Dataloader 3.0: A new algorithm to solve the N+1 Problem

We've developed a new algorithm to solve the N+1 Problem in GraphQL by loading data breadth-first instead of depth-first. This approach allows us to batch together requests for sibling fields, reducing the number of concurrent operations and making the code more maintainable and efficient. The benefits include eliminating the need for concurrency synchronization, reducing performance overhead, and improving code understandability. By resolving fields breadth-first, we can automatically batch together requests for sibling fields, reducing the complexity of our code and improving performance by up to 5x compared to the traditional DataLoader pattern. This approach has significant implications for GraphQL servers, frameworks, and gateways, encouraging a reevaluation of the trade-offs involved in using the DataLoader pattern.