Skip to content

Development

Jump to bottom
github-actions[bot] edited this page Oct 30, 2026 · 4 revisions

Development

Layout

brucetyper/
  __init__.py       the version
  assays.py         every primer and expected band size, and nowhere else
  insilico_pcr.py   primer matching and amplicon finding
  genotype.py       sizes to canonical bands to a call
  cli.py            argument parsing and the reports; the brucetyper entry point
bruce_ladder.py     a shim at the root, so python bruce_ladder.py still works
bruce_rates.py      legacy 2023
suis_rates.py       legacy 2023

pip install -e . for a working install from a clone. See Installation.

Tests

python -m unittest discover -s tests -v

The suite runs the CLI as python -m brucetyper.cli, so it exercises the same entry point the installed brucetyper command uses, and runs unchanged from a clone or from an unpacked source distribution. One test covers the bruce_ladder.py shim and skips when it is absent, as it is in the distribution.

The suite builds synthetic genomes whose products reproduce each expected band profile, runs the CLI over them and checks the calls. It exercises the pipeline end to end; it is not a validation of the assays against real Brucella sequence.

Legacy scripts

bruce_rates.py and suis_rates.py are the original 2023 scripts, kept as they were apart from three fixes:

  • they no longer hard-code an input path
  • bruce_rates.py no longer divides by zero on a species that is absent from the report, which killed it on its first iteration for any report without an abortus genome
  • its wrongPreID and notID counters now match the call against the species list by substring rather than by equality, which is what that list holds: 'abort' has to be found inside abortus. wrongPreID compared the species to the list it had just been removed from, so it could never fire at all

A third 2023 script, suis_ladder.py, was removed in 1.1.0. It called genotypes from the consolidated report of a separate in silico PCR tool, a step brucetyper now does itself, and it was never finished: write_output was a stub that wrote the literal string test, and do_amos_pcr wrote one - row per profile that did not match instead of one row per sample. It remains in the history as git show e13310a:suis_ladder.py.

Both *_rates.py scripts score calls against species or biovar names embedded in file names, and take the report to score as their first argument:

python bruce_rates.py results/report_bruce.tsv

Coverage

Coverage is measured in CI and reported to Codecov. The suite drives the CLI through subprocess, and brucetyper/cli.py fans out over a multiprocessing pool, so .coveragerc turns on parallel mode and the workflow sets COVERAGE_PROCESS_START; without that only the parent process is measured and the figure reads about 35% instead of the real one.

Every file in the repository is measured. Nothing is excluded from the figure except the suite itself, so the number is not flattered by an omit list:

Module Coverage
brucetyper/__init__.py 100%
brucetyper/assays.py 100%
brucetyper/cli.py 100%
brucetyper/genotype.py 100%
brucetyper/insilico_pcr.py 100%
bruce_ladder.py (the shim) 100%
bruce_rates.py (legacy 2023) 100%
suis_rates.py (legacy 2023) 100%
Total 100%

This arrived in 1.1.0. Before it the two legacy *_rates.py scripts had no tests at all and the total was 59%, and the pair of bugs in bruce_rates.py above is exactly what untested code had been hiding. tests/test_rates.py now runs both scripts the way this page documents and reads their printed blocks back, including the cases that used to crash.

brucetyper

Repository · Issues

Clone this wiki locally