2.1.0 - 2026-03-27

Release Notes

Now ~25% faster than the old C implementation of OrfM (v1.4). Cleaned up API, fixed multi-line FASTA bug.

Added

• Correctly handle wrapped (multi-line) FASTA sequences. Previously raw_seq() was
  used which includes embedded newlines, producing incorrect ORFs for wrapped input.
• FASTQ validation: the first record is checked character-by-character for
  valid IUPAC DNA/RNA characters; subsequent records detect invalid characters
  via the translation table (a codon containing a non-IUPAC byte produces the
  sentinel value 0, triggering a panic with position and character details).
• flate2 switched to the zlib-ng backend for all gzip decompression
  (needletail and the fastq crate), roughly halving wall-clock time on
  gzipped inputs compared to the previous miniz_oxide default.
• Orf::orfm_id() method returns the ORF identifier
  ({source_seq_name}_{start}_{frame}_{orf_number}).
• Orf::name() method returns the full ORF name (orfm_id plus comment if present).

Fixed

• revcom_base now handles all IUPAC ambiguity codes (R, Y, S, W, K, M, B, D, H, V)
  and RNA uracil (U). Previously these were all mapped to N, losing information in
  reverse-strand transcripts.
• Parse errors in FASTA/FASTQ input now produce an error message and exit instead of
  silently truncating output. Previously while let Some(Ok(...)) in the iterators
  and process_with_needletail would treat any Err as end-of-file.

Changed

• FASTQ parsing switched from needletail to the fastq crate (zero-copy,
  callback-based). Format is auto-detected by inspecting the first byte of input;
  FASTA input continues to use needletail.
• Aho-Corasick automaton is now case-insensitive, removing a per-sequence uppercase
  copy allocation.
• Newline detection is integrated into the Aho-Corasick scan itself: raw_seq() is
  used optimistically (zero-copy), and if a newline is encountered during the stop
  codon search, the sequence is re-processed via seq(). Once a newline is detected,
  all subsequent records in the file use seq() directly.
• find_orfs is now the public single-sequence entry point and always returns
  Vec<Orf> (normalizing wrapped input automatically). See the find_orfs docs
  for the performance trade-off vs. the iterator API.
• Orf.seq_name renamed to Orf.source_seq_name for clarity. The header() method
  has been replaced by orfm_id() and name().
• Orf.seq_name and Orf.seq_comment changed from String to Arc<str>, reducing
  per-ORF heap allocations to a cheap atomic reference count increment.
• Library iterators (OrfFileIter, OrfReaderIter) no longer clone each Orf on
  iteration; they yield owned values directly via IntoIter.
• Library API cleaned up: version_at_least, get_codon_table, revcom_base,
  and is_valid_iupac are no longer public.
• Benchmark data files moved to benchmark/data/.
• Benchmark summary now prints a box-drawing table with C vs Rust wall-clock
  times and ratio (median of 3 replicates).

Removed

• -r VERSION flag and version_at_least function (dead code).
• Orf::header() method — use orfm_id() or name() instead.

Install orfm 2.1.0

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/wwood/OrfM/releases/download/2.1.0/orfm-installer.sh | sh

Install prebuilt binaries via powershell script

powershell -ExecutionPolicy Bypass -c "irm https://github.com/wwood/OrfM/releases/download/2.1.0/orfm-installer.ps1 | iex"

Download orfm 2.1.0

File	Platform	Checksum
orfm-aarch64-apple-darwin.tar.xz	Apple Silicon macOS	checksum
orfm-x86_64-apple-darwin.tar.xz	Intel macOS	checksum
orfm-x86_64-pc-windows-msvc.zip	x64 Windows	checksum
orfm-aarch64-unknown-linux-gnu.tar.xz	ARM64 Linux	checksum
orfm-x86_64-unknown-linux-gnu.tar.xz	x64 Linux	checksum
orfm-x86_64-unknown-linux-musl.tar.xz	x64 MUSL Linux	checksum