This made me think about indexing exported animation frames—the sprite sheets produced by https://spritefuse.com/ are one example. A useful test case might be duplicate record keys arriving after the bound is hit: round-robin inserts could spread repeats across filters, so I would be interested in whether Hudi handles duplicate keys before insertion or simply accepts the extra bit saturation. The exact-lookup step is what makes that performance question separate from data correctness.
This made me think about indexing exported animation frames—the sprite sheets produced by https://spritefuse.com/ are one example. A useful test case might be duplicate record keys arriving after the bound is hit: round-robin inserts could spread repeats across filters, so I would be interested in whether Hudi handles duplicate keys before insertion or simply accepts the extra bit saturation. The exact-lookup step is what makes that performance question separate from data correctness.
That's a really good idea. I have to double check, but AFAIK such monitoring is not implemented yet. A good opportunity to contribute to the project!