

There should be other bug in reusing CSE that can cause the following plan to occur. I will delve deeper into how to fix this in next PR.

SELECT [

col("__POLARS_CSER_14783041659764827719").slice(offset=1, length=[(count()) - (1)]).alias("1"),

col("__POLARS_CSER_14783041659764827719").slice(offset=2, length=[(count()) - (1)]).alias("2"),

[(count()) - (1)].alias("__POLARS_CSER_14783041659764827719")] FROM

  DF ["a", "b"]; PROJECT 1/2 COLUMNS; SELECTION: "None"

[ritchie46]

I think it should. Why did this panic?

[reswqa]

@ritchie46

TBH, I can not reproduce panic. But the output is not correct indeed.

>>> df = pl.DataFrame({
...     "a": [1,2,1,2,1,2],
...     "b": [None, None, 1,1,1,1]
... })
>>> df.lazy().select(
...     pl.col("a").slice(offset = 1, length=pl.count() - 1).alias("1"),
...     pl.col("a").slice(offset = 1, length=pl.count() - 1).alias("2")
... ).collect()
shape: (0, 2)
┌─────┬─────┐
│ 1   ┆ 2   │
│ --- ┆ --- │
│ u32 ┆ u32 │
╞═════╪═════╡
└─────┴─────┘

It seems that col ("a") has been mistakenly replaced with cse expr

SELECT [

col("__POLARS_CSER_14783041659764827719").slice(offset=1, length=[(count()) - (1)]).alias("1"),

col("__POLARS_CSER_14783041659764827719").slice(offset=1, length=[(count()) - (1)]).alias("2"),

[(count()) - (1)].alias("__POLARS_CSER_14783041659764827719")] FROM

  DF ["a", "b"]; PROJECT 1/2 COLUMNS; SELECTION: "None"

[reswqa]

All right I think I found the culprit, will create PR to fix this.