Lazy MEAN Aggregator doesn't support weights

[DPeterK]

Despite being a weighted aggregator, you can't actually pass weights to iris.analysis.MEAN when producing a lazy aggregation:

b = da.from_array(np.array([1,2,3,4,5]))
MEAN.lazy_aggregate(b, 0, weights=[2, 4, 1, 6, 3])

Traceback (most recent call last):
  File "/.../miniconda3/envs/scitools_dev/lib/python3.6/site-packages/IPython/core/interactiveshell.py", line 2963, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-42-60626d069686>", line 1, in <module>
    MEAN.lazy_aggregate(b, 0, weights=[2, 4, 1, 6, 3])
  File "/.../git/iris/lib/iris/analysis/__init__.py", line 471, in lazy_aggregate
    return self.lazy_func(data, axis, **kwargs)
  File "/.../git/iris/lib/iris/analysis/__init__.py", line 1460, in inner_stat
    dask_result = dask_stats_function(array, axis=axis, **kwargs)
TypeError: mean() got an unexpected keyword argument 'weights'

This functionality is not tested by iris.tests.unit.analysis.test_MEAN.

One solution to this would be to use da.average rather than da.mean - but note there might be an implementational problem with da.average when the input array is masked.

[valeriupredoi]

hey guys. we have this issue documented in ESMValTool - with specific examples and work started by Ben towards a solution, if it helps https://github.com/ESMValGroup/ESMValTool/issues/887

[bouweandela]

It looks like dask.array.ma.average is now implemented:
https://github.com/dask/dask/blob/44948eca09b046cc06169dbdfc3cf4a5750a2598/dask/array/ma.py#L262-L264

so I think the solution could be as simple as replacing da.mean by da.ma.average here:

iris/lib/iris/analysis/__init__.py

Lines 1486 to 1487 in 337f6f0

	MEAN = WeightedAggregator('mean', ma.average,
	lazy_func=_build_dask_mdtol_function(da.mean))

and pinning dask to an appropriately recent version, i.e. >= 1.1.0
http://docs.dask.org/en/latest/changelog.html#id16

[valeriupredoi]

good find @bouweandela 🍺