You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@iceberg.apache.org by "vvarma (via GitHub)" <gi...@apache.org> on 2026/05/11 17:26:26 UTC

[I] Predicates on nested-leaf columns fail [iceberg-rust]

vvarma opened a new issue, #2432:
URL: https://github.com/apache/iceberg-rust/issues/2432

   ### Apache Iceberg Rust version
   
   0.6.0 (latest version)
   
   ### Describe the bug
   
   Nested-leaf predicates in the Arrow reader are rejected even when the predicate targets a primitive leaf inside a struct.
     For example, a predicate like:
   
     ```sql
     nested.value = 20
   ```
     fails during row-filter construction because PredicateConverter::bound_reference resolves the predicate field ID to a Parquet leaf column, then rejects it when the Parquet column root is a group/struct.
   
   ### To Reproduce
   
   _No response_
   
   ### Expected behavior
   
   _No response_
   
   ### Willingness to contribute
   
   I can contribute a fix for this bug independently


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org
For additional commands, e-mail: issues-help@iceberg.apache.org


Re: [I] Predicates on nested-leaf columns fail [iceberg-rust]

Posted by "idan57570-art (via GitHub)" <gi...@apache.org>.
idan57570-art commented on issue #2432:
URL: https://github.com/apache/iceberg-rust/issues/2432#issuecomment-4424478878

   "I’ve analyzed the issue regarding nested-leaf predicates in the Arrow reader. While there is a PR linked (#2433), it's crucial that we ensure the PredicateConverter doesn't just resolve the field ID but also correctly handles the ProjectionMask for nested Parquet structures.
   
   My proposed addition to the fix: We should implement a recursive check in bound_reference that validates the leaf type against the Arrow schema, even when the Parquet root is a group. This will prevent the row-filter from failing when it encounters struct-wrapped primitives.
   
   I can assist in verifying the fix with additional unit tests for multi-level nested structs if needed.
   
   Wallet: RTC972526299366 @Scottcjn"


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org
For additional commands, e-mail: issues-help@iceberg.apache.org


Re: [I] Predicates on nested-leaf columns fail [iceberg-rust]

Posted by "LuciferYang (via GitHub)" <gi...@apache.org>.
LuciferYang commented on issue #2432:
URL: https://github.com/apache/iceberg-rust/issues/2432#issuecomment-5087763625

   I hit this and opened #2906 for it.
   
   The fix resolves a predicate's field to its Parquet column path (root to leaf) rather than a single leaf index, and walks that path through the projected `RecordBatch`, descending struct children by name to reach the primitive leaf, so `person.age > 25` builds a row filter instead of erroring on the group root.
   
   Scope note: this covers top-level and struct-nested leaves. Fields inside a list or map get no accessor in `Schema::build_accessors`, so a predicate on one already fails at bind time and never reaches the row filter, no change needed there. Row-group and page-index pruning resolve nested leaves by field id and didn't error on this path, so I left them out; happy to look at them separately if it's worth tightening.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org
For additional commands, e-mail: issues-help@iceberg.apache.org