Skip to content

CHANGELOG.md

Latest commit

 

History

History
27 lines (19 loc) · 1.15 KB

CHANGELOG.md

File metadata and controls

27 lines (19 loc) · 1.15 KB
Raw

HypGrep Changelog

[0.3.3]

  • Coalesce adjacent candidate blocks into a single read per row group, reducing the number of HTTP range requests per search

[0.3.2]

  • Read one row group per fetch when building the index, reducing memory use on large sources

[0.3.1]

  • Add textColumns option to index only a subset of the source's string columns

[0.3.0]

  • Default n-gram tokenizer now keeps the JSON delimiters " : { } inside n-grams, making structured/JSON greps highly selective at almost no index-size cost
  • The kept-character set is configurable via the ngramChars option and recorded in index metadata; indexes built before this version keep working unchanged

[0.2.1]

  • Partition buckets during index build to handle V8's Map size limit on large datasets

[0.2.0]

  • Switch to n-gram grep over Parquet: case-insensitive substring search via a precomputed index
  • Regex/grep query support, extracting literal n-grams to prune candidate blocks

[0.1.1]

  • parquetFind matches rows as a contiguous phrase, like ctrl-F

[0.1.0]

  • Initial release with keyword search on Parquet files via a compact index, with BM25 ranking.