zearch — searches regex directly on Re-Pair grammar-compressed text without decompression
grep — GNU grep on plain (uncompressed) text
ripgrep — fast ripgrep on plain (uncompressed) text
hyperscan — Intel Hyperscan engine on plain (uncompressed) text
lz4|hyperscan — decompress with LZ4, then search with Hyperscan
zstd|hyperscan — decompress with Zstandard, then search with Hyperscan
lz4|grep — decompress with LZ4, then search with grep
zstd|grep — decompress with Zstandard, then search with grep
lz4|ripgrep — decompress with LZ4, then search with ripgrep
zstd|ripgrep — decompress with Zstandard, then search with ripgrep
repair — Re-Pair decompression only (no search; decompression baseline)
lz4 — LZ4 decompression only (no search; decompression baseline)
zstd — Zstandard decompression only (no search; decompression baseline)
gzip — Gzip decompression only (no search; decompression baseline)
Overview
The running time shown for each regular expression is the confidence interval computed over 30 runs, measured after a "warming up" run. When the confidence intervals of two experiments do not overlap then we have enough statistical evidence to claim that one tool outperforms the other on the given experiment.
If an execution takes more than 10 times the time required by zearch it is considered a timeout.