master: project roadmap and phase tracking

[timini]

Project Roadmap

This issue tracks the overall plan for getting the horn simulation pipeline from its current state (scaffolding works, physics don't) to a working tool that produces meaningful acoustic simulation results.

Current State

The pipeline has good scaffolding — containerised packages, Nextflow orchestration, CI/CD, parallel frequency bands — but the physics are non-functional:

• BEM coupling (bempp.x.dolfinx) has never existed in any pip release
• SPL metric is an L2 norm over the full volume (physically meaningless)
• Only conical horn geometry is implemented (spec calls for 4 profiles)
• Analysis is limited to a single SPL-vs-frequency plot
• mesh_size is never forwarded to the solver
• Groovy integer division silently drops frequency samples

Phase 1: Correctness — FEM-only pipeline producing real results

Goal: Fix the bugs that make the current pipeline output meaningless. After this phase, running the pipeline with a conical horn should produce physically plausible SPL data.

• bug: SPL metric is physically meaningless (L2 norm over full volume) #10 — Replace L2-norm SPL with outlet-surface pressure integration
• bug: params.mesh_size is never forwarded to the solver in main.nf #11 — Forward params.mesh_size from main.nf to the solver CLI
• bug: max_freq_mesh parameter accepted but never used for adaptive meshing #12 — Implement frequency-adaptive meshing (max_freq_mesh → element size < λ/6)
• bug: Groovy integer division drops frequency samples in main.nf #14 — Fix Groovy integer division dropping frequency samples
• bug: geometry tests use wrong kwarg (output_dir instead of output_file) #4 — Fix geometry test kwarg (output_dir vs output_file)
• milestone: FEM-only pipeline producing physically meaningful results #40 — Milestone: FEM-only pipeline producing physically meaningful results

Phase 2: BEM Research — determine if hybrid FEM-BEM is viable

Goal: Answer the question: can we get exterior radiation modelling working? This runs in parallel with Phase 3.

• spike: investigate BEM coupling approaches — is bempp.x.dolfinx viable? #39 — Spike: investigate BEM coupling approaches
  • Option A: Build bempp with dolfinx coupling from source
  • Option B: Use bempp standalone (manual data transfer at boundary)
  • Option C: PML absorbing layers in pure FEM (no bempp needed)
  • Option D: Analytical radiation impedance BC at the mouth
• bug: BEM coupling in solver.py is a non-functional placeholder #9 — Replace the non-functional BEM placeholder based on spike outcome
• bug: run_simulation nf-test passes dummy STEP file — test will always fail #16 — Fix run_simulation nf-test (needs real STEP fixture from geometry stage)

Phase 3: Features — horn profiles, analysis, driver DB

Goal: Expand the pipeline from "one conical horn with one plot" to the feature set described in the spec. Can start once Phase 1 is done.

• feat: implement exponential and hyperbolic horn geometry generators #33 — Implement exponential and hyperbolic horn geometry generators
• feat: implement analysis features (impedance plots, polar plots, scoring, driver DB) #35 — Implement analysis features (impedance plots, polar plots, scoring)
• bug: horn-analysis Docker build fails due to unused psycopg2-binary/pdfplumber deps #3 — Clean up horn-analysis deps (psycopg2/pdfplumber are for future driver DB)
• bug: get_driver_params.py references non-existent data/drivers.json #5 — Fix or remove get_driver_params.py (references non-existent drivers.json)
• test: horn-analysis test_analysis.py is a placeholder (assert True) #21 — Write real tests for horn-analysis (currently assert True)

Phase 4: Cleanup and polish

Goal: Pay down tech debt. Can be done opportunistically alongside other phases.

• cleanup: remove dead solver_runner.py #15 — Remove dead solver_runner.py
• infra: CI test_packages pushes to DockerHub on every PR — decouple build from publish #19 — Decouple CI build from DockerHub publish
• arch: horn-core data structures are unused — integrate shared data model #22 — Integrate horn-core shared data model (currently unused)
• cleanup: horn-core __init__.py exports nothing #23 — Make horn-core __init__.py export its data structures
• bug: dolfinx and gmsh missing from horn-solver pyproject.toml dependencies #24 — Add dolfinx and gmsh to horn-solver pyproject.toml
• infra: Dockerfiles hardcode python3.10 in COPY paths #25 — Remove hardcoded python3.10 paths in Dockerfiles
• cleanup: remove committed *.egg-info directories from git #27 — Remove committed *.egg-info directories
• cleanup: remove archived horn-bem-solver package #28 — Remove archived horn-bem-solver package
• cleanup: remove dead params.test_outdir from main.nf #29 — Remove dead params.test_outdir from main.nf
• bug: plotter.py never calls plt.close() — memory leak in batch use #30 — Fix plt.close() memory leak in plotter
• cleanup: add compare_horns as a pyproject.toml script entry point #31 — Add compare_horns as a pyproject.toml script entry point

Phase 5: Full FEM-BEM coupling

Goal: Implement the "gold standard" hybrid FEM-BEM simulation from the design spec. Depends entirely on the outcome of #39.

• feat: implement real FEM-BEM coupling for acoustic simulation #34 — Implement real FEM-BEM coupling for acoustic simulation

Recommended execution order

Phase 1 (correctness)
    |
    +---> Phase 2 (BEM spike) -----> Phase 5 (if viable)
    |         |
    +---> Phase 3 (features) <------ inform from spike
    |
    +---> Phase 4 (cleanup, anytime)

Phase 1 is the critical path. Phase 2 and 3 can run in parallel once Phase 1 is done. Phase 4 is background work.

[timini]

Roadmap update — horn profiles

PR #72 adds 4 new horn profiles (tractrix, OS, Le Cléac'h, CD), bringing the total from 3 to 7. This completes the deferred tractrix work from #46 and significantly expands the design space for auto/fullauto pipelines.

Phase 3 progress:

• feat: implement exponential and hyperbolic horn geometry generators #33 ✅ (exponential + hyperbolic, done in feat: Phase 3 — horn profiles + analysis features #55)
• feat: exponential, tractrix, hyperbolic horn profiles #46 ✅ (tractrix now done in feat: add tractrix, OS, Le Cléac'h, and CD horn profiles #72, plus OS/Le Cléac'h/CD)
• feat: implement analysis features (impedance plots, polar plots, scoring, driver DB) #35 ✅ (analysis features done across multiple PRs)