Counting Unique Values by group in an xarray Dataset

Question

I have an xarray dataset of shipment line items called shipDS:

<xarray.Dataset>
Dimensions:   (lineItem: 10)
Dimensions without coordinates: lineItem
Data variables:
    shipID    (lineItem) int64 1 1 2 3 4 4 5 5 5 5
    prodID    (lineItem) int64 90 92 92 90 90 91 92 93 94 95
    category  (lineItem) <U1 'A' 'B' 'B' 'A' 'A' 'A' 'B' 'C' 'D' 'D'

Each line item has a category. The question is how many unique categories were included in every shipment (i.e. shipID)?

Here is the code to create the xarray dataset:

import xarray as xr

## construct dataset of shipments
## each shipment might have multiple products
shipIDs = [1,1,2,3,4,4,5,5,5,5]
productIDs = [90,92,92,90,90,91,92,93,94,95]
category = ["A","B","B","A","A","A","B","C","D","D"]

shipDS = xr.Dataset(data_vars = 
    {
        "shipID": ("lineItem", shipIDs),
        "prodID": ("lineItem", productIDs),
        "category": ("lineItem", category)
     }
)

## How many product categories were in every shipment (i.e. in each shipID)?

To answer this in Pandas, I would do the following:

## pandas answer
(
    shipDS
    .to_pandas()
    .groupby(["shipID","category"])
    .count()
    .reset_index()
    .groupby("shipID")
    .agg(numUniqCategories=('category', 'count'))
)

and get the following result:

    numUniqCategories
shipID  
1   2
2   1
3   1
4   1
5   3

I am trying to become an xarray expert, but cannot seem to figure out the right way to do this solely in xarray.

Answer 1

At least for now, there's no built-in xarray method to count unique values. Maybe this discussion might help: xarray-forum. The most viable way to do this is your pandas solution, although it's not scalable for huge xr.Datasets.