I watched <this> amazing talk on ktor graphql from...

# graphql-kotlin

[asavio]

I watched

this▾

amazing talk on ktor graphql from KotlinConf’23, Amsterdam yesterday. Amazing talk! I’m now curious about how ktor and kotlin graphql solves the n+1 problem that we generally encounter in graphql servers. I couldn’t find any resources on how ktor solves that. Does any one know how that’s done in ktor? Thanks.

[Dariusz Kuc]

👋 folks generally use data loader pattern to batch requests downstream (see https://opensource.expediagroup.com/graphql-kotlin/docs/server/data-loader/#dataloaders-and-coroutines)

[asavio]

Oh, this is great. Thanks, will check it out!

[Gerard Klijs]

This▾

presentation about data loader 3.0 might also be interesting.

[Dariusz Kuc]

I watched it expecting some fundamental rewrite.... but it was nothing new

graphql-java resolves queries in a breadth first matter (afaik all graphql libs do) - it will gather all available futures (for given level) and then await for the first leaf value. It actually even does an extra optimization by checking for sync leafs - so it will actually traverse far down the tree until all fields are futures before waiting for them.

This is actually what Sam implemented with sync exhaustion strategy for the batch dataloader.

[Dariusz Kuc]

As for federated resolver -> thats what the query plans are for (so using Apollo router you end up with 4 fetches and definitely dont use thread per id)... so again nothing new on that front 🤷

[Dariusz Kuc]

I did like the will it compose game though :)