Prepare SaveMoney for future updates

[mamu0]

Impact-free refactoring to prepare SaveMoney for future implementations:

• Unified Finding model + adapter
• Interface wrapping existing analyzers
• Limiter
• In-memory cache
• Rewritten orchestrator

Resolves: CES-2077

[Copilot]

Pull request overview

This PR refactors the @pagopa/dx-savemoney package to prepare for future extensions by introducing a unified “Finding” domain model, a plugin-style analyzer registry, and new runtime helpers (limiter + in-run metrics caching), while keeping the current report output shape (AzureDetailedResourceReport) unchanged.

Changes:

• Introduces a unified Finding model and an adapter to bridge legacy AnalysisResult into Finding[].
• Adds a plugin architecture for Azure analyzers (registry + types) and rewrites the orchestrator to use it.
• Adds a small concurrency limiter and an in-memory cache for Azure Monitor metric calls (scoped to a single run).

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
packages/savemoney/src/index.ts	Re-exports new Phase 0 public APIs (Finding model, analyzer plugin layer, limiter).
packages/savemoney/src/finding.ts	Adds unified Finding types and an adapter from legacy AnalysisResult.reason.
packages/savemoney/src/concurrency.ts	Introduces a lightweight async concurrency limiter helper.
packages/savemoney/src/azure/utils.ts	Adds run-scoped in-memory caching for getMetric() plus reset/test hooks.
packages/savemoney/src/azure/types.ts	Extends AzureConfig with an optional concurrency setting.
packages/savemoney/src/azure/analyzers/types.ts	Defines analyzer plugin interfaces (Analyzer, AnalyzerContext, AzureClients).
packages/savemoney/src/azure/analyzers/registry.ts	Wraps existing per-resource analyzers into the new registry.
packages/savemoney/src/azure/analyzers/index.ts	Provides re-exports for the analyzer plugin layer.
packages/savemoney/src/azure/analyzer.ts	Rewrites orchestration to use analyzer registry + limiter; resets metrics cache at run start.
.nx/version-plans/version-plan-1779281691024.md	Declares a patch version bump for the refactor.

Comments suppressed due to low confidence (1)

packages/savemoney/src/concurrency.ts:35

• createLimiter can temporarily allow more than max tasks to run. When a task completes, next() decrements active and resolves the next waiter; before the resumed waiter increments active, a new caller can observe active < max and start too, causing active to exceed max and breaking the concurrency guarantee. Consider a pool implementation that starts queued tasks without releasing the slot (or uses a while (active >= max) await ... loop plus slot reservation) so there is no window where capacity is double-claimed.

  return async <T>(fn: () => Promise<T>): Promise<T> => {
    if (active >= max) {
      await new Promise<void>((resolve) => queue.push(resolve));
    }
    active++;

[Copilot]

Pull request overview

Copilot reviewed 15 out of 15 changed files in this pull request and generated 6 comments.

Comments suppressed due to low confidence (1)

packages/savemoney/src/azure/tests/utils.test.ts:167

• Same issue here: as unknown as Parameters<typeof getMetric>[0] hides type mismatches in the failing-client mock. Using a small MonitorClientLike interface (or a Pick<> of the real client) would let this mock type-check without assertions and prevent accidental drift from the production call shape.

function makeFailingMonitorClient(calls: number[] = []) {
  return {
    metrics: {
      list: vi.fn().mockImplementation(async () => {
        calls.push(1);
        throw new Error("network error");
      }),
    },
  } as unknown as Parameters<typeof getMetric>[0];
}

[Copilot]

Pull request overview

Copilot reviewed 23 out of 24 changed files in this pull request and generated 2 comments.

Files not reviewed (1)

• pnpm-lock.yaml: Language not supported