Rename `rle()` struct fields to `len` and `value` about polars HOT 3 OPEN

Agreed on the rename.

I don't think the index should be part of the RLE method by default. It is not an essential part of the RLE definition. Though possibly an include_index parameter would make sense - but please open a separate issue for that.

from polars.

@cmdlineluser I am tempted to also change the field order of the struct to value/len. That way it matches value_counts. What do you think?

EDIT: Nevermind, it's probably not a good idea as the standard RLE places len before value, e.g. 12W1B12W3B24W1B14W.

from polars.

Side note: Would it make sense for rle() to also return the row index? {index, value, len}

The particular use-case being wanting the original row index after performing a .filter()

df = pl.DataFrame({"foo": ["a", "a", "a", "b", "c", "c"]})

df.select(pl.col("foo").rle())
# shape: (3, 1)
# ┌───────────┐
# │ foo       │
# │ ---       │
# │ struct[2] │
# ╞═══════════╡
# │ {3,"a"}   │
# │ {1,"b"}   │
# │ {2,"c"}   │
# └───────────┘

We can calculate it from the length, but it's a little awkward:

(df.select(pl.col("foo").rle())
   .with_columns(
      index = pl.col("foo").struct["lengths"].cum_sum().shift().fill_null(0)
   )
   #.filter(...)
)
# shape: (3, 2)
# ┌───────────┬───────┐
# │ foo       ┆ index │
# │ ---       ┆ ---   │
# │ struct[2] ┆ i32   │
# ╞═══════════╪═══════╡
# │ {3,"a"}   ┆ 0     │
# │ {1,"b"}   ┆ 3     │
# │ {2,"c"}   ┆ 4     │
# └───────────┴───────┘

from polars.