Reader Reference

Reader Options

Read-time behaviour is configured with an immutable ReaderConfig, passed to ParquetFileReader.open(...). It is separate from HardwoodContext, which holds the shared runtime resources (the decode thread pool and native decompression pools): a single context can back reads with different ReaderConfigs.

import dev.hardwood.HardwoodContext;
import dev.hardwood.reader.ParquetFileReader;
import dev.hardwood.reader.ReaderConfig;

ReaderConfig config = ReaderConfig.builder()
        .option("hardwood.fixed-list-fast-path", "true")
        .build();

try (HardwoodContext context = HardwoodContext.create();
     ParquetFileReader reader = ParquetFileReader.open(inputFile, context, config)) {
    // ...
}

Obtain the defaults with ReaderConfig.defaults(). The open(inputFile) and open(inputFile, context) overloads use the defaults.

Options are string-keyed and keys are matched case-sensitively; boolean option values are compared case-insensitively. An unrecognised key is ignored, so a retired option does not break callers, and is reported in the log.

Option	Default	Description
hardwood.fixed-list-fast-path	false	Set to "true" to decode fixed-size LIST columns (e.g. embedding vectors, where every row holds the same number of non-null elements) without reconstructing per-row definition and repetition levels.
hardwood.metadata-filtering	true	Set to "false" to disable metadata-based filtering: no row groups or pages are skipped from min/max statistics, bloom filters, dictionary pages, or page indexes, and filter predicates are instead evaluated against every decoded row. Filtered results then depend only on the data pages, for files whose footer or page-index metadata is unreliable.

Limits

A column chunk read without an OffsetIndex must be at most 2 GB of compressed data; a chunk with an OffsetIndex is read in page groups and has no such limit. Files larger than 2 GB are read both locally and from S3, except by the in-memory (ByteBuffer) backend and by an S3 file opened with RangeBacking.SPARSE_TEMPFILE, which the dive TUI uses for every file; both are limited to 2 GB per file. For datasets that don't fit a single supported file, split the data into multiple files at write time and read them as one; see Read Multiple Files as One Dataset.