Dynamically splitting a topic into multiple derived streams (by field)

[mrawding]

Say I have an input topic like:

/robots/status

/3d-tracks

Is it possible to use the topic converter example (or another supported pattern) to dynamically split a single topic into multiple derived topics at runtime?

Example 1: Split by name

Given a topic:

/robots/status

Where each message contains:

msg.name (string)

msg.location (sensor_msgs/NavSatFix)

Is there a supported way to dynamically create new topics like:

/robots/<msg.name>/gps

Each publishing a derived LocationFix (or NavSatFix) message for that specific robot?

In other words, the set of output topics is not known ahead of time and depends on values inside the message.

Example 2: Split by ID from an array

Given a topic:

/3d-tracks

Where each message contains:

msg.tracks[]

--- for track in tracks ----

track.id

track.pos (3D vector)

Is it possible to emit one output topic per track ID, e.g.:

/tracks/<id>

Each becoming its own data stream over time?

Key question:
Does Foxglove currently support creating dynamic output topics from a converter (or extension) based on message contents, or are output topics required to be declared statically at initialization time?

If this is not supported today:

Is there a recommended workaround or pattern?

Or is this something only achievable by preprocessing the data outside Foxglove (ROS node, MCAP transform, etc.)?

Thanks!

[jacob-foxglove]

Hey @mrawding, thank you for reaching out. Please see the following notes:

Does Foxglove currently support creating dynamic output topics from a converter (or extension) based on message contents, or are output topics required to be declared statically at initialization time?

Answer:
No, topic converters require a definition of what the contents / schema when registered.

In this case your best bet would be to do something like this as you said via a pre-processing step with some sort of ROS node, MCAP transform for example.