hyl.st / reams
Automerge backed Markdown editor for the Atmosphere
Star
Fork 0
Watch
reams docs testing coverage-gate.md
5.7 kB
Markdown
at main

Coverage gate #

The pre-commit hook runs npm run test:coverage and requires 100% line, statement, function, and branch coverage per file. The committed scope is the portable core boundary in src/utils/**/*.ts, src/types/**/*.ts, the metadata services boundary (heading-slug, extract-metadata, metadata-cache, metadata-cache-persistence), the service layer (derived-index-engine, derived-index-persistence, document-manager, file-index, background-indexer, search-service, search-index-persistence, file-stat-from-history, fuzzy-score, image-resolver, tab-awareness, url-router, vault-root-url), and the sync layer (heads, sync-server-config, sync-connection-status), and the format-preservation utilities (hash, lezer, value-blob, synthesis-cache, registry, core-registry, reconcile-context, mark-emitter), and the automerge-markdown core (mutation-context, paste-bounded, errors, markdown-analysis, claim-set, phase0, markdown-block-analysis, rich-text-projection, native-block-operations); type declarations are not executable and are excluded. The layout layer covers gesture-command-utils, gesture-settings, layout-state, layout-persistence, sidebar-leaf-persistence, search-create, create-item-path, file-tree-utils, and file-tree-persistence; type declarations are not The metadata tests cover Markdown normalization, duplicate disambiguation, fragment lookup, frontmatter parsing, code-range detection, section building, link/embed/tag/heading/footnote extraction, backlink resolution, display titles, resolved links, and IndexedDB persistence. The service layer tests cover CRDT document management, derived index engine work generation and snapshot coherence, IndexedDB persistence, file index live subscription reconciliation and timestamp wiring, background indexer scheduling, retry/backoff, profiling, and persistence migration, MiniSearch full-text search with filter predicates and defensive fallbacks for inconsistent state, search index IndexedDB persistence with upgrade callbacks and malformed-data validation, FileStat derivation from Automerge change history with clock-skew defense and binary content support, fuzzy string scoring with tier ordering and subsequence caps, image blob URL resolution with cache eviction and concurrent deduplication, cross-tab awareness via BroadcastChannel announce/ack/close protocol, URL hash fragment routing with foreign-doc adoption and replaceState dedup, and vault root URL first-boot persistence. The sync layer tests cover unordered Automerge heads comparison and canonical key generation, sync-server URL resolution with localStorage/env fallback and ws/wss scheme validation, and live WebSocket connection-status registry with adapter hot-swap and pre-init URL suppression. The format-preservation tests cover FNV-1a hashing, mark emission, Lezer markdown parsing, mark value blob encoding/decoding with owner and version validation, LRU-bounded synthesis cache with heads-keyed eviction, construct registry with owner validation and reserved-name enforcement, core semantic registry singleton caching, reconcile context with claim bounds and parsed-tree resolution, and low-level mark-at-path emission with rich-text projection. The legacy reconcile adapter, content-path mark emission wrapper, draft coordinator, adapter schema vocabulary, and content-path block normalization were relocated to tests/format-preservation/harness/legacy; the characterization corpus that exercises them still runs, but those modules are no longer part of the production coverage gate. The automerge-markdown tests cover mutation context tracking with guarded mutations, paste range validation, draft reconciliation error classification, markdown analysis with deep-freeze cloning, claim set interval merging with kind promotion and claim validation, and phase 0 cross-construct cleanup with shielded range trimming and mark expansion preservation, and markdown block analysis with prefix parsing, list/quote depth tracking, code fence shielding, and continuation parent assignment, and rich-text projection with path validation, coordinate guarding, mark expansion probing, block descriptor decoding, mark peer splitting and merge, and visible text extraction, and native block operations with compatibility checking, snapshot reconciliation, block stream updates, and mutation context guarding. The layout layer tests cover gesture command execution with user-facing notices, gesture settings localStorage persistence with version migration and type-validated field fallbacks, layout state sidebar/right-panel/editor mode transitions with toolbar offset computation, layout persistence with v1-v5 schema migration and clamped numeric fields, sidebar leaf persistence with v1-to-v2 migration and old panel-views incorporation, search-create preview resolution and execute-time re-resolution with template-backed and blank-file creation paths, file-tree create-item path parsing with URL adoption and validation, file-tree utility icon mapping and deprecated sort compatibility, and file-tree expanded folder persistence with corrupted-data defense.

Performance smoke tests, browser invariants, and integration tests remain separate executable lanes. They are excluded from instrumentation because V8 instrumentation changes timing and browser-only behavior. Run npm run test, npm run test:perf, and npm run test:invariants for those lanes.

Reports are emitted to coverage/ in text, JSON, JSON summary, and HTML formats. The scope should be expanded only when its functional tests can meet the same four 100% thresholds without weakening the performance or browser budgets.