Start using STI for decision model

[hellcp-work]

Expands the decision model into an STI model with two new classes.

Keep in mind this uses data migration, so if you try using this in your development setup, start off by running a rails data:migrate

Remaining tasks:

• Fix the canned responses not working with this
• Fix the events that fail in the tests

[saraycp]

@hellcp-work The changes related to canned responses LGTM. Just fixing linters complaints is missing.

[eduardoj]

I would rename the last commit which includes all the changes to the specs from "Fix feature specs for canned responses with the new decision sti" to something more generic, like, for example, "Fix specs with the new decision sti".

[saraycp]

With the proposed changes, we can get something like this:

=> #<DecisionFavored:0x00007f9001b9a648
[...]
 kind: "cleared",
 type: "DecisionFavored">

a newly created Decision where type is correct and kind is not. I know the reason is that we are no longer using kind and we are about to remove it, so we simply ignore it and the database takes the default value (cleared).

However, there is a gap between the code changes being applied and the data migration run, where someone can create a new decision and kind will be wrong. And then, when running the data migration we are going to take the wrong value and set it into the type.

I know we have little risk for this to happen, but who knows? We have to take OBS and IBS in mind. And the fix shouldn't be too difficult: write to both kind and type temporarily. It's also what strong_migrations suggests.

We've just done something similar in PR #16017, I could take care.

What do you think?

[hellcp-work]

I know we have little risk for this to happen, but who knows? We have to take OBS and IBS in mind. And the fix shouldn't be too difficult: write to both kind and type temporarily. It's also what strong_migrations suggests.

We've just done something similar in PR #16017, I could take care.

What do you think?

It's not difficult to implement, we might as well do it

The renaming of the spec commit message is missing

[danidoni]

I would rename the last commit which includes all the changes to the specs from "Fix feature specs for canned responses with the new decision sti" to something more generic, like, for example, "Fix specs with the new decision sti".

Done

[eduardoj]

I would rename the last commit which includes all the changes to the specs from "Fix feature specs for canned responses with the new decision sti" to something more generic, like, for example, "Fix specs with the new decision sti".

Done

I'm not fully convinced with "Make feature specs work with Decision STI". I would get rid of "feature". We are touching more specs than feature specs...

[danidoni]

I'm not fully convinced with "Make feature specs work with Decision STI". I would get rid of "feature". We are touching more specs than feature specs...

Alright