Parquet: Diagnose Row-Group Statistics and Pruning

Parquet: Diagnose Row-Group Statistics and Pruning

Choose one answer. All questions are included, with explanations after checking. No time limit. Training target: 26/32. For data engineers diagnosing Parquet scans. Covers flat scalar columns, trusted bounds, SQL null semantics, compound filters, clustering, projection, partition pruning and evidence from scan metrics. Reader capabilities are stated in each scenario; mathematical skip eligibility is not a promise that every engine performs it. No floating-point NaN, nested repeated-column counting, vendor-specific optimizer or complete Parquet certification coverage. Sources inspected 2026-11-19; Arrow documentation displayed v26.0.1, while format-level questions make no engine-version assumption. Sources: https://parquet.apache.org/docs/file-format/ https://parquet.apache.org/docs/concepts/ https://arrow.apache.org/docs/python/generated/pyarrow.parquet.Statistics.html https://raw.githubusercontent.com/apache/parquet-format/master/src/main/thrift/parquet.thrift https://arrow.apache.org/docs/python/generated/pyarrow.parquet.read_table.html https://arrow.apache.org/docs/python/generated/pyarrow.parquet.write_table.html https://arrow.apache.org/docs/python/parquet/index.html https://parquet.apache.org/docs/file-format/nulls/ https://parquet.apache.org/docs/file-format/pageindex/ https://parquet.apache.org/docs/file-format/bloomfilter/ https://www.postgresql.org/docs/current/functions-comparison.html
1 из 32
A flat nullable column has 80 rows and a valid null_count of 0. Min/max are absent. For x IS NULL, what may a reader using null counts do?