fix: coerce UUID to String in readable_metrics to avoid ClassCastException in Spark

[kadai0308]

Spark expects all StringType fields to be castable to CharSequence, but Iceberg's
readable_metrics lower_bound/upper_bound may decode to java.util.UUID for UUID-typed
columns. This causes a runtime ClassCastException when Spark tries to read those
metrics as UTF8String.

This commit fixes the issue by converting UUID values to string when generating
readable metric values for Spark metadata tables.

Closes: #13077 (comment)

[singhpk234]

thank you for the fix @kadai0308 ! can you please add an UT for this as well

[kadai0308]

thank you for the fix @kadai0308 ! can you please add an UT for this as well

@Fokko
updated! Please help me review the PR. Thx!

[github-actions]

This pull request has been marked as stale due to 30 days of inactivity. It will be closed in 1 week if no further activity occurs. If you think that’s incorrect or this pull request requires a review, please simply write any comment. If closed, you can revive the PR at any time and @mention a reviewer or discuss it on the [email protected] list. Thank you for your contributions.

[kadai0308]

@Fokko
Hi, I am not sure about what's the next step of this PR?

[aperture147]

I'm having the same problem when I run catalog_name.system.rewrite_manifests. When will this be merged in the next release of iceberg/spark? Thanks.

[kadai0308]

@Fokko @singhpk234 Just a friendly reminder about this PR.

[Fokko]

This makes sense to me @kadai0308 👍

[ebyhr]

This PR broke one of the tests (TestIcebergParquetSystemTables.testFilesTableReadableMetrics) in Trino Iceberg connector.

I understand this change is mainly intended for Spark, but it will impact Java API users as well. I had to introduce a redundant cast into our connector. Ideally, Spark specific change should be done in Spark module, not API.

[Fokko]

@ebyhr Good catch, and I agree that we should fix this on the Spark side 👍