Allow full table updates again

[laurenceisla]

Full table updates where disallowed in 86574c8. This PR brings back that functionality because of the reasons mentioned here: #2195 (comment). This issue is also related, mentions discussions on how to handle bulk updates: #1959

• Allow full table update
• Add tests

[steve-chavez]

As it is, the PR causes that the following:

PATCH /table

{"id": 1, "name": "new-name"}

No longer does a full table update(doesn't trigger a pg-safeupdate error if installed) but instead it would update the row with id 1. However if we do:

PATCH /table

{"name": "new-name"}

This would do a full table update because there's no id in the body.

Somehow this feels like surprising behavior to me, the url no longer denotes the operation. It could be very confusing for users.

I still think that bulk patch(also delete) would be more explicit and consistent with underscore operators. However we cannot add a single one like on #2405, we need a consistent design for all <- this could take a while and it'd be better for a next release.

---

What if we reopen #1959, revert back #2311 and leave bulk patch for the next release?

[wolfgangwalther]

What if we reopen #1959, revert back #2311 and leave bulk patch for the next release?

Yes, I came to the same conclusion yesterday, after reading through everything.