consider using sort(key=...) function

[sgratzl]

it looks quite difficult how the sorting is handled:

delphi-epidata/src/acquisition/covidcast/database.py

Lines 330 to 335 in d6dd046

	sorting_fields = "data_source signal time_type geo_type".split()
	sortable_fields_fn = lambda x: [(field, x[field]) for field in sorting_fields]
	prepended_sortables_fn = lambda x: sortable_fields_fn(x) + list(x.items())
	tuple_representation = list(map(prepended_sortables_fn, meta))
	tuple_representation.sort()
	meta = list(map(dict, tuple_representation)) # back to dict form

can't the key function as in .sort(key=lambda x: (x['data_source'], x['signal'], x['time_type'], x['geo_type'])) be used instead?

[dshemetov]

@melange396 does this have to do something with threads? It seems like all the thread work is done before this chunk?

[melange396]

yeah, this is done post-threading (because the threaded workers arent likely to retain the original ordering). .sort(key=) would be a much less ugly way to do this.

[nolangormley]

Looking into this. Seems like something we could clear up during the v4 integration.

[dshemetov]

This meta computations code is getting rewritten in #947, so it will be fixed there. (For context, that work will go into #646, which in turn should be merged in after v4. It'll look something like this when done.)