ENH: Sliding window by value interval on non-time index?

[Hoeze]

Is your feature request related to a problem?

I would like to compute the number of values in a sliding window like this:

import pandas as pd
df = pd.DataFrame({"group": ["A", "B"] * 5, "distance": [0.1, 0.3, 1.7, 3.2, 2.6, 3.7, 8.3, 0.6, 5.9, 6.3]})
df

Out[5]: 
  group  distance
0     A       0.1
2     A       1.7
4     A       2.6
8     A       5.9
6     A       8.3
1     B       0.3
7     B       0.6
3     B       3.2
5     B       3.7
9     B       6.3

# possible API:
df.groupby("group").sliding(on="distance", window_size=3.0, step_size=2.0, min=0, max=10).count()

                   count
group  distance
A      [0, 3)      3
       [2, 5)      1
       [4, 7)      1
       [6, 9)      1
       [8, 11)     0
B      [0, 3)      2
       [2, 5)      2
       [4, 7)      1
       [6, 9)      0
       [8, 11)     0

Describe the solution you'd like

There should be some sliding window function that slices by value instead of rows:

def sliding(
        self,
        on: str, # 
        window_size: numeric,
        step_size: numeric=None, # equals to window size by default
        min: numeric=None, # minimum value in group by default 
        max: numeric=None, # maximum value in group by default
    ):
    pass

API breaking implications

No idea, maybe integrate to rolling?

Describe alternatives you've considered

1. Create IntervalArray:

intervals = pd.arrays.IntervalArray.from_arrays(np.arange(10), np.arange(10) + 2)
Out[53]: 
<IntervalArray>
[(0, 2], (1, 3], (2, 4], (3, 5], (4, 6], (5, 7], (6, 8], (7, 9], (8, 10], (9, 11]]

2. Join on intervals: No idea how to do this...

Additional background

This question also appeared on StackOverflow:
https://stackoverflow.com/questions/43538064/how-to-do-sliding-window-by-value-interval-on-non-time-index-in-pandas

[mroeschke]

Thanks for the request, but it appears this request hasn't gotten much interest over the years so closing