Proposal: Faster Algorithm For OverlappingFieldsCanBeMerged

[acceleratesage]

At XING we are using the Sangria implementation of GraphQL. The implementation of the OverlappingFieldsCanBeMerged validation in Sangria follows the one in this reference implementation. We experienced performance problems with it related to the number of fragments. After we invested some time to solve it, I came up with a better algorithm, that we are using now.

• Here is the description of the algorithm.
• And here is our PR for Sangria.

The proposal would be to use this algorithm in the reference implementation. Below is an example benchmark graph showing the performance improvement for multiple fragments in the same selection set.

[IvanGoncharov]

@SimonAdameit Thanks a lot for sharing 👍 Its super useful ATM we preparing for TS conversion so we in a feature freeze state but it would be at the top queue after conversion is finished.
Also, I think we need to use this opportunity and address edge cases raised in #1497 at the same time.

[yaacovCR]

but it would be at the top queue after conversion is finished.
Also, I think we need to use this opportunity and address edge cases raised in #1497 at the same time.

with TS conversion finished, this should be in the news!