Row-Group Control, Faster S3 Reads, Footer Caching: Hardwood 1.1.0.Beta2 Is Out

Hardwood 1.1.0.Beta2, a Java library for reading and writing Apache Parquet files, adds explicit row-group boundary control, the ability to skip bad records during writes without aborting, column/offset indexes, and Bloom filters for point queries. Reader performance is significantly improved, especially for S3-backed reads, via coalesced bloom filter and page index fetches, prefetching, lazy multi-file opening, filter-column skipping using row-group stats, and dictionary-space predicate evaluation. New features include an experimental MetadataSource API for caching Parquet footers (keyed by S3 ETag) and id-based dictionary access for faster aggregation over dictionary-encoded columns. The release also adds new filter predicates, nanosecond-precision timestamp support, in-memory output files, several CLI improvements, bug fixes, and some breaking changes (reworked exception model, reordered index-based accessors) ahead of the 1.1 release candidate.

Nguồn: https://www.morling.dev/blog/row-group-control-faster-s3-reads-footer-caching-hardwood-1-1-0-beta2. 8 Sync News chỉ tóm tắt và dẫn link; bản quyền nội dung thuộc tác giả và nguồn gốc.