BUG: Fix rolling median and quantile with closed='left' and closed='neither' (#26005)

[ihsansecer]

• Fixes issues with rolling median (and also quantile) closed='left' and closed='neither' in offset-based rolling window, multiple issues with closed='left' #26005
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Here is my understanding of how it should work:

*Same as test cases, there is one day between each element and window size is 3 days.

For closed='both' (() denotes Nth window and - is just to state window is out of ):

N=0 | (-, -, - [0), 1, 2, 3, 4, 5, 6, 7, 8, 9]
N=1 | -, (-, - [0, 1), 2, 3, 4, 5, 6, 7, 8, 9]
N=2 | -, -, (- [0, 1, 2), 3, 4, 5, 6, 7, 8, 9]
N=3 | -, -, - [(0, 1, 2, 3), 4, 5, 6, 7, 8, 9]
...

For closed='right':

N=0 | -, (-, - [0), 1, 2, 3, 4, 5, 6, 7, 8, 9]
N=1 | -, -, (- [0, 1), 2, 3, 4, 5, 6, 7, 8, 9]
N=2 | -, -, - [(0, 1, 2), 3, 4, 5, 6, 7, 8, 9]
N=3 | -, -, - [0, (1, 2, 3), 4, 5, 6, 7, 8, 9]
...

For closed='left':

N=0 | (-, -, -) [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
N=1 | -, (-, - [0), 1, 2, 3, 4, 5, 6, 7, 8, 9]
N=2 | -, -, (- [0, 1), 2, 3, 4, 5, 6, 7, 8, 9]
N=3 | -, -, - [(0, 1, 2), 3, 4, 5, 6, 7, 8, 9]
...

For closed='neither':

N=0 | -, (-, -) [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
N=1 | -, -, (- [0), 1, 2, 3, 4, 5, 6, 7, 8, 9]
N=2 | -, -, - [(0, 1), 2, 3, 4, 5, 6, 7, 8, 9]
N=3 | -, -, - [0, (1, 2), 3, 4, 5, 6, 7, 8, 9]
...

[TomAugspurger]

Thank @ihsansecer. It looks like #26005 reported multiple issues. Do your changes fix the memory skiplist issues? If so, can you add tests / release note? Otherwise, edit the original post to remove the "closes", so that it remains open.

cc @WillAyd

[codecov]

Codecov Report

Merging #26910 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #26910      +/-   ##
==========================================
- Coverage   91.97%   91.97%   -0.01%     
==========================================
  Files         180      180              
  Lines       50756    50760       +4     
==========================================
  Hits        46685    46685              
- Misses       4071     4075       +4

Flag	Coverage Δ
#multiple	90.56% <ø> (ø)	⬆️
#single	41.82% <ø> (-0.09%)	⬇️

Impacted Files	Coverage Δ
pandas/io/gbq.py	88.88% <0%> (-11.12%)	⬇️
pandas/core/frame.py	96.89% <0%> (-0.12%)	⬇️
pandas/core/internals/construction.py	95.95% <0%> (+0.02%)	⬆️
pandas/core/dtypes/cast.py	90.56% <0%> (+0.03%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b9b081d...51e89c4. Read the comment docs.

[ihsansecer]

@TomAugspurger not yet checked issues with memory skiplist and max function but I am planning to have a look and open another PR. Thanks

[jreback]

comments

[ihsansecer]

It seems quantile is also producing unexpected results same as stated in the issue. If there is no problem with this fix I will apply the same to quantile function.

[jreback]

yep, can you merge master and push again here.

[ihsansecer]

@jreback Merged and changed roll_quantile as mentioned

[jreback]

thanks @ihsansecer nice patch!