fix(packages): await metadataPayloadBuilder.build() in CreatePackageUseCase

[ifoche]

Summary

• Fix missing await on metadataPayloadBuilder.build() in CreatePackageUseCase, which caused a raw Promise to be stored as package contents instead of the resolved metadata payload — making all module-generated packages appear empty
• Remove unnecessary generic type parameters from TransformationRepository.mapPackageTo/mapPackageFrom to prevent Promise objects from being silently accepted as payload (compile-time safety)
• Add unit tests for CreatePackageUseCase (5 test cases)

Related Tasks

• Module package imports successfully but contains no data

Test plan

• Run yarn test — all 393 tests pass (48 files)
• Run yarn build core-app — builds cleanly
• Manual: create a module with a dataset, generate a package, verify the package contains the dataset metadata
• Manual: import the generated package on another instance, verify elements appear in the result summary
• Manual: download package as JSON, verify metadata objects are present
• Manual: compare package with local instance, verify diff shows actual differences

🤖 Generated with Claude Code

[bundlemon]

BundleMon

No change in files bundle size

Groups updated (1)

Status	Path	Size	Limits
✅	Build Folder ./**/*	2.67MB (+43B 0%)	+20%

Final result: ✅

View report in BundleMon website ➡️

---

Current branch size history | Target branch size history

[xurxodev]

thanks @ifoche

This PR fixes a real bug: a missing await on metadataPayloadBuilder.build() in CreatePackageUseCase was storing an unresolved Promise as package contents, makinage empty. The fix itself is correct andminimal. The branch originally also narrowed TransformationRepository.mapPackageTo/mapPackageFrom from generic to a fiMetadataPackage type as an extra safety net a — that part has since been reverted on this branch back to development's fully generic signature, because it broke other legitimate callers once merged with development (see below). What's left to fix ile.

1. Should fix - src/domain/packages/usecases/tests/Creat doubles don't follow the project'sestablished pattern. The rest of the codebase builds test doubles with ts-mockito (mock/when/instance) plus givenA... helper functions that keep construction out osrc/domain/metadata/builders/tests/MetadataPayloadBuilder.spec.tsx, e.g. givenMetadataPayloadBuilderOfCategory). Tnew spec instead hand-rolls every dependency lt inline in beforeEach, and reimplementsStorageClient/DynamicRepositoryFactory collaborators by hand instead of using mock(...)/instance(...). Worth aligning with the existing convention for consistency ng out of sync with how the rest of the suitemocks the same collaborators (DynamicRepositoryFactory, UserRepository, TransformationRepository).

2. Recommendations non blocking

• Worth a short follow-up task to revisit TransformationRepository.mapPackageTo/mapPackageFrom's type safety. The original commit on this branch narrowed the sackage (Input/Output removed) specifically tomake an unresolved Promise a compile-time error — a reasonable goal given this bug's root cause. After merging development into this branch, that narrowing t legitimately pass non-MetadataPackage shapes: DownloadPayloadUseCase/DownloadPayloadFromSyncRuleUseCase (pass SynchronizationPayload, a wider union) and TEID2ApiRepository (passes TrackerPostRequest a bug — by teiTransformations-api.spec.ts,which exercises teiTransformations against DHIS2 wire-format data end-to-end). Reverting to the fully generic signature was the safe choice for this PR's scope, but -leak hole is back open at the type level forany future caller. A follow-up should explore a narrower fix — e.g. a whitelist-style conditional type covering the known valid payload shapes (SynchronizationPayload than touching this again inside an unrelated PR.

[xurxodev]

Hi @ifoche, pushed a commit addressing the "Should fix" point from the review:

• Tests (7edbfe62): rewrote CreatePackageUseCase.spec.ts to use ts-mockito (mock/when/instance) with givenA... helpers, matching the convention used elsewhere (e.g. MetadataPayloadBuilder.spec.tsx), instead of hand-rolled vi.fn() doubles and a real DynamicRepositoryFactory instance.

  While at it, also simplified the assertions: the original tests verified that internal collaborators were called (metadataPayloadBuilder.build(), transformationRepository.mapPackageTo()) rather than the use case's actual observable output. Those collaborator-call checks were redundant — if build() weren't called, or mapPackageTo() were skipped or wrong, the persisted contents would already fail to match, since it flows directly into what's saved. Went with 3 tests focused purely on what execute() persists (package contents resolved instead of a Promise, provided contents used as-is, module version updated), which is both less coupled to implementation and a closer fit to the bug this PR actually fixes.

All 441 tests pass.

The non-blocking recommendation about TransformationRepository.mapPackageTo/mapPackageFrom type safety (reverted to fully generic after merging development) is left as a follow-up, as originally suggested — it needs a proper design (e.g. a whitelist-style conditional type) to avoid breaking DownloadPayloadUseCase/DownloadPayloadFromSyncRuleUseCase/TEID2ApiRepository again.