The EDA tool uses its own instance of the Safe-DS services. This requires careful synchronization with the services in the language server and leads to duplicate work for parsing and validation. It also makes it harder to implement extensions for other IDEs. For now, the EDA tool should rather send requests to the language server to execute code and get results.
If that extra hop introduces too much latency, we can still request the language server to generate code (usually a smaller response) and then send it directly to the runner via an extra websocket connection. This does not need another full set of services.
The EDA tool uses its own instance of the Safe-DS services. This requires careful synchronization with the services in the language server and leads to duplicate work for parsing and validation. It also makes it harder to implement extensions for other IDEs. For now, the EDA tool should rather send requests to the language server to execute code and get results.
If that extra hop introduces too much latency, we can still request the language server to generate code (usually a smaller response) and then send it directly to the runner via an extra websocket connection. This does not need another full set of services.