Skip to content

TESTING.md

Latest commit

 

History

History
77 lines (59 loc) · 4.25 KB

TESTING.md

File metadata and controls

77 lines (59 loc) · 4.25 KB
Raw

Testing fhir-dsl

This document indexes per-package test prompts used to drive comprehensive unit-test generation against the fhir-dsl codebase. The prompts are meant to be fed — verbatim — to a capable coding AI (Claude, GPT, Gemini, etc.). Each prompt is self-contained: a model that has only the repo and its package's PROMPT.md has everything it needs to produce useful, non-duplicate tests.

Per-package prompts

Package Prompt Scope
@fhir-dsl/cli packages/cli/test/PROMPT.md fhir-gen generate flag parsing, validation, exit codes
@fhir-dsl/core packages/core/test/PROMPT.md client, search/read/transaction builders, validation chain, type-level inference
@fhir-dsl/example packages/example/test/PROMPT.md consumer smoke test against committed generated tree
@fhir-dsl/fhirpath packages/fhirpath/test/PROMPT.md FHIRPath builder grammar + evaluator against the HL7 spec
@fhir-dsl/generator packages/generator/test/PROMPT.md parser, every emitter, end-to-end generate() to tmpdir
@fhir-dsl/runtime packages/runtime/test/PROMPT.md retry matrix, pagination, abort, FhirRequestError shape
@fhir-dsl/smart packages/smart/test/PROMPT.md SMART App Launch v2, PKCE, backend services JWT, scope parser
@fhir-dsl/terminology packages/terminology/test/PROMPT.md registry, ValueSet compose + expansion, CodeSystem hierarchy
@fhir-dsl/types packages/types/test/PROMPT.md compile-only *.test-d.ts coverage of base FHIR types
@fhir-dsl/utils packages/utils/test/PROMPT.md naming, type-mapping, logger

Workflow

  1. Hand the target prompt to a coding-capable AI. The prompt tells it which files to read, what the existing coverage is, what scenarios to cover, and where to write tests.

  2. The model writes new tests under packages/<pkg>/test/. Existing colocated tests in packages/<pkg>/src/ are not touched unless explicitly extending them.

  3. Before accepting the model's work, run all three gates from the repo root:

    pnpm test
pnpm lint
pnpm -r typecheck

    All three must be green. No skipped tests without a documented reason.

Where tests live

vitest.config.ts already picks up both locations:

  • packages/*/src/**/*.test.ts — colocated tests (existing style).
  • packages/*/test/**/*.test.ts — separate test/ dir (home for AI-written tests + these prompts).
  • packages/*/test/**/*.test-d.ts — type-level tests (compiled with tsconfig.test.json).

No vitest config change is needed to adopt the separate test/ dir.

Conventions each prompt enforces

  • Vitest with globals: true. Biome for lint/format. Strict TS.
  • No emojis in source or tests.
  • Prefer real behavior over mocks (e.g. run the real emitters and assert on emitted strings). Mock only at external boundaries — fetch, fs, timers.
  • No snapshot tests without a meaningful assertion alongside them.
  • Do not duplicate scenarios already covered by colocated tests — the prompts list existing coverage so the model can fill gaps rather than re-write.
  • No source refactors, no API changes, no new runtime dependencies, no tests that hit real FHIR servers. Devtime-only deps (e.g. msw) are fine if the prompt allows them.

Gates (what a run must satisfy)

Gate Command Purpose
Unit + type tests pnpm test All vitest files green, including *.test-d.ts
Lint / format pnpm lint Biome clean
Typecheck pnpm -r typecheck Every workspace package typechecks in isolation

If a model produces changes that break any of these, the work is not done.

Out of scope for these prompts

  • No production code changes. If a test surfaces a real bug, file it separately — the prompt's job is coverage, not refactoring.
  • No new CI config, no new scripts. The three gates above exist already.
  • No hitting real FHIR servers. Fixtures and mocked fetch only.