Upstream contributions

Changes to FastQC

The Rust rewrite works as a development sandbox for FastQC. Ideas get prototyped and benchmarked here, and the equivalence tests catch bugs. What works then goes back to Java FastQC as pull requests. The aim is a better canonical tool, not a replacement for it. Many of these changes shipped in FastQC v0.13.0.

Released in FastQC v0.13.0

Performance

#197 — About 3x faster with parallel processing. Each file now runs through a small pipeline: one reader thread (decompression and parsing) feeds up to three processor threads, each of which owns a subset of the QC modules. -t is now a total CPU budget. This work started with Paolo Di Tommaso's experiments. Each change was then benchmarked on its own to find the smallest change with the biggest effect.
#199 — Removes the per-sequence char[] allocations in seven QC modules.

SVG graphs

#200 — SVG is now the default graph format in the HTML report. Graphs are sharp at any zoom level and easy to edit for publication. Use --png to get PNG graphs again.
#188 — Smaller SVG output. Uses presentation attributes, merges line segments into polylines and merges runs of same-colour rectangles. Data lines are also thicker.

Correctness

#187 — Adapter Content showed empty trailing rows when --min_length was set. The Rust equivalence tests found this. Upstream then reworked --min_length into a real read-length filter, and added --max_length.
#209 — Encoding detection misread high-quality data (no base below Q31) as Illumina 1.5, so every quality score was 31 too low. The Rust rewrite found this with BAM input. In v0.13.0 Phred+33 is the default for all input, and --phred64 is now an opt-in.
#203 — Basic Statistics now reports mean and median read length.

Project infrastructure

#156 — GitHub Actions workflow that builds the release assets automatically.
#155 — Markdown README and installation docs.

Merged into a feature branch

These are merged into the upstream liberationfont branch, but that branch is not yet in master (#216). #217 merges it into master.

#185, #186, #194 — Bundle the Liberation Sans font and anti-alias graph text. This removes the dependency on system fonts, so graphs look the same on all machines. It also aligns text metrics between Java and Rust.

Open pull requests

#205 — Mean read length to one decimal place, and the standard median for an even number of reads.
#196, #195 — Move the HTML report layout into template fragments, and fix some small HTML problems.
#161 — HTML report design refresh. The Rust rewrite already has this as the modern template (--template modern).

How it works

The Rust rewrite follows upstream releases (see strategy). The equivalence tests run the Java and Rust versions on the same inputs and compare their outputs. A difference is either a Rust bug or a Java bug. Java bugs become upstream issues and pull requests. When upstream releases a fix, the Rust rewrite gets the same change, so the two outputs stay identical.