Optional ReadonlynameThe caller's own name for this output.
Set from the key when a request uses outputs. It rides back on the
Fact and OutputInfo so a consumer reads the name it
chose rather than parsing a derived id ([PND-PROCSLOT]).
ReadonlyonOptional ReadonlyoutputWhich output of a multi-output op. Defaults to all of them.
Only meaningful for a column-producing node — a fact has one shape.
A selector says what to surface, not what to compute.
It used to say both:
{ on, reduce: 'percentileRank' }named a node and a fold to run over it, from an enum this file owned. The fold is a node now ([PND-PROCFOLD]), so a selector's whole job is to point at one and name it — and what comes back is whatever that node produces: a fact from a fold, columns from anything else.That is the simplification the change was for. There is no
reduce, nopoints, and nocolumns: true; asking for a bounded sample means pointing at ashapenode, which is a thing with an id that caches.