html_url,issue_url,id,node_id,user,created_at,updated_at,author_association,body,reactions,performed_via_github_app,issue
https://github.com/pydata/xarray/issues/3656#issuecomment-777878102,https://api.github.com/repos/pydata/xarray/issues/3656,777878102,MDEyOklzc3VlQ29tbWVudDc3Nzg3ODEwMg==,14808389,2021-02-12T00:05:51Z,2021-02-12T00:05:51Z,MEMBER,"I assume you figured this out somehow. If not, feel free to reopen.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,543825272
https://github.com/pydata/xarray/issues/3656#issuecomment-605525025,https://api.github.com/repos/pydata/xarray/issues/3656,605525025,MDEyOklzc3VlQ29tbWVudDYwNTUyNTAyNQ==,14808389,2020-03-28T21:58:40Z,2020-03-28T21:58:40Z,MEMBER,"is this still an issue, @mada0007? If it is, I tried a bit more (it is still incomplete since I don't really understand what happens, especially with the last groupby-agg)
```python
In [7]: def get_grps(s, axis, thresh=-1, Nmin=2): 
   ...:     s = xr.DataArray(s, dims=(""time"", ""lon"", ""lat"", ""window"")) 
   ...:     m = ( 
   ...:         xr.concat( 
   ...:             [ 
   ...:                 s.shift(window=-i).pipe(lambda arr: arr <= thresh) 
   ...:                 for i in range(Nmin) 
   ...:             ], 
   ...:             dim=""shifted"", 
   ...:         ) 
   ...:         .reduce(lambda arr, axis: np.logical_and.reduce(arr, axis=axis), dim=""shifted"") 
   ...:     ) 
   ...:     if Nmin > 1: 
   ...:         m = ( 
   ...:             m.where(lambda arr: arr == False, np.nan) 
   ...:             .ffill(limit=Nmin - 1, dim=""window"") 
   ...:             .fillna(False) 
   ...:         ) 
   ...:      
   ...:     gps = (m != m.shift(window=1)).cumsum(dim=""window"").where(m) 
   ...:      
   ...:     groups = s.groupby(gps.rename(""gps_groups"")) 
   ...:     return xr.where( 
   ...:         gps.isnull().all(dim=""window""), 
   ...:         0, 
   ...:         1,  # use `groups` here once it has the same size as `gps` 
   ...:     ) 
   ...:  
   ...: spi = xr.DataArray(dims=(""time"", ""lon"", ""lat""), data=np.random.randn(118, 18, 36)) 
   ...: spi.rolling(time=59).reduce(get_grps)
```
which mirrors your function except it does the calculation for all points in your space-time-grid at once.

Of course, once again I can't really check for errors since I don't understand what I'm computing :grin:","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,543825272
https://github.com/pydata/xarray/issues/3656#issuecomment-570056960,https://api.github.com/repos/pydata/xarray/issues/3656,570056960,MDEyOklzc3VlQ29tbWVudDU3MDA1Njk2MA==,14808389,2020-01-01T14:19:00Z,2020-01-01T14:19:00Z,MEMBER,"the reduce function needs to be a function that takes a 4D array (`time x lat x lon x window` in your case). It should use the elements of the `window` dimension (`1 x 1 x 1 x 59`) to calculate a `time x lat x lon` array.

The only thing wrong with your function is that it ignores the `axis` parameter: it can only work on 1D arrays (with a shape of `59` in this case).

A simple albeit horribly slow way to use your function without rewriting it would be `numpy.apply_along_axis` (slow because the function is called `time * lat * lon == 20995200` times):
```python
spi.rolling(time=59).reduce(lambda x, axis: np.apply_along_axis(get_grps, axis, x))
```

I'm not too familiar with writing numpy reduce functions so you might want to wait for the opinion of someone more experienced.

I started to rewrite your function using `xarray.DataArray` instead of `pandas.Series`, but I don't have time to finish that right now. Hopefully this points you in the right direction.
```python
In [32]: axis = 3
    ...: thresh = -1
    ...: Nmin = 2
    ...: # function code starts here
    ...: dim_names = [""time"", ""lat"", ""lon""]
    ...: # fails if axis == -1 but axis from reduce is always a positive number
    ...: dim_names.insert(axis, ""window"")
    ...: # use s = xr.DataArray(s, dims=dim_names) in the function
    ...: # so we can interactively write and try the function code
    ...: s = spi.rolling(time=59).construct(""window"")
    ...: shifted = xr.concat([s.shift({""window"": i}) for i in range(Nmin)], dim=""shifted"")
    ...: thresholded = shifted <= thresh
    ...: # xarray does not support `ufunc.reduce`, so we need to fall back to numpy arrays
    ...: m = r.copy(data=np.logical_and.reduce(thresholded.values, axis=0))
    ...: if Nmin > 1:
    ...:     m = m.where(~m).ffill(limit=Nmin - 1, dim=""window"").fillna(False)
    ...: gps = m != m.shift({""window"": 1}).cumsum().where(m)
```","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,543825272
https://github.com/pydata/xarray/issues/3656#issuecomment-569650603,https://api.github.com/repos/pydata/xarray/issues/3656,569650603,MDEyOklzc3VlQ29tbWVudDU2OTY1MDYwMw==,14808389,2019-12-30T11:18:12Z,2019-12-30T13:09:03Z,MEMBER,"the reason for the exception is that you are trying to convert the input data (a 4D `numpy.ndarray` of shape `324 x 180 x 360 x 59`) to a `pandas.Series` which only takes 1D data. You will have to rewrite your function to accept a 4D array and reduce it to a 3D array along the given axis (see `numpy` functions like `std` or `mean`).

We should definitely improve the documentation, right now the docstring of `DataArrayRolling.reduce` and `DatasetRolling.reduce` is not really clear on the requirements for the reduce function and the examples don't show the use of custom functions.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,543825272