Ecto Multi Support Proposition

[DiscoStarslayer]

Initial rough draft of a PaperTrail.Multi API that wraps Ecto.Multi to allow creating multi step transactions using a pipeline.

Inspiration taken from this ticket #48.

This PR extracts the logic from the PaperTrail module and breaks it down into transaction creation + committing steps in PaperTrail.Multi. These functions provide the Multi API and are then reused within PaperTrail to implement the non-Multi functions.

Since even the non-Multi functions require a transaction to apply the paper trail logic, I thought this might be a nice approach.

API wise, it's almost 1-to-1 with Ecto.Multi, the only exceptions are:

• no delete_all
• no insert_all
• no insert_or_update
• no update_all
• Multi transaction should be committed to the DB with PaperTrail.Multi.commit instead of simply calling Repo.transaction on the multi chain.

There is no tests or documentation, but the existing tests pass fully which I think makes a good smoke test as insert, update, and delete are re-implemented using the Multi-api.

I figured I would open this now and see if I'm barking up the right tree here, and if so I will continue to add tests and documentation around the feature.

[gshaw]

Not having Multi support is a deal breaking for using paper_trail so I'm happy to see this work started. 🎉

Will this API allow inserting a versioned and non-versioned record in the same transaction? How would it look?

PaperTrail.Multi.new()
|> PaperTrail.Multi.insert(versioned_changeset, options)
|> Multi.insert(non_versioned_changeset_in_different_table) # would this work?
|> PaperTrail.Multi.commit()

[gshaw]

@DiscoStarslayer I tried your branch and the above sample code works exactly as I had hoped. Great work, I hope this PR will move forward and get released. I'll keep using this branch to help with the testing and will comment back if I find any problems.

Thanks for getting this going. Much appreciated.

[DiscoStarslayer]

@gshaw That's good to hear! I was going to comment saying theoretically there is no reason why that use case should not work as this is simply reusing the existing Ecto.Multi struct. Great to hear you tested it out and confirmed it though!

[izelnakri]

Hi @DiscoStarslayer, I just noticed your PR. Im reviewing it now, thank you for all the work and sorry for responding late!

[izelnakri]

This is awesome @DiscoStarslayer! Proposed changes enhance the usability of the library without adding much complexity. Frankly we should have done this earlier :) I'm merging this PR and cut a release!