[Fix] Columns of type "Object" merging with old value on Postgres

[jaeggerr]

After this discussion #4808

The PR fixes an issue with Postgres due to this commit #2984.

When updating a field of type object (JSON object), the previous and new values are merged.
This is not the expected behaviour as it is not documented and not happening in MongoDB.

[codecov]

Codecov Report

Merging #4815 into master will decrease coverage by 0.13%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #4815      +/-   ##
==========================================
- Coverage   92.82%   92.68%   -0.14%     
==========================================
  Files         119      119              
  Lines        8684     8684              
==========================================
- Hits         8061     8049      -12     
- Misses        623      635      +12

Impacted Files	Coverage Δ
...dapters/Storage/Postgres/PostgresStorageAdapter.js	97.15% <100%> (-0.36%)	⬇️
src/Adapters/Auth/meetup.js	84.21% <0%> (-5.27%)	⬇️
src/Adapters/Auth/linkedin.js	81.81% <0%> (-4.55%)	⬇️
src/Adapters/Auth/facebook.js	80% <0%> (-4%)	⬇️
src/Routers/PushRouter.js	92.85% <0%> (-3.58%)	⬇️
src/Adapters/Auth/facebookaccountkit.js	72.72% <0%> (-3.04%)	⬇️
src/Controllers/SchemaController.js	96.48% <0%> (-0.24%)	⬇️
src/Controllers/DatabaseController.js	94.64% <0%> (-0.2%)	⬇️
src/RestWrite.js	93.61% <0%> (-0.19%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1afc34e...e67edd1. Read the comment docs.

[flovilmart]

LGTM! Tests pass which is amazing!

[flovilmart]

Using async/await, there’s no need for done as a param and call at the end of the method

[dplewis]

@flovilmart I think more test are needed for this. See as getting rid of COALESE doesn’t change anything here.

The use of ‘unset’ is a workaround.

Also {} merged with { c: “d” } is the same as {} updated to { c: “d” }

[flovilmart]

@dplewis, are you able to write a failing test? The unset, because not saved after I believe has no effect?

[dplewis]

@flovilmart Here is the failing test per #4808 (comment)

The unset here looks weird to me. Also COALESE is just to prevent null values don't see how that fixes this.

[flovilmart]

Uhm ok. Not a PG expert. I assumed the test was failing before te fix :)

[jaeggerr]

I am not a PG expert neither but I did not just remove the COALESE but I replaced
( COALESCE($${index}:name, '{}'::jsonb) by {}'::jsonb

As far as I understand, it will take the previously stored value or {} if it is null.
Then, there is a || that merges the left part with the right part.
The purpose of the PR is to remove $${index}:name to prevent the old value to be merged with the new one.

[dplewis]

@jaeggerr Just ran a few more tests looks right. So {} gets set then you add and remove from it right?