[lqiu96]

Summary & Bottleneck

• In BigQueryImpl.transformTableData and ArrowDeserializer, every FieldValueList in pageNoSchema is already constructed with schema.getFields() attached.
• However, TableResult.getValues() and TableResult.iterateAll() unconditionally wrap the underlying page iterable in addSchema(...), which invokes Iterables.transform(iter, list -> list.withSchema(getSchema().getFields())).
• On every row visited via getValues() or iterateAll(), TableResult calls getSchema().getFields() inside Function.apply, and FieldValueList.withSchema(FieldList) allocates a new FieldValueList instance and copies this.row via ImmutableList.copyOf even when this.schema is already the exact same FieldList.
• In addition, wrapping pageNoSchema.getValues() in Iterables.transform returns a lazy FluentIterable view that does not implement java.util.Collection, stripping known-size Collection and RandomAccess fast-paths for callers collecting or streaming page results.

Changes

• In java-bigquery/google-cloud-bigquery/src/main/java/com/google/cloud/bigquery/FieldValueList.java:
  • Short-circuit withSchema(FieldList) to return this when this.schema == schema || Objects.equals(this.schema, schema).
  • Expose package-private getSchema() so TableResult can inspect whether a page's rows already carry the target FieldList.
• In java-bigquery/google-cloud-bigquery/src/main/java/com/google/cloud/bigquery/TableResult.java:
  • Hoist getSchema().getFields() outside the Iterables.transform Function in addSchema.
  • In getValues(), return pageNoSchema.getValues() directly when it is a Collection whose elements already carry the matching FieldList.
• In java-bigquery/google-cloud-bigquery/src/test/java/com/google/cloud/bigquery/FieldValueListTest.java and java-bigquery/google-cloud-bigquery/src/test/java/com/google/cloud/bigquery/TableResultTest.java:
  • Add unit tests verifying that withSchema returns the same instance when the schema matches and that TableResult.getValues() and TableResult.iterateAll() preserve Collection and row identity when rows already carry the schema.

Expected Impact

• Eliminates $O(\text{rows})$ redundant FieldValueList allocations (24 bytes per row) on every TableResult.getValues() and TableResult.iterateAll() traversal when rows already carry the schema.
• Preserves the Collection and List interfaces on TableResult.getValues() when the underlying page holds a materialized collection.

Verification

• Ran mvn test -pl java-bigquery/google-cloud-bigquery -Dtest=FieldValueListTest,TableResultTest,BigQueryImplTest -Dclirr.skip=true -Denforcer.skip=true -Dflatten.skip=true -Dcheckstyle.skip=true -Dfmt.skip=true (169 tests run, 0 failures).