Target manifest¶

A manifest lists targets explicitly. Use one for a runner that has no discovery, or to adjust what discovery finds. diffcone discover -o targets.json writes a manifest of the discovered targets, and --targets targets.json reads one.

{
  "source_roots": ["src", "."],
  "targets": [
    {
      "runner": "pytest",
      "runner_id": "tests/test_calc.py::test_add",
      "entry_symbol": "tests.test_calc.test_add",
      "lifecycle_dependencies": ["tests.conftest.db"]
    },
    {
      "runner": "asv",
      "runner_id": "bench_calc.TimeCalc.time_add",
      "entry_symbol": "benchmarks.bench_calc.TimeCalc.time_add",
      "lifecycle_dependencies": ["benchmarks.bench_calc.TimeCalc.setup"]
    }
  ]
}

Field	Description
`source_roots`	Optional. The source roots the names below use. `--source-root` overrides it.
`runner`	Which runner runs the target. `diffcone run` supports `pytest` and `asv`; other runners can be planned but not run.
`runner_id`	The name the runner uses for the target: a pytest node ID (without parameters) or an ASV benchmark name.
`entry_symbol`	The test or benchmark function, as an import path: module, then classes, then function.
`lifecycle_dependencies`	Code the runner runs for this target outside its own function: pytest fixtures, ASV `setup` and `setup_cache`, or a module (so module-level code such as `pytestmark` counts).

A manifest may also be just the list of targets, without the surrounding object.

When you combine a manifest with --discover, a manifest entry replaces a discovered target with the same runner and runner_id.

A dependency written fixture:<name> marks a fixture whose definition couldn't be found. diffcone can't tell what such a fixture depends on, so it selects the target on every change.