B612-Asteroid-Institute / adam_core / 21831350943

83%

Build Type

push

github

Committed by web-flow

Commit Message

Preallocate chunk outputs for 2-body propagation/ephemeris (#182)

* Optimize 2-body chunk accumulation via preallocation

Replace repeated np.concatenate() in 2-body propagation and ephemeris generation with preallocated output buffers, and add regression tests ensuring padded chunk rows are not included.

* Add Ray backend for 2-body propagation and ephemeris

Add optional Ray-based parallelization to `propagate_2body` and `generate_ephemeris_2body` via `max_processes` and task `chunk_size`, reusing the existing JAX padded-chunk kernels inside each worker. Include tests verifying Ray and serial outputs match.

* Assume Ray is always available

Remove optional Ray import/guards now that Ray is a required dependency, and run the Ray parity tests unconditionally.

* ray: disable dashboard and avoid fixed metrics port

Pass include_dashboard=False and _metrics_export_port=0 in initialize_use_ray for both auto-connect and local start paths, and update tests.

* ray: disable metrics export endpoint

Default initialize_use_ray to include_dashboard=False and _metrics_export_port=-1 (disable metrics export) to avoid metrics-related setup/timeouts in library usage.

* ray: stop overriding metrics export port

Remove the default _metrics_export_port override from initialize_use_ray and update mocked tests accordingly.

* Fix VariantOrbits ephemeris observer pairing

Ensure VariantOrbits propagation output is stably ordered by (orbit_id, variant_id, time) so ephemeris generation pairs propagated states with the correct observer rows.

Add a regression test covering time-major variant propagation.

* Phase angle as proper function, experimental sbdb implementation

* Commit to specific ordering for prop and ephemeris

* include light time in variants

* linting

* separate functions and combined for phase and magnitude

* linting

* remove unused import

* one more try

Run Details

1004 of 1127 new or added lines in 20 files covered. (89.09%)

2 existing lines in 2 files now uncovered.

11939 of 14499 relevant lines covered (82.34%)

0.82 hits per line