Improvement

[superstar54]

Issue 1

instead of running

wg = my_workflow.build(1, 2, 3)

The user will run the following code directly

wg = my_workflow(1, 2, 3)

Better rasie an error if user run outside a graph.

Can the user view the graph without inputs?

issue2

wg.run() the second time, gives wrong error message.

issue3

add_multiply.run(1, 2, 3)
wg = add_multiply.build(1, 2, 3)
wg.run(inputs={'x": 1, "y": 2, "z": 3})

The syntax is not consistent.

improve doc

provide a cheat sheet to list what the user gets if the run

• add(1, 2)
• add_multiply.build(1, 2)

Giovanni want to understand what are "Task", "Handler", "TaskSocketNamespace" etc.

Issue4

Could we also support .build, .run, .submit for aiida-core's process class (WorkChain, CalcJob)

For Workchain, we may have an issue, because WorkChain has its own submit, but we may rename it to submit_child

For the tutorial, for the moment, we update the aiida-core to support `engine.submit(WorkGraph, intpus=)

Open an issue for the discussion for the aiida 3.0, to update aiida-core process regarding adding .run and .submit.

explain the difference between task and task.graph

• a task can not call another task inside
• a task.graph can not return raw value

improve while docs

not show wrong example at the beginning.

change wg.process to wg.node

[edan-bainglass]

After some discussion in the AiiDA Core team, we are considering that perhaps supporting .run and .submit is not a great idea. Perhaps it is best to only keep AiiDA's native engine.submit(<some-workgraph>, **inputs). Support for this is being introduced in aiidateam/aiida-core#7261. I would keep .build of course, or maybe rename it as .compile (TBD).

My current thinking is that THE USER submits A WORKGRAPH, i.e., submit(workgraph), rather than THE WORKGRAPH submitting itself, i.e., workgraph.submit().

@superstar54 good to get your thoughts on this.