Skip to content

fix(data): stale notes, one outline for the descriptions (T2 of #449) - #460

Merged
ocots merged 5 commits into
mainfrom
fix/t2-review
Nov 20, 2026
Merged

ocots merged 5 commits into
mainfrom
fix/t2-review

Conversation

@ocots

@ocots ocots commented Nov 19, 2026 •

Copy link
Copy Markdown
Member

Refs #449 (T2, findings). Applied conventions: ct-conventions v0.14.0, ct-agents v0.5.0.

Fixes findings 2–4 of the T2 review (finding 1 was wrong, see below).

Changes

  1. bryson_denham: JuMP :draft → :valid. Reverted: the note is right. The OptimalControl model declares x ∈ R², whose components are x₁, x₂, and the JuMP model names them x1, x2; the first CI run failed on it (test_features). The review had only looked at the parameter names. Unchanged here.
  2. mountain_car: note rewritten. The gap (102.381 against 102.398) comes from modeler=exa solver=madnlp, in both OptimalControl modelings. It does not come from the scalarized modeling: Ipopt gives 102.3976 in every modeling.
  3. One outline for the descriptions, checked by test_outline in test_consistency.jl:
    • the order is Mathematical formulation, Parameters, Qualitative behaviour, Characteristics, References;
    • Parameters and Qualitative behaviour are optional, the others required;
    • robbins gets its "Mathematical formulation" heading, space_shuttle its "References" (it was "References and relevance"), and the horizontal rules between sections are removed (4 files).
  4. "Characteristics" keeps what cannot be computed (one source of truth). The bullets that restated the dimensions, the free or fixed final time, or the kinds of constraints are removed: the problem page shows them from features (docs/problems.jl, facts). What remains is the nature of the dynamics, the structure of the solution, and why the problem is a benchmark.

Also: a broken \text in the formulation of balanced_field (a tab had replaced \t: ext{ m (35 ft)}).

Left open

  • The hand-written parameter sections (17 descriptions, three names): option (a) chosen by the maintainer, in a follow-up. The problem page already generates a ### Parameters section from the metadata (parameters_section in docs/problems.jl), so 5 pages show two sections named "Parameters", and the values are written twice. The hand-written ones also give the meaning and the unit of each parameter, which the metadata do not have. Either the metadata gain the meaning and unit of each parameter, shown by the generated section, or the descriptions keep only the meaning, without values. This is the maintainer's call, in a follow-up.
  • balanced_field has no "Qualitative behaviour". Its source (the Dymos example) does not describe the solution. The section stays optional in the descriptions; the page of a problem without one gets a placeholder that points to the figures (with_behaviour in docs/problems.jl), so every page has the same outline. It can be written from the cron plots.

Test plan

  • Locally (Julia 1.12): Pkg.test(; test_args=["core"]): 1334 passed, 1 broken (the existing @test_skip of test_discovery_cache.jl).
  • CI.

🤖 Generated with Claude Code

- bryson_denham: JuMP is valid (it passes the convergence and robustness
  crons with the objective of OptimalControl); its note on ASCII names
  was no longer true.
- mountain_car: the note puts the 1.6e-4 gap on the configuration
  exa + MadNLP, in both OptimalControl modelings, not on the scalarized
  modeling.
- Descriptions: "Characteristics" keeps what cannot be computed; the
  dimensions, final time and kinds of constraints are shown on the
  problem page from `features` (one source of truth). Horizontal rules
  removed, `robbins` gets its "Mathematical formulation" heading,
  `space_shuttle` its "References", and a broken `\text` of
  `balanced_field` is fixed.
- test_consistency: every description follows one outline.

Refs #449

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@ocots ocots added the run ci github-runner Trigger the CI workflow on GitHub-hosted runners label Nov 19, 2026
ocots and others added 2 commits November 19, 2026 22:23
… the figures

Refs #449

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The OptimalControl model declares `x ∈ R²`, whose components are x₁, x₂;
the JuMP model names them x1, x2. The comparison of the modelings fails on
it (test_features).

Refs #449

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ocots and others added 2 commits November 20, 2026 08:23
# Conflicts:
#	ext/MetaData/mountain_car.jl
The 102.398 with Ipopt is the one of JuMP and the OptimalControl
modelings; the ModelingToolkit modeling (#452) solves another problem
(shooting).

Refs #449

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@ocots
ocots merged commit 0e2f080 into main Nov 20, 2026
10 of 11 checks passed
@ocots
ocots deleted the fix/t2-review branch November 20, 2026 07:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

run ci github-runner Trigger the CI workflow on GitHub-hosted runners

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant