Update badgerDb version

[AndrewSisley]

Check for existence of IM mode first https://dgraph.io/docs/badger/get-started/#in-memory-mode-diskless-mode

IPFS has a wrapper for v2, but might be worth wrapping the newer v3 ourselves - have a think, and if leaning towards v2 make sure it has the IM option.

Badger link

[jsimnz]

Check for existence of IM mode first https://dgraph.io/docs/badger/get-started/#in-memory-mode-diskless-mode

IPFS has a wrapper for v2, but might be worth wrapping the newer v3 ourselves - have a think, and if leaning towards v2 make sure it has the IM option.

Badger link

🦡 😂 got me with that one

Seems like it is prob best to create a v3 wrapper. Or at least open up an issue and see what the IPFS team thinks about the current strategy is w.r.t Badger versions and their wrappers. I can't think they want to stick with this approach they have for every released Badger version.

[jsimnz]

I havent fully grokked the in-memory implementation yet, however Badger V2 does support the in-memory mode, so we can at least start with the v2 IPFS version for now.

[AndrewSisley]

:D wasn't sure anyone would find it lol (or if the email would plain-text it)

Ah looks like they have had a plan to change the approach since Jan, I guess they just haven't gotten round to it: ipfs/go-ds-badger#109

You'd rather go v2 then v3 and hit our users twice (if we don't get to v3 before the next release)?

[jsimnz]

I just meant try out the v2 mode, to test the in-mem version of Badger, and if all looks good, we can either ping the IPFS team on that issue, and see if they have some idea of timeline, or we:

1. Create our own v3 wrapper
2. Fork the existing wrapper, and implement the structure/plan they are already talking about, and see if we can merge it upstream.

[AndrewSisley]

I pushed very WIPy v2 here: https://github.com/sourcenetwork/defradb/pull/new/sisley/chore/I52-badger-2
v3 is here: https://github.com/sourcenetwork/defradb/pull/new/sisley/chore/I52-badger-3

There are a couple of unit test failures, which might actually be an uncaught production bug to do with transaction disposal

Also, they haven't released v2 with context yet - I had to go-get the master branch of the IPFS branch, so I guess that rules out using v2 in the short term

[jsimnz]

fixed in #56