[question] restructuring dependencies?

[seriousme]

Currently whenever we update aedes-persisistence to facilitate aedes-persistence-redis or aedes-persistence-mongodb we need to update the version of aedes-persistence in aedes-cached-persistence as well.
This can be quite annoying I imagine.

So a few suggestions:

1. merge aedes-cached-persistence into aedes-persistence.
2. remove the passthrough from aedes-cached-persistence and let redis and mongodb import aedes-persistence directly

Happy to help if required.

Kind regards,
Hans

[robertsLando]

I would be ok in merging cached persistence in persistence as long as that doesn't downgrade performances. BTW the ultimate fix for this would be the one I suggested to move to a monorepo

[seriousme]

Thx, and yes I think too that a monorepo would help. But for now we can simplify stuff.

[seriousme]

@robertsLando
I suggest to do the following:

1. move abstract.js and test.js to a /test folder so its clear that they are not used during normal operation
2. update the test scripts in package.json to use the new /test folder
3. mv promisifed.js to a /lib folder so its clear that it is not called directly by Aedes
4. copy callbackifiedPersistence.js from cached-persistence into /lib
5. rename index.js from cached-persistence to clusterBroadcast.js in /lib and also rename the class because it doens't cache anything, it just broadcasts added/removed subscriptions (if we are doing a major semver change then we'd better do it right).
6. update typescript defs where required
7. create an upgrade.md that explains the changes
   7a. do we need this or are we pretty sure Redis and MongoDB are the only direct dependents on aedes-cached-persistence that are actually used and maintained ?
   7b. if mongo and redis are the only ones then we might be able to simplify clusterBroadcast.js even further
8. update the readme.md of aedes-persistence
9. update the readme.md of aedes-cached-persistence pointing to aedes-persistence
10. replace the dependency on aedes-cached-persistence by aedes-persistence in Redis and MongoDB
11. done

We could even go further (got some ideas on that as well), but for now this is already quite a change.
None of the changes touch lines of code except for the rename of the class in 4 and the simplification in 6b. It should all be possible with zero performance impact.
What do you think?

Kind regards,
Hans

[robertsLando]

About point 7: AFAIK the only persistences that are used are redis mongo and level. Anyway we can mark this with a major release. Considering we are introducing a breaking change we could also consider to drop callback based calls and use promises instead? This of course will require a major release also of aedes.

@mcollina thoughts?