ENH: access sliced dataframe from rolling.cov

[srkunze]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

In a current project, I iterate over df.rolling(window).cov(pairwise=True). Currently, I back-calculate from the index value of the cov() and the window offset what I suspect to be the start of the window. Then I slice the original df again into the window.

It would be great to iterate efficiently over the original df simultaneously with the cov values (and possibly with all the other window functions).

Feature Description

An idea off the top off my head:

for window, cov in df.rolling(window).roll("window", "cov_pairwise"):
    ...
    # window equals df.loc[start:end]
    # cov equals df.loc[start:end].cov()
    # start equals window.index[0]
    # end equals window.index[-1]
    ...

Alternative Solutions

I don't know any. Maybe there is already a way to do this.

Additionally, roll could allow efficient slicing to avoid useless calculations

for window, cov in df.rolling(window).roll("window", "cov_pairwise")[-1000:]:
    ...

Additional Context

No response

[MartinBraquet]

For increased clarity, could you please add a minimal and fully reproducible example of your current methodology and explain how your proposed feature would improve efficiency (in terms of time complexity)?