Hi all,

I would like to discuss an opt-in way for a Polaris catalog to treat
namespace and table names as case-insensitive.

Polaris stores those names exactly as the client sent them. Hive stores
them in lowercase, and Spark and Trino do not send the same spelling for
the same SQL text. A table created from one engine can be missing in the
other, or the catalog can end up with both Employee and EMPLOYEE.

The proposal keeps today's behavior unless a catalog opts in when it is
created. The setting cannot be changed later. Existing catalogs stay
case-sensitive. The detailed design, including the request flow,
configuration, and what does not change, is here:
Case-insensitive namespace and table names
<https://docs.google.com/document/d/1gB4DXUbLhKCfqnzEbRMFBjnLGbpvGVGc-y-dSRNpvus/edit?tab=t.0>

This follows up on https://github.com/apache/polaris/issues/5842. An
earlier thread is here:
https://lists.apache.org/thread/qsyj5ly2z1ydo3s14nnp4oyypco259g8

I would appreciate feedback before starting the implementation.

Thank you,

Reply via email to