nx_fastq.nx

nx_fastq.nx -- FASTQ single-record stream parser. license_tier: INDEPENDENT_REDERIVE genealogy_id: international-research-sources/cock-2010-fastq-format G2.0b of NISHI_GENOMICS_SUBSTRATE_ROADMAP.md. The bridge between raw sequencer-emitted byte streams and the substrate's typed DNA + Phred primitives. Parses ONE record per call (4-line format) and emits offset metadata; the caller composes downstream with: dna_pack_ambig(buf + seq_start, seq_len, out_bases, out_nbits) nx_phred33_decode_string(buf + qual_start, qual_len, out_qs) FASTQ canonical 4-line format (Cock 2010): line 1: '@' + identifier [+ optional description] line 2: nucleotide sequence (single line in G2.0b -- see deferred) line 3: '+' [+ optional repeated identifier] line 4: quality string, len == seq len, Phred+33 encoded Validation enforced: - record starts with '@' (offset position 0) - separator line starts with '+' - seq_len == qual_len What G2.0b does NOT do (deferred): - Multi-line sequence / quality (rare in modern data; complicates parsing because qual chars include '@' and '+' so line-by-line scanning ambiguates without lookahead) -- G2.0b.2 - gzip / bgzip decompression -- callers pre-decompress; nx_gzip wrapper exists, integration is G2.0b.3 - Header field parsing (Illumina-style "INSTRUMENT:RUN:FLOWCELL:..." metadata) -- G2.0b.4 - Phred encoding auto-detection (+33 vs +64) -- G2.0a.1 heuristic - Paired-end record pairing across R1/R2 files -- G2.0b.5 API: nx_find_lf(buf, start, end) -> i64 offset of next LF or -1 fastq_parse_one_record(buf, buf_len, offset, out_meta) -> i64 out_meta layout (7 i64s):