Hi! I'm trying to migrate over to the new normaliz...

# apollo-kotlin

[Seb Jachec]

Hi! I'm trying to migrate over to the new normalized cache, having followed the docs for Relay-style pagination, but I'm getting a NullPointerException crash in ConnectionMetadataGenerator.metadataForObject that I'm having a hard time understanding 🧵

[Seb Jachec]

I've created a CacheManager that I'm providing when building my ApolloClient, using ConnectionMetadataGenerator(Pagination.connectionTypes) and a ConnectionRecordMerger, alongside both Apollo's in-memory and SQLite caches chained together. I've got a number of schema extensions (happy to share in full), but I thought the key part would be that I've specified connectionFields on our Query type via a @typePolicy directive.

Wrapping ConnectionMetadataGenerator and forwarding metadataForObject so that I can examine the obj and context being passed in, I can see a large number of null objects.

I'm not exactly sure how to debug this exception and crash though – I'd love any pointers as I'm probably making a mistake somewhere!

[bod]

is it a project you could share so I could have a look?

[Seb Jachec]

It isn't something I'm able to share unfortunately (plus our project and schema is pretty sizeable)

[bod]

Maybe you could extract only the relevant pieces in a minimal project that shows the issue?

From what you said it looks like you did all that's needed but there's probably a small detail missing 😅. There's a sample using pagination here, maybe you could try to compare with what you have.

[bod]

are you in the "simple" case, aka vanilla relay style?

[Seb Jachec]

Yep, simple vanilla Relay style

[Seb Jachec]

I'll double check against that sample

[Seb Jachec]

Comparing against that sample, nothing jumped out as being obviously wrong in our project. But, I've narrowed down the issue to a specific set of fields being requested (via fragments) as part of a single query, without them there's no issue.

[bod]

Hmm interesting. That doesn't ring a bell... Might be a bug!

[Seb Jachec]

Would there be any benefit to sharing a trimmed down version of our schema (relevant parts), plus query & fragments, either here or in a GH issue? Or would that not be particularly helpful in diagnosing this?

[bod]

Yes that would be perfect!

[Seb Jachec]

Hi, it was a bit tricky to just get the relevant parts of our schema, queries and extensions but I've zipped them up here. If you're able to give any insights (or anything that seems potentially odd for normalized cache writes) just from that, I'd be really grateful.

I've left a comment or two, but essentially if defaultTags aren't fetched in the calendar fragment (calendar.graphql), or calendar isn't fetched as part of the event fragment (event.graphql) 'upstream' from that when querying events, we don't see a NullPointerException

Trimmed GraphQL Schema.zip

[bod]

Thanks! I'm off today but will have a look tomorrow!

[Seb Jachec]

Thank you!

[bod]

All right I've made a little project using your schema/operations - at the moment I don't repro the NPE. This is probably because of the response I crafted here, which may not trigger the issue? Could you tweak the project, maybe replace with a response that triggers the issue?

[Seb Jachec]

Ah thanks, tweaking now! This looks like a really good isolated way to test – I hadn't thought about testing with a mock Apollo server admittedly

[Seb Jachec]

Hi, sorry for the silence – I hoped to have a more satisfying update or resolution, but right now I haven't even been able to reproduce the original issue in our own project, perplexingly

[bod]

Odd! 👀

[bod]

were you always using the same version?

[Seb Jachec]

Yep, always using the 1.0.0-alpha.1 normalized cache artifacts with Apollo 4.2.0, that hasn't changed since I started migrating over

[bod]

all right. Well, odd 🙂 Hope you'll catch it so we can continue trying to understand the issue.

[Seb Jachec]

I'll let you know if I get to dig into it again!