A reference to a node in the graph under construction.
The fold methods add a node and hand back a handle to it, rather
than producing a selector as they used to. That is the whole shape of
[PND-PROCFOLD] at the authoring layer: z.percentileRank() always
read like an op call, and now it is one.
The derived slot is <slot>:<fold> — with the fold's params folded in
when it has any, e.g. <slot>:shape(points=100). Deterministic in all
three, so calling .last() twice on the same node returns the same
node rather than colliding, while shape({points: 20}) and
shape({points: 100}) are two nodes rather than the second silently
answering with the first's 20. Safe against a source column name,
which cannot contain a colon.
A reference to a node in the graph under construction.
The fold methods add a node and hand back a handle to it, rather than producing a selector as they used to. That is the whole shape of [PND-PROCFOLD] at the authoring layer:
z.percentileRank()always read like an op call, and now it is one.The derived slot is
<slot>:<fold>— with the fold's params folded in when it has any, e.g.<slot>:shape(points=100). Deterministic in all three, so calling.last()twice on the same node returns the same node rather than colliding, whileshape({points: 20})andshape({points: 100})are two nodes rather than the second silently answering with the first's 20. Safe against a source column name, which cannot contain a colon.