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.
I deprecated the debug flags myself here:
https://github.com/neo4j-graphql/neo4j-graphql-js/pull/288
You can now debug the queries run by `neo4j-graphql-js` by starting
the backend like this:
```bash
DEBUG=neo4j-graphql-js yarn run backend
```
- it's good to return the pinnedAt date for ordering
- move test to a better describe block
- remove unneeded outdated variables from graphql/PostQuery UpdatePost
- fix indentation in Post.gql
- fix pinnedAt to return pinned.createdAt, not post.createdAt
Co-authored-by: Mike Aono <aonomike@gmail.com>
- this was causing the noftications to break since they were using the
post fragment and pinnedAt was returning null
- remove unused bits of code
- revert changes to cypress since they were not causing the error
- can be changed in another PR
- following @roschaefer's PR review suggestions
- simplify UpdatePost by using pinPost/unpinPost
- did not remove filtering because attempting to have two queries caused
problems as well to do with duplicate records, etc... and it's working
now
- when a user updates a post, we should not override every property in
our database with the new params, since we have read-only properties
like createdAt that we don't want to go deleting aimlessly.
It's sometimes unbelievable how many bugs you find when doing
refactoring. This time there was some apparent confusion about
`commentsCount` and `commentedCount`. The counters on the post card were
never showing the correct number.
I think the validation for categories was too complex and the graphql
query 'PostEmotionsByCurrentUser' was erroneously allowed for
unauthenticated users.
- if there are no ids, we shouldn't add an empty array since it adds unneccessarily to our auto-generated post query and affects greatly performance. see issue #1390
@mattwr18 I prefer (I believe it's even best practice) that a delete
mutation should return the deleted object. If you run the delete
mutation again, it should return `null` because there is no object like
that anymore. That way the client knows if a delete mutation has changed
any state in the database.
Also I fixed another bug in the resolver. If your graphql mutation looks
like this:
```gql
mutation {
RemovePostEmotions(to:{ id:"p15"}, data:{emotion: angry}) {
from {
id
name
}
to {
id
title
}
emotion
}
}
```
Then you get errors because your resolver does not return the name for
the user or the title for the post anymore. Just use spread operator...
and it's fixed.
