Add kGraphGRC message for setting/getting a YAML graph #553
Conversation
iamsergio
changed the title
|
It's done, but will undraft only after #555 is merged, so I can rebase (change required for tests to work). |
This allows Graph to use the yaml importer without having a cyclic dependency between headers, and without moving stuff into .cpp files. Part of task #246 Signed-off-by: Sergio Martins <[email protected]>
Part of task #246 Signed-off-by: Sergio Martins <[email protected]>
|
There was a problem hiding this comment.
Looks good in general, just some questions on how this will interact with the lifecycle state of the graph and its blocks, see below. Might all be fine as it is and just need some more explanation for me to update my understanding of how we are going to use this.
| @@ -343,6 +350,7 @@ class Graph : public gr::Block<Graph> { | |||
| propertyCallbacks[graph::property::kEmplaceEdge] = std::mem_fn(&Graph::propertyCallbackEmplaceEdge); | |||
| propertyCallbacks[graph::property::kRemoveEdge] = std::mem_fn(&Graph::propertyCallbackRemoveEdge); | |||
| propertyCallbacks[graph::property::kGraphInspect] = std::mem_fn(&Graph::propertyCallbackGraphInspect); | |||
| propertyCallbacks[graph::property::kGraphGRC] = std::mem_fn(&Graph::propertyCallbackGraphGRC); | |||
There was a problem hiding this comment.
I feel like the semantics of the graph receiving a message and then replacing itself are a bit weird. For my mental model, the scheduler getting that message as the owner of the graph would make a lot more sense?
This might also be that my model of of the interplay between the components is a bit outdated and the graph has become a lot more dynamic lately.
| } else if (message.cmd == message::Command::Set) { | ||
| const auto& data = message.data.value(); | ||
| auto yamlContent = std::get<std::string>(data.at("value"s)); | ||
| *this = gr::loadGrc(*_pluginLoader, yamlContent); |
There was a problem hiding this comment.
If I read this correctly, this will just replace the current graph with a new one which has been read from the yaml, right? It might do some cleanup in the old graph's destructor, but it cannot do blocking actions like correctly stopping the current graph, before removing it.
Question: Should it be/Is it possible to do this on a running graph, or does it have to be in stopped state? On the other hand this would make sending the message more complicated, since effectively you would need to trigger and wait for the state change before.
There was a problem hiding this comment.
While integrating in OD, I'm seeing a crash. I'll first add a unit-test for "Running Graph" case, so we capture the crash.
@ivan-cukic can comment on the rest
|
being folded into #570 |
Add kGraphGRC message for setting/getting a YAML graph
For #246