A modern programming language that makes developers happier.

kotlinlang

Hey All, we are facing an issue where on Apollo v3 the query watchers are not triggered when doing a manual `publish` , wanted to know if there is a change how query watchers are triggered as compared to Apollo v2? I did verify via the `dump`  and we do have a record with the cache key I am trying to publish. There is no actual change in the record it's just a trigger to update the UI state.

I’m AFK so would need to double check but it’s possible v3 doesn’t listen to “object” keys, only “field” keys

• object key: `User:42`
• field key: `User:42.name` 

I am guessing object key in your example if we have an object
```{
 __typename: "user",
id : "42",
flag: true //or any other fields
}```
`publish("42")`

This was made so that we have more granular notifications

And maybe try `publish("42.flag")` too (id is quite a special flag)

That worked, not sure I read about this anywhere in the migration guide :thank-you-color:
I am a bit confused though, field publishes are required to filter out unwanted publishes by the object itself?

Also forgot to mention it worked with the field  `publish("42.id")` ? Should I not be using `id`?

&gt; not sure I read about this anywhere in the migration guide 
Good callout. I’ll add something.
&gt; Also forgot to mention it worked with the field  `publish("42.id")` ? Should I not be using `id` 
Using `id` is fine the important thing is to use a field that is aueried by your query
&gt; I am a bit confused though, field publishes are required to filter out unwanted publishes by the object itself?
Imagine you have a top level `post` and a 2 queries, 1 that gets the likes and the other get the contents.
If we notify based on the post id, both queries are potentially refetched when you only want to refetch likes (and contents is much larger)

<https://github.com/apollographql/apollo-kotlin/pull/6288/files>

Understood, thanks for the help, much appreciated.