performFullSync only ever cached each collection's root-level documents
— a document's own sub-documents were never cached at all, so nested
content stayed unreachable offline even with a full sync (and even
after implicitly re-fetching individual documents, since the walk that
seeds the cache never visited them). Now recurses depth-first into
every document's children via cacheDocumentTree, terminating naturally
once a branch runs out of sub-documents.
Also caches each collection individually under "collection:<id>" (was
only ever cached as part of a paginated list blob), and adds
OfflineCacheStore.loadAll(keyPrefix:) plus
cachedDocumentsIndex()/cachedCollectionsIndex() on
CachingOutlineAPIClient to read everything back as a flat local index
— no per-id lookup needed, no network involved.
Found and fixed a real infinite-recursion bug in the process: an
existing test's stub ignored parentDocumentId and kept returning the
same root documents at every recursion depth, hanging swift test
indefinitely once the real code started recursing into children. Fixed
the stub, added a dedicated regression test for the recursive behavior
itself. 78/78 tests passing.
Puranjay Savar Mattas
bef194d493