Clarify the use of period = `'1i'` when passing an int-type `index_column` in `.rolling`

[mkleinbort-ic]

Clarify the use of period = `'1i'` when passing an int-type `index_column` in `.rolling`

[mkleinbort-ic]

Yes, but to me the part that got me confused was

'3i'

does that mean 3 consecutive rows? or rows where the integers are within a window of 3

[cmdlineluser]

Yeah, I recently asked if your final example should raise an exception:

• #14722

[MarcoGorelli]

or rows where the integers are within a window of 3

This is it

So, if you have

In [13]: df
Out[13]:
shape: (5, 2)
┌───────┬───────┐
│ index ┆ value │
│ ---   ┆ ---   │
│ i64   ┆ i64   │
╞═══════╪═══════╡
│ 0     ┆ 1     │
│ 4     ┆ 4     │
│ 5     ┆ 2     │
│ 6     ┆ 4     │
│ 8     ┆ 1     │
└───────┴───────┘

Then the windows are:

• (-2, 0]
• (2, 4]
• (3, 5]
• (4, 6]
• (6, 8]

(where ( means "open interval" and ] "closed interval")

In [14]: df.rolling('index', period='2i').agg(pl.col('index').alias('aggregated'))
Out[14]:
shape: (5, 2)
┌───────┬────────────┐
│ index ┆ aggregated │
│ ---   ┆ ---        │
│ i64   ┆ list[i64]  │
╞═══════╪════════════╡
│ 0     ┆ [0]        │
│ 4     ┆ [4]        │
│ 5     ┆ [4, 5]     │
│ 6     ┆ [5, 6]     │
│ 8     ┆ [8]        │
└───────┴────────────┘

[mkleinbort-ic]

Which is great! But could be added to the documentation to make things clear to new users.

[MarcoGorelli]

yup - fancy making a PR?