@loradb/lora-node 0.30.0 → 0.31.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +17 -3
  2. package/package.json +8 -8
package/README.md CHANGED
@@ -236,6 +236,11 @@ Resuming:
236
236
  covers, so history starts at the last `sync()`.
237
237
  - An LSN the database no longer retains fails with
238
238
  `LORA_CHANGES_TRUNCATED`: start a new feed and re-read current state.
239
+ - On a WAL-backed database a batch is delivered once its commit is durable
240
+ (fsynced), so a crash can never take back an LSN a consumer has seen. The
241
+ feed asks for the fsync at once; under a steady stream of writes batches
242
+ trail the commit by up to the WAL sync interval (50 ms by default, 1 s
243
+ for a named database).
239
244
 
240
245
  Writers never wait for a feed. Each feed buffers up to `bufferSize` batches
241
246
  (default 1024); a consumer that falls further behind receives the buffered
@@ -509,9 +514,18 @@ See `ts/types.ts` (`LoraErrorCode`) for the full list.
509
514
  deadlocking: the writer holding the lock always runs). Prefer
510
515
  `await`-in-a-loop or a single batched query for heavy write workloads.
511
516
  While an interactive transaction or a mutating stream is open (it holds
512
- the lock until JavaScript finishes it), auto-commit writes, `clear()`,
513
- `sync()` and `loadSnapshot()` wait on lora's own threads instead, so the
514
- pool stays free for `fs`, `dns`, `zlib` and `crypto`.
517
+ the lock until JavaScript finishes it), writes wait off the pool instead,
518
+ so it stays free for `fs`, `dns`, `zlib` and `crypto`. Auto-commit
519
+ `execute()` and `profile()` writes then wait in a first-in, first-out
520
+ queue per database handle, run by one thread: 2 000 writes behind a long
521
+ transaction take one thread, not 2 000. A queued write still rejects with
522
+ `LORA_TIMEOUT` when its `timeoutMs` (or `queryTimeoutMs`) passes or its
523
+ `signal` aborts, and the time it spends queued counts against the
524
+ timeout. `dispose()` rejects the writes still queued. `clear()`,
525
+ `sync()`, `loadSnapshot()` and reads never wait behind that queue:
526
+ `sync()` on an in-memory or WAL database finishes while writes are
527
+ queued (on a database opened by name it takes the writer lock, so it
528
+ waits for the transaction like a write).
515
529
  Interactive transactions wait on their own threads, not the pool, and so
516
530
  do mutating `stream()`s while the lock may be busy: one opens without
517
531
  blocking the event loop, and its first `next()` waits for the writer
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@loradb/lora-node",
3
- "version": "0.30.0",
3
+ "version": "0.31.0",
4
4
  "description": "Node.js / TypeScript bindings for the Lora in-memory graph database",
5
5
  "license": "BUSL-1.1",
6
6
  "packageManager": "yarn@4.5.1",
@@ -96,12 +96,12 @@
96
96
  "vitest": "^2.1.1"
97
97
  },
98
98
  "optionalDependencies": {
99
- "@loradb/lora-node-linux-x64-gnu": "0.30.0",
100
- "@loradb/lora-node-linux-arm64-gnu": "0.30.0",
101
- "@loradb/lora-node-linux-x64-musl": "0.30.0",
102
- "@loradb/lora-node-linux-arm64-musl": "0.30.0",
103
- "@loradb/lora-node-darwin-x64": "0.30.0",
104
- "@loradb/lora-node-darwin-arm64": "0.30.0",
105
- "@loradb/lora-node-win32-x64-msvc": "0.30.0"
99
+ "@loradb/lora-node-linux-x64-gnu": "0.31.0",
100
+ "@loradb/lora-node-linux-arm64-gnu": "0.31.0",
101
+ "@loradb/lora-node-linux-x64-musl": "0.31.0",
102
+ "@loradb/lora-node-linux-arm64-musl": "0.31.0",
103
+ "@loradb/lora-node-darwin-x64": "0.31.0",
104
+ "@loradb/lora-node-darwin-arm64": "0.31.0",
105
+ "@loradb/lora-node-win32-x64-msvc": "0.31.0"
106
106
  }
107
107
  }