gffbase 0.2.1

gffbase 0.2.1

on Python PyPI

A correctness release. Several 0.2.0 defects returned plausible results that were wrong, with nothing raised. Upgrade.

pip install --upgrade gffbase

abi3 wheels for Linux (x86_64, aarch64), macOS (Intel, Apple Silicon) and Windows (x86_64) cover CPython 3.10 through 3.14; no Rust toolchain is needed.

Silent-data-loss fixes

• A query inside a feature loop ended the loop early. db.children(gene) inside for gene in db.features_of_type("gene") — the canonical gffutils loop — stopped at the first 10,000 rows without an error. create_introns, create_splice_sites and iter_by_parent_childs were truncated the same way.
• update() lost ids, parents and duplicates. New features ignored their ID attribute (the introns of db.update(db.create_introns()) became intron_6 instead of e1-e2), got no Parent edges, and a duplicate id was stored instead of refused. update() now runs new data through the create_db pipeline, atomically, as gffutils does.
• Empty or unrecognized input built an empty database. Empty, header-only, FASTA-only and binary files — and gzip not named .gz (.bgz, .GZ, no extension) — now raise EmptyInputError or are decompressed by content.
• A GTF with one bad first line was read as GFF3 by the Rust engine, so its gene/transcript hierarchy was never built.
• Quoted values (ID="g001") read differently from what was stored, and re-serialized with doubled quotes.
• DataIterator refused files that create_db reads, and a transform returning None kept the feature instead of dropping it.
• Six create_db parameters were ignored (gtf_transcript_key, gtf_gene_key, dialect, verbose, and pragmas, which never reached the ingest); a failed ingest left DuckDB's spill directory behind.

Behaviour you may notice

Each change brings gffbase in line with gffutils or its own documentation: EmptyInputError for input with no features; update() raises DuplicateIDError on a taken id (write an edit back with merge_strategy="replace"); checklines samples checklines + 1 records; add_relation(level=...) other than 1 raises; inferred GTF parents are labelled gffutils_derived in compat mode. The full list is under Changed in the changelog.

The documentation was corrected where it was wrong — versioned GENCODE ids, RefSeq's region rows, memory claims — and the migration guide now gives measured per-call latencies against gffutils: region() is faster; single-row children() and db[id] are slower, which is what the batched API is for.

How this release was verified

• Qualified in CI on Linux, macOS and Windows across Python 3.10–3.14 on both spatial index paths, with a five-corpus correctness gate, the gffutils parity suite, the pure-Python engine without the extension, and the minimum supported dependencies.
• Every fix lands with a regression test that fails on 0.2.0.
• The binaries attached here are byte-identical to the files on PyPI.

Docs: https://khchao.com/gffbase/ · Changelog: CHANGELOG.md · PyPI: https://pypi.org/project/gffbase/0.2.1/