calculating cumsums on a groupby object

[tommylees112]

How do I go about calculating cumsums on a groupby object?

I have a Dataset that looks as the following:

lat = np.linspace(-5.175003, 5.9749985, 224)
lon = np.linspace(33.524994, 42.274994, 176)
time = pd.date_range(start='1981-01-31', end='2019-04-30', freq='M')
data = np.random.randn(len(time), len(lat), len(lon))
dims = ['time', 'lat', 'lon']
coords = {'time': time, 'lat': lat, 'lon': lon}

ds = xr.Dataset({'precip': (dims, data)}, coords=coords)

Out[]:
<xarray.Dataset>
Dimensions:  (lat: 224, lon: 176, time: 460)
Coordinates:
  * time     (time) datetime64[ns] 1981-01-31 1981-02-28 ... 2019-04-30
  * lat      (lat) float64 -5.175 -5.125 -5.075 -5.025 ... 5.875 5.925 5.975
  * lon      (lon) float64 33.52 33.57 33.62 33.67 ... 42.12 42.17 42.22 42.27
Data variables:
    precip   (time, lat, lon) float64 0.006328 0.2969 1.564 ... 0.6675 2.32

I need to groupby year and calculate the cumsum for each year. That way I will have a value for each month (timestep) and each pixel (lat - lon pair).

But the cumsum operation doesn't work on a groupby object

ds.groupby('time.year').cumsum(dim='time')

Out[]:
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-12-dceee5f5647c> in <module>
      9 display(ds_)
     10 
---> 11 ds_.groupby('time.year').cumsum(dim='time')

AttributeError: 'DatasetGroupBy' object has no attribute 'cumsum'

Is there a work around?

INSTALLED VERSIONS
------------------
commit: None
python: 3.7.0 | packaged by conda-forge | (default, Nov 12 2018, 12:34:36) 
[Clang 4.0.1 (tags/RELEASE_401/final)]
python-bits: 64
OS: Darwin
OS-release: 18.2.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: en_US.UTF-8
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8
libhdf5: 1.10.4
libnetcdf: 4.6.2

xarray: 0.12.2
pandas: 0.24.2
numpy: 1.16.4
scipy: 1.3.0
netCDF4: 1.5.1.2
pydap: None
h5netcdf: None
h5py: 2.9.0
Nio: None
zarr: None
cftime: 1.0.3.4
nc_time_axis: None
PseudonetCDF: None
rasterio: 1.0.17
cfgrib: 0.9.7
iris: None
bottleneck: 1.2.1
dask: 1.2.2
distributed: 1.28.1
matplotlib: 3.1.0
cartopy: 0.17.0
seaborn: 0.9.0
numbagg: None
setuptools: 41.0.1
pip: 19.1
conda: None
pytest: 4.5.0
IPython: 7.1.1
sphinx: 2.0.1

[dcherian]

I wonder if this is as easy as adding ops.inject_cum_methods(Dataset.groupby) at the end of core/groupby.py?

[shoyer]

It looks like ds.groupby('time.year').apply(lambda x: x.cumsum(dim='time')) mostly works for now.

But yes, it would be great to add this.

[dcherian]

@tommylees112 Are you up for sending in a PR. It's an easy fix...

[tommylees112]

Would love to! Sorry have been away this weekend. Do i just clone the repo write the code and send in a PR in a new branch?

(first PR on a public repo!)

[nbren12]

Xarray has a pretty extensive contributor's guide that you might find helpful. In short, the way to contribute changes is to create your own fork of xarray, commit/push some changes, and finally submit a pull request (PR).

[stale]

In order to maintain a list of currently relevant issues, we mark issues as stale after a period of inactivity

If this issue remains relevant, please comment here or remove the stale label; otherwise it will be marked as closed automatically