feat(storage): add raw binary-blob primitive to every storage adapter

Introduce a first-class binary-blob storage primitive on the StorageAdapter
contract and implement it across all storage backends. This stores opaque byte
payloads verbatim instead of base64-in-JSON, eliminating the ~33% inflation and
full-materialization cost of the JSON envelope. It unblocks zero-copy,
mmap-able column-store segments and batch vector I/O at billion scale.

New methods (declared abstract on BaseStorageAdapter, the class that implements
StorageAdapter, and added to the StorageAdapter interface):

  saveBinaryBlob(key, data)    raw write, atomic on real filesystems
  loadBinaryBlob(key)          exact bytes, or null if absent
  deleteBinaryBlob(key)        idempotent (missing is ignored)
  getBinaryBlobPath(key)       real local fs path where one exists, else null

Shared key -> location convention across every adapter: the key's
"/"-separated segments nest under a `_blobs/` prefix and are suffixed with
`.bin`, e.g. "graph-lsm/source/sstable-123" ->
"<root>/_blobs/graph-lsm/source/sstable-123.bin". Blobs are not branch-scoped
(COW): they are immutable producer-managed segments.

Per-adapter behavior:
- FileSystemStorage: writes under <rootDir>/_blobs via tmp+rename; returns the
  real on-disk path so native code can mmap it directly. Path convention matches
  the existing MmapFileSystemStorage subclass byte-for-byte.
- S3CompatibleStorage / R2Storage / GcsStorage / AzureBlobStorage: put/get/delete
  raw octet-stream objects; getBinaryBlobPath returns null (remote stores have no
  local path).
- MemoryStorage: defensive-copied Map<string, Buffer>; null path; cleared on
  clear().
- OPFSStorage: stores raw bytes in the OPFS tree; null path.
- HistoricalStorageAdapter: read-only — save/delete throw; load resolves the
  blob from the historical commit tree; null path.

Tests: tests/unit/storage/binaryBlob.test.ts exercises save/load round-trip
(byte-identical, incl. non-UTF8 bytes), overwrite, delete-then-load, load-missing,
and getBinaryBlobPath behavior for all eight adapters. Cloud adapters run against
in-memory client fakes that drive the real adapter code; OPFS runs against an
in-memory FileSystem Access API mock; the historical adapter commits a blob into
a real COW tree. 59 new tests; full unit suite (1398 tests) green.
This commit is contained in:
David Snelling 2026-05-27 11:49:49 -07:00
parent 547721ae14
commit 298b572671
12 changed files with 1552 additions and 0 deletions

View file

@ -89,6 +89,10 @@ export class FileSystemStorage extends BaseStorage {
private indexDir!: string // Legacy - for backward compatibility
private systemDir!: string
private lockDir!: string
// Root for raw binary blobs (`<rootDir>/_blobs`). Blobs are stored verbatim
// (no JSON envelope, no compression) so native code can mmap them directly via
// getBinaryBlobPath(). Set in init() once the path module is loaded.
private blobsDir!: string
private activeLocks: Set<string> = new Set()
private lockTimers: Map<string, NodeJS.Timeout> = new Map() // Track timers for cleanup
private allTimers: Set<NodeJS.Timeout> = new Set() // Track all timers for cleanup
@ -213,6 +217,7 @@ export class FileSystemStorage extends BaseStorage {
this.indexDir = path.join(this.rootDir, 'indexes')
this.systemDir = path.join(this.rootDir, SYSTEM_DIR)
this.lockDir = path.join(this.rootDir, 'locks')
this.blobsDir = path.join(this.rootDir, '_blobs')
// Create the root directory if it doesn't exist
await this.ensureDirectoryExists(this.rootDir)
@ -242,6 +247,9 @@ export class FileSystemStorage extends BaseStorage {
// Create the locks directory if it doesn't exist
await this.ensureDirectoryExists(this.lockDir)
// Create the binary blobs directory if it doesn't exist
await this.ensureDirectoryExists(this.blobsDir)
// Initialize count management
this.countsFilePath = path.join(this.systemDir, 'counts.json')
await this.initializeCounts()
@ -946,6 +954,91 @@ export class FileSystemStorage extends BaseStorage {
}
}
// ===========================================================================
// Raw binary-blob primitive (mmap-friendly)
// ===========================================================================
/**
* Resolve a blob key to its on-disk path under `<rootDir>/_blobs`.
*
* The key's "/"-separated segments become nested directories and the file is
* suffixed with `.bin`, e.g. `"graph-lsm/source/sstable-123"`
* `<rootDir>/_blobs/graph-lsm/source/sstable-123.bin`. This convention is
* shared with cortex's `MmapFileSystemStorage` so native code and the TS layer
* agree on exactly where each blob lives.
*
* @param key - The blob key.
* @returns The absolute on-disk path for the blob.
* @private
*/
private blobPath(key: string): string {
// `path` and `blobsDir` are populated in init(). If a caller resolves a blob
// path before init() (e.g. native code probing for a mmap target), fall back
// to a POSIX-style join off rootDir so this synchronous method never throws.
if (path && this.blobsDir) {
return path.join(this.blobsDir, ...key.split('/')) + '.bin'
}
return `${this.rootDir}/_blobs/${key}.bin`
}
/**
* Persist a raw binary blob verbatim under `key`, using an atomic
* temp-file + rename so concurrent readers never observe a torn write.
* Parent directories are created on demand.
*
* @param key - The blob key (see {@link getBinaryBlobPath} for the convention).
* @param data - The exact bytes to store.
*/
public async saveBinaryBlob(key: string, data: Buffer): Promise<void> {
await this.ensureInitialized()
const filePath = this.blobPath(key)
await this.ensureDirectoryExists(path.dirname(filePath))
const tmpPath = filePath + '.tmp'
await fs.promises.writeFile(tmpPath, data)
await fs.promises.rename(tmpPath, filePath)
}
/**
* Load the raw bytes stored under `key`, or `null` if the blob does not exist.
*
* @param key - The blob key.
* @returns The blob bytes, or `null` if absent.
*/
public async loadBinaryBlob(key: string): Promise<Buffer | null> {
await this.ensureInitialized()
try {
return await fs.promises.readFile(this.blobPath(key))
} catch {
return null
}
}
/**
* Delete the blob stored under `key`. Missing blobs are ignored.
*
* @param key - The blob key.
*/
public async deleteBinaryBlob(key: string): Promise<void> {
await this.ensureInitialized()
try {
await fs.promises.unlink(this.blobPath(key))
} catch {
/* ignore missing files */
}
}
/**
* Return the real on-disk path for `key` so native code can mmap the file
* directly. The path is returned whether or not the file currently exists
* callers are expected to write before mapping.
*
* @param key - The blob key.
* @returns The absolute on-disk path (never `null` for filesystem storage).
*/
public getBinaryBlobPath(key: string): string | null {
return this.blobPath(key)
}
/**
* Get multiple metadata objects in batches (CRITICAL: Prevents socket exhaustion)
* FileSystem implementation uses controlled concurrency to prevent too many file reads