Hi Iceberger, I have a proposal to simplify the tableExists API in the Hive catalog, which involves a behavior change, and I’d like to hear your thoughts.
Currently, in our catalog interface[1], the tableExists method is implemented as a default API by invoking the loadTable method. It returns true if the table can be loaded without exceptions. This behavior implies two checks: The table entry exists in the catalog. The latest metadata.json for the table is not corrupted. The behavior change I’m proposing focuses only on the first condition—checking if the table entry exists in the catalog. This separates the concerns of table existence and table health (e.g., metadata not corrupted). Such a change could improve the performance of existence checks, especially for RESTcatalog where table existence is abstracted as an HTTP HEAD request [2]. I also reviewed the current usage of the tableExists API in the Iceberg codebase to ensure that this optimization would not have any negative impact. I’d love to hear everyone’s feedback on this! If there’s consensus, I can follow up with a similar optimization for the viewExists method in the Hive catalog. [1]: https://github.com/apache/iceberg/pull/11597 [2]: https://github.com/apache/iceberg/blob/3badfe0c1fcf0c0adfc7aa4a10f0b50365c48cf9/open-api/rest-catalog-open-api.yaml#L1129-L1133 Best regards, Steve Zhang
