By default, no partition column, no sort and no window function yet
Optionalrest: JsonObjectReadonlyaggregationsThe window functions, each a new column
ReadonlycolumnsThe partition columns, as listed; none is a window over all the rows
ReadonlyidUnique within the query and shown to users, e.g. filter101
ReadonlykeyAssigned at construction from a monotonic sequence: the node's identity for change detection
ReadonlyrestThe keys of the node's saved JSON that this version does not know, kept so a re-save writes them back (PLAN §10.3); edits keep them too
ReadonlysortsThe window's sort keys, most significant first; may be empty
Whether edits may change which nodes feed this one. Not for a node whose port semantics are unknown, such as an Unknown node.
Where its window functions are: a window, sorted when it has sort keys
Whether the node takes rows by their order, as a Limit does: the emitter then writes its input's order as a sort just before it (PLAN §11.4)
Labels for the input ports, shown on the edges that feed them
The named input slots, in order: none for a source, one for a unary transform, two for a binary one
The type discriminator, e.g. filter
A one-line description for the canvas and the details panel
The description without the values users typed, which may be sensitive, for logs and telemetry. By default the description, which has none.
Whether the node has none of the settings it needs yet, e.g. a Join with no key pair: the canvas shows it as not set, not by its description
The partition columns first: none twice, then each checked
(validatePartitionColumn), every one reported; none at all is one
window over all the rows. Then the sort keys, each checked
(validateSortKey), every one reported, then none twice; none at all is
an unsorted window. Then some window functions, each checked
(validateColumnAggregation, as a window's), every one reported. The
engine fails on a duplicate with no location, so Cube refuses one first.
Optionalerrors: string[]A new partition with the same id, and these window functions
A new partition with the same id, and these partition columns
A new partition with the same id, and these sort keys
The node to keep once its two inputs swap ports: by default this node. A node whose settings name its inputs by side, such as a Join's key columns, returns a new node (same id) with those settings swapped too.
Adds a column per window function to every row (spec §7.13): each computed over the rows that share the row's partition column values (all the rows without any), in the window's sort order. With a sort, an aggregate runs from the partition's first row to the current one, rows tied on the sort sharing a value, as SQL's default frame does (D5, PLAN §11.6); without one, it covers the whole partition. The rank and row functions need a sort (PLAN §11.9). The input's columns and rows stay as they are, in the input's order.