Snapshotting and mutator initial state about eventsourcing

[yarbelk]

Yup, that is the method I was talking of - it switches the argument order from the @singledispatch - which needs F(event:E, state:S) -> S to F'(state:S, event:E) -> S, which is needed by reduce. This took me a bit of time to catch, as its a rather subtle difference when you're just reading through the code.

I added some comments to PR #60 , I like the refactors: it feels cleaner.

Looking forward to the next release.

[yarbelkgazia]

Actually, you need to use the classmethod on the repository, which has the call order reversed so it works with reduce, and sets up the class if the initial state is None. Wasn't clear to me initially. should document this.

[johnbywater]

On the develop branch, the ExampleApplication enables snapshotting by constructing a snapshot strategy, then passing it to the event sourced repository. That causes the repository to check for a snapshot before it gets all the events. And if there is a snapshot, it only gets the events that were applied after the snapshot was taken. It also uses older snapshots when getting an entity at a historical version. There's a test covering it.

        self.example_repo = ExampleRepository(
            event_store=self.version_entity_event_store,
            snapshot_strategy=self.snapshot_strategy,
        )

However, I had two issues with this code: firstly snapshots used timestamps not version numbers, and also there wasn't a snapshotting policy to automatically create snapshots.

So I just reworked the snapshotting. The snapshot records are now in a dedicated table. Snapshots use integers which correspond to event entity version attribute, rather than event timestamps. I added a section to the README file about using snapshots (will move it to the new docs when the code snippets in the new docs are covered with a test, with something like the test_readme.py). And in that new section added an example snapshotting policy that automatically takes a snapshot every five events.

The changes are in PR #60. Let me know what you think?

(Not sure which class method you meant, was it mutate()? I think it's correct that the snapshot events aren't supported by the mutator function, if that's what was happening in your first comment.)

[johnbywater]

Just to link things up, there is now a page in the documentation about snapshotting, here:

http://eventsourcing.readthedocs.io/en/latest/topics/user_guide/snapshotting.html