Event Sourcing with commanded: How to handle side effects?

Hey there!

currently looking into event sourcing.

I stumbled over commanded and it looks pretty neat. Followed the documentation and set the example bank account commands/events.

Now I am at a step where I would like to transfer what I learned but I am certainly missing some understanding.

What I want to store and track are some documents (files on disk) and a graph with their relations. That means, at some point, I actually have to write the files and adjust the graph. But I do not get where or how I would do that in a way that things are not run repeatedly but are still replayable.

As an example, a document should have an owner, which is stored as an edge/connection in the graph. I do not know how/when/where I should create this edge.

Example 1: Create relation in execute

1. Dispatch command AssignOwner
2. execute:
   1. Check if owner can be assigned, the document exists etc.
   2. Update the Graph by adding the relation
   3. Generate Event OwnerAssigned

Now … lets say I’d like to debug something locally that happened on the server, I would download the events and … the relation in the graph would be missing, because the event OwnerAssigned does not create it and the command AssignOwner is no longer present.

Example 2: Create relation in apply

1. Dispatch command AssignOwner
2. execute:
   1. Check if owner can be assigned, the document exists etc.
   2. Generate Event OwnerAssigned
3. In apply create the edge in the graph

I guess now, every time the events are aggregated, the edge would be created?

Example 3: Create additional commands from event

1. Dispatch command AssignOwner
2. execute:
   1. Check if owner can be assigned, the document exists etc.
   2. Generate Event OwnerAssigned
3. React to event OwnerAssigned and trigger additional commands that alter the graph.

How would this differ from Example 2?