Skip to content

curriculum(APP19): competency-derived TypeScript types and validated API evidence - #167

Merged
christophersesugh merged 2 commits into
mainfrom
curriculum/app19-competency-first-reassessment
Nov 18, 2026
Merged

christophersesugh merged 2 commits into
mainfrom
curriculum/app19-competency-first-reassessment

Conversation

@christophersesugh

@christophersesugh christophersesugh commented Nov 18, 2026 •

Copy link
Copy Markdown
Collaborator

M3 / APP19 — competency-first academic reassessment

1. Baseline, authority and scope

  • Baseline: main at 9ddddcc642fb373437004366def519d7a5fbfb2c, with APP1 #161, APP17 #163 and APP18 #165 merged. Branch: curriculum/app19-competency-first-reassessment. Curriculum source changed: curriculum/APP/APP19.md. CI regression repair also changes: tests/test_app_catalog.py (only the two APP19 malformed-Markdown negative tests; no normative/content change).
  • Normative authority: Course Format, Evidence Rules, Status Labels, Conformance. Review guidance: Design Guide; temporary review protocol revision 1.0 draft PR #150 (protocol remains unmerged) and APP architecture baseline M0 issue #152.
  • Before: APP19 — TypeScript Programming, course, core, v0.2.0 draft, APP18 direct prerequisite, 12 modules, 82 numbered lesson entries, 12 required outcomes, five required assessments, provisional 10–14 weeks/10–14 focused hours/week.
  • After: 10 coherent modules, 58 numbered lesson entries, same 12 outcome IDs, same five required assessment IDs, no metadata, prerequisite, status, duration, binding rules, cross-program, APP catalog or other course change. Count difference is a consequence, not a desired target.

2. Independent TypeScript exit competence and sources

Authentic exit performance: A learner who already writes and tests JavaScript (APP18) can predict structural type checking, represent absent/variant states, diagnose unsound assertions and incompatible consumers, design relational generic APIs with proportionate type transformations, maintain a strict and reproducible compiler/module boundary, validate untrusted runtime payloads, evolve an existing JavaScript module and independently defend a bounded typed library under unexpected change. The goal is testable TypeScript engineering, not maximum language-syntax coverage or an exercise per type feature.

Evidence informing the course boundary:

Prerequisite and specialization ownership: APP18 owns JavaScript behavior, promises, ES modules, local Node tooling and tests. APP19 owns static TypeScript modeling, compiler checking, public declaration and consumer contracts, incremental migration and type/runtime distinction. APP2 consumes this for browser APIs; APP20 adds React; APP21 handles the Node/V8/libuv/OS/service runtime; APP3 adds backend frameworks; elective APP15 adds TypeScript–Rust FFI. No DOM, server, React, Rust, remote package registry or private vendor becomes a hidden core requirement.

3. Outcome-by-outcome disposition and coverage challenges

Retain APP19-O1–O12 without rewriting their stable meaning or silently changing old completion decisions:

  • O1/O9: tsc static checking, compiler diagnostics versus emitted/runtime JavaScript, Node-native type stripping limits and local reproducibility.
  • O2/O3: inference, literals, union/intersection/tuple, objects, structural compatibility, readonly/optional state and why excess-property checks are not validation.
  • O4/O5: discriminated control flow, never exhaustion, assertion/predicate unsoundness, unknown and type holes.
  • O6: typed calls/callback compatibility, overloads and async result types with actual runtime failure behavior.
  • O7/O8: meaningful generic input/output/key relations, suitable utility/mapped/conditional/template-literal transformations, complexity critique and static negative cases.
  • O10: runtime-versus-type-only module surfaces, declarations, JavaScript interoperation, two real consumer contracts and version/change consequences.
  • O11: unknown-data validation, missing/null/version handling and adversarial negative evidence.
  • O12: controlled migration of a working JS slice, actual tests, a bounded unfamiliar change, clean reproduction and independent defense.

Identified evidence gap: Existing O8 explicitly says utility, mapped, conditional and template-literal types, but prior AS3 only asked for “a justified utility/mapped/conditional type”: this could pass without testing a template-literal relationship, and without demonstrating the difference between a mapped type and a canned utility type. Revised AS3 requires bounded, compiler-verifiable examples of each family; a small isolated consumer is acceptable where publishing the more complex type would damage the API. The examiner also judges when the public API should choose simpler types. O8 is not a license to make a gimmicky type-language course.

4. Module-by-module decisions and revised lesson intent

Former modules → revised module Entries Competency-grounded disposition
M1 → M1 Static contract and first program 6 → 4 Combine course-history/orientation/setup fragments into immediate JavaScript-versus-TS investigation, a real strict compiler check and separate run; keep historical context brief and purposeful
M2 → M2 Inference, literals, missing values 7 → 6 Merge fragmented inference/value catalog; preserve unions/intersections, exact absence, literal precision, any/unknown/never and safe assertions, observing compiler/runtime differences
M3 → M3 Structural object contracts 7 → 5 Preserve type/interface shape, excess-property counterexample, optional/readonly, index/key relationships and compile-time class contracts; do not reteach APP18 class implementation
M4 → M4 Narrowing and exhaustive states 7 → 6 Focus on observable JavaScript guards, discriminated modeling, truthiness and mutable-state limits, never exhaustion, unsound predicates, and introduced-variant repair
M5 → M5 Functions and typed async results 7 → 5 Merge parameter/receiver tutorials; retain callback variance, overload-vs-union-vs-generic selection and actual Promise rejection evidence without repeating APP18 scheduling
M6 + M7 → M6 Generics and type transformations 14 → 8 One conceptual progression from relational generic inference/constraints/key access to utility/mapped, conditional/distributive and template-literal design. Tie each to working consumer constraints and a simplicity decision, not seven unrelated exercises in “advanced” syntax
M8 + M9 → M7 Compiler, modules and typed API delivery 13 → 8 Build one coherent tsc/host/module/declaration workflow. Eliminate repeated type-only import, module-resolution, declaration, JavaScript interop and package-consumer treatments across two modules; retain meaningful strictness, emission, native stripping limitations and independently tested APIs
M10 → M8 Runtime trust boundaries 7 → 6 Separate runtime validation from erased type annotations; keep versioned payload, optional/null, truthful predicates, inherited-key safety and adversarial rejection without relabelling a cast as validation
M11 → M9 Incremental JS migration and dual verification 7 → 5 Replace repeated debugging/how-to-test topics with one APP18-to-TS migration and clear static negative vs runtime positive/negative checks, plus compatibility diagnosis
M12 → M10 One typed library and defense 7 → 5 Reuse assessed work, do not start a fifth independent project; require new bounded change, consumer reactions, corrected static/runtime outcomes and clean reproduction

Actual new distribution: M1 4, M2 6, M3 5, M4 6, M5 5, M6 8, M7 8, M8 6, M9 5, M10 5 — not equal or quota-derived. The M6 and M7 modules are longer because the mechanisms and feedback are linked; splitting them again for equal counts would recreate the original problem. Each described lesson has a primary capability and an observation, counterexample or engineering task.

5. Outcome → required assessment → evidence review

Retain APP19-AS1–APP19-AS5, all explicit public Work/Criteria/Evidence, each with different decisions:

  • AS1 — O1/O2/O3/O5/O9: predict static compatibility, literal/null/readonly and excess-property differences; deliberately find incorrect assertions by comparing actual JavaScript execution and tsc diagnostics.
  • AS2 — O2/O4/O5/O11: safely decode synthetic unknown input, report malformed/versioned data, add a new discriminant, force exhaustive consumer rework and verify adverse cases. No fake automatic runtime type guard.
  • AS3 — O3/O6/O7/O8/O10: same evolving project; one relational generic API, keyof/indexed access; distinct utility/mapped, conditional and template-literal checked examples with meaningful use or explicit exclusion from public API; two independent valid/invalid consumers, actual callback/async behavior and proportionality judgment.
  • AS4 — O1/O5/O9/O10/O12: migrate an APP18-style module with supported baseline and tsconfig; show tsc static check and a separate live executable path (actual JS emission or appropriately supported native erasable syntax). Seed and resolve module/config/type-import failure; verify old runtime regression cases offline.
  • AS5 — O1–O12: integrate one typed library and at least two consumers, reuse prior accepted evidence by reference, introduce an unfamiliar bounded contract/variant change, retest separately at compile and runtime, disclose breaking compatibility and defend trust, type-system limits and reproducibility.

Bidirectional checks: Every retained outcome is named in summative definitions; the AS3 O8 criteria now genuinely require distinct type-family evidence; the other assessment ID mappings were preserved. These criteria allow written/recorded/accessible defenses and offline local tools; they reject unfounded claims that successful tsc, .d.ts, or Node stripping validates untrusted records. No hidden mandatory course labs or five copies of the same library.

6. Cross-program effects, workload and migration

  • Direct prerequisite APP18 remains; actual merged APP18 includes explicit JS language/module/async/testing/toolchain evidence. APP2 and APP20 consume rather than repeat typed language semantics; APP21 owns Node service/runtime depth.
  • No required changes to APP catalog, APP9 graduation pre-entry evidence, other course outcome/assessment IDs, binding OEE spec or temporary protocol.
  • The 10–14-week duration and 10–14 focused hours/week remain draft planning estimates. No real observed learner times were available; a shorter outline is not evidence the schedule should automatically shrink. Required work remains substantial; future calibration must measure real practice and assessments.
  • No new paid compiler/tool, remote registry, private CI, proprietary editor, browser DOM, React, Node HTTP service or Rust specialization required.
  • This is a material editorial/course-criteria clarification with no renumbering of course outcomes or assessment IDs; historical learner evaluations remain tied to their original source/spec revision, and no automatic equivalence is asserted.
  • Any post-merge course-consumer effect should be evaluated against the actual published learner requirements, not inferred from matching course labels.

7. Findings (locator, consequence, severity, disposition)

  1. Major — M6/M7, AS3/O8: resolved. Isolated advanced type constructs did not follow one coherent real generic API and failed to explicitly assess every O8 type family. Combined the competency sequence; tightened O8 evidence without requiring type tricks in a public interface.
  2. Major — M8/M9, AS4: resolved. Compiler configuration, type-only imports, declarations, resolution, package exports and consumer compatibility were taught twice with fragmented accountability. One coherent compiler+typed API delivery stage retains diagnostic and independent runtime mechanisms.
  3. Major — AS4/AS5: resolved. Existing evidence wording permitted avoidable duplicate projects/logs. Made one evolving library and references-first evidence operational while protecting all five separate completion judgments.
  4. Minor/Major — compiler+host reasoning: resolved. Explicitly separate native Node erasable TypeScript execution, tsc --noEmit, compiler-emitted JS, tsconfig, ambient declaration vs real runtime API and true module-resolution behavior; avoid false safety claims.
  5. Minor — M1–M5, M10–M12: resolved. Condensed isolated history/setup, repeated JavaScript concepts, type nomenclature, redundant static test and capstone checklists where APP18 already establishes their foundations; retained actual new failure reasoning.
  6. Retained: structural vs runtime meaning, truthfulness of predicates, public compatibility and negative testing. The revised course intentionally has two eight-lesson modules because complexity warrants a longer coherent stage; this is not a new equal-length pattern.
  7. Outstanding: measured learner workload, actual independent assessor usability, runtime compiler/version-specific experiments and educational effectiveness require real evidence. Do not promote draft to tested/verified from editorial checks.

8. Verification completed and not completed

  • Branch refetched from GitHub and compared against pinned main: initial curriculum commit: one modified course file; after CI repair: exactly two changed files, curriculum/APP/APP19.md and tests/test_app_catalog.py. The latter changes only broken test-fixture lookup, not curriculum requirements.
  • Static checks confirmed: ten sequential numbered modules; module Outcomes and Lessons sections; all 58 individual lessons correctly numbered within each module; all twelve unique declared outcome IDs; five stable required assessment IDs; Outcomes/Work/Criteria/Evidence for each required assessment; every outcome appears in required assessments; required top-level sections remain intact. Cross-checked existing APP19 catalog tests and repository outcome checker for no APP19-specific fixed lesson/module quota.
  • Reviewed actual merged APP18 and neighboring APP2/APP21 boundaries; checked relevant official TypeScript, Node and academic sources (links above).
  • Observed CI (signed head f9a81b4): validate-app, markdownlint, inventory, internal/public links, and DCO passed. The outcome-curriculum workflow executed 263 tests: two failed in tests/test_app_catalog.py before the curriculum validator step, because test mutations assumed obsolete exact APP19-AS4 prose. After commit 6d29aa8, those same tests locate the stable APP19-AS4 summary row and inject malformed backtick/pipe states irrespective of editorial wording; checks on the updated head are pending/re-running. The new repair commit needs the maintainer's DCO sign-off. Not executed locally: full Python suite, local TypeScript/Node experiments, independent learner evaluation. Do not imply academic conformance, verified readiness or CI success while checks are unfinished.
  • Human reviewer and CI must independently assess outcome-to-evidence quality and actual validity before merge.

9. Decision and next course

Submit for maintainer academic review and repository CI. Merge only with maintainer approval; do not merge/delete this branch or merge temporary protocol PR #150 as part of this course work. Next APP M3 course in prerequisite-valid sequence after APP19 is merged: APP4 — Databases & Data Modeling, then APP2, APP21 and subsequent core courses per #152.

Branch: curriculum/app19-competency-first-reassessment. Base: main.

CI correction after initial review (2026-11-18)

  • Outcome curriculum CI on signed head f9a81b4 ran 263 unit tests and failed exactly two tests: test_app19_assessment_summary_table_rejects_unclosed_code and test_app19_assessment_summary_table_rejects_truncated_row. Both used hardcoded strings from the former APP19-AS4 evidence summary, so they raised before invoking the malformed-Markdown validator after the new curriculum text was authored. No content-only APP19 validation failure was reported by that run.
  • Follow-up 6d29aa8 updates only those two negative tests. They identify the real APP19-AS4 table row by stable assessment ID, remove a closing inline-code backtick in one mutation, and remove its final pipe separator in the other; both assert that the validator reports its expected malformed-row diagnosis. The original 4→3 pipe and even→odd backtick mutations were independently checked before pushing; the rest of the test suite is unchanged.
  • The updated GitHub Actions run must be observed for completion. DCO sign-off on this new repair commit is still required from the maintainer, even though the prior course commit had already been signed. No branch was merged, closed or deleted.

… type/runtime evidence

Signed-off-by: Christopher S. Aondona <christohybrid185@gmail.com>
@christophersesugh
christophersesugh force-pushed the curriculum/app19-competency-first-reassessment branch from d78af02 to f9a81b4 Compare November 18, 2026 13:13
Signed-off-by: Christopher S. Aondona <christohybrid185@gmail.com>
@christophersesugh
christophersesugh force-pushed the curriculum/app19-competency-first-reassessment branch from 6d29aa8 to cb0762b Compare November 18, 2026 13:25
@christophersesugh
christophersesugh merged commit 46b4b31 into main Nov 18, 2026
7 checks passed
@christophersesugh
christophersesugh deleted the curriculum/app19-competency-first-reassessment branch November 18, 2026 13:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant