1D (and 2D) laser-scans are a common data type in robotic applications.

Unfortunately we currently have to convert this data into points (or lines) on the client-side before logging.

The data is generally made up of an array of range-values (1D or 2D tensor data) and a mapping of index values to angular offsets.

For example, see: https://github.com/ros2/common_interfaces/blob/rolling/sensor_msgs/msg/LaserScan.msg

This is conceptually very similar to the depth projection mapping we are now doing for depth-images, but with a radial (or other paremetric) projection rather than a pinhole model.

An additional interesting complication with LaserScan data is that the data in a single sample typically represents data collected over a range of time (as a function of the rotational speed of the scanner). Once we support a concept of events with Duration, LaserScan would be a good candidate for representing this.