josephmckinsey
commented
2 years ago @areiman suggested this.
We should be able to run in mixed run-times. For example, most of the simulation would be managed by Docker, except for one which would be managed locally.
Even locally, this could be useful. We could have all of the components be automatically started, but one component would be manually managed by the user. In this case, they would only have to encode and decode types and interact with HELICS. After they are confident the HELICS interaction, then they can add the standardized input_mapping.json andstatic_inputs.json as configuration and component_definition.json as the "type" of the component.
This would allow users to use their standard tools for development and limits our own duplication of existing tools.
It should be possible to run the simulation where you have access to gdb in one of the processes. This can be accomplished ad-hoc by removing the entry of interest from the helics runner JSON and then running it yourself.