Compress

Processor that concatenates columns and stacks the rows of each group into per-group 2D arrays.

Packing many columns into one array gives them all a single dtype, so alongside the column names it records the dtype each one had, letting a reader put a column back the way it was written. override_dtypes records the packed dtype instead, for a pack where the stored type is the one that matters.

Configuration

Selected as name: compress.

Option	Description	Type	Default
to	Destination frame for the compressed result.	str	required
by	Columns to group rows by (one output row per group).	column(s)	the event ids
cols	Columns concatenated and stacked into the per-group array. Use ‘__all__’ to compress all non-by cols.	column(s)	required
out	Output column holding the stacked array.	str | int	required
dtype	Data type of the stacked array.	str	null
record_names	Store the source column names in the attributes.	bool	true
record_offset	Store the per-column offsets in the attributes.	bool	true
override_dtypes	Record every column as the dtype it was packed into, rather than the one it came in with. For a pack whose stored type is the one consumers should see.	bool	false

- name: compress
  kwargs:
    to: pulses
    cols: [ x, y, z, charge, time ]
    out: features