- Refactoring without tests makes it riskier
- Move some tests from resolver to middleware unit tests to live closer
to where the validation happens, remove duplicate tests
- DRY out code
We had this error in our neo4j pod recently:
```
2019-12-02 08:29:42.680+0000 ERROR Unable to schedule bolt session 'bolt-1018230' for execution since there are no available threads to serve it at the moment. You can retry at a later time or consider increasing max thread pool size for bolt connector(s).
2019-12-02 08:29:42.680+0000 ERROR Unable to schedule bolt session 'bolt-1018224' for execution since there are no available threads to serve it at the moment. You can retry at a later time or consider increasing max thread pool size for bolt connector(s).
2019-12-02 08:29:42.681+0000 ERROR Unable to schedule bolt session 'bolt-1018352' for execution since there are no available threads to serve it at the moment. You can retry at a later time or consider increasing max thread pool size for bolt connector(s).
2019-12-02 08:29:42.682+0000 ERROR Unable to schedule bolt session 'bolt-1018243' for execution since there are no available threads to serve it at the moment. You can retry at a later time or consider increasing max thread pool size for bolt connector(s).
```
Apparently the default is 400 threads. So we must have a leak somewhere.
- update script
- use readTxResult for validateReview
- favor more verbose variables
- do not set review.closed as we close the report and the rule at the
moment is set to 'latestReviewUpdatedAtRules' rule, so it's clear the
last review must have been the one that closed the report
- Don't throw error if a report already exists since we use MERGE and
that does not create a new resource if it exists. That is tested at the
neo4j(database) level.
- We also have a test to make in reports.spec.js that no duplicate is
created to ensure we didn't make some error on our part.
- Fix some faulty logic in validateReview
- We have been using http requests in our set up in this file, which can
be avoided. We are not trying to test that the report/review workflow
works properly here, we are trying to test that if a resource was
reported, and then disabled, some conditions are true. Certainly, we
will have tests that check that the report, review workflow works as
expected.
Since we decided to open Human Connection for public registration, we
don't need to create artificial shortage of signups. The original
intention here was too:
1. Grow the network slowly and under control
2. Create an incentive for users to invite their friends
Point 2 is now obsolete, everyone can signup herself. Point 1 is based
on the (seemingly) wrong assumption that there will be a "run" on the
network if we open it for public registration. Our growth was fairly
stable so far.
Ideally we would not have this inconsistent edge case that we filter
through the user type. Much better would be a graphql query like:
```graphql
{
User(filter: { primary_email: { email: "something@example.org" } }) {
id
name
slug
}
}
```
Apparently, `neo4j-graphql-js` *replaces* the schema partly. When we use
join models like `EMOTED`, `neo4j-graphql-js` puts in a join model with
the two names of the connected types.
This registers the language Russian in the backend. Unfortunately, the
locations seem to be implemented with a hard coded attribute for each
language. 😞
We should refactor this.
@KapilJ your solution is right now the best solution, I think. Probably
the ideal solution would be if we could implement the `CreateComment`
resolver in such a way that it is doing eager loading of the
`comment->post->author` relationship and resolving a commment object
which has the required objects all set. Then you wouldn't have to
refetch all the stuff.
But I think this is OK for now 👍
