Before: Used recursive setTimeout calls via unpause, creating nested closures and multiple timer contexts.
After:
- Tracks a single timeoutId to manage timer lifecycle
- Eliminates nested closure creation
- Cleaner timer chain when calls are pending
2. batch (line 1204-1227)
Before: Used throttle internally, inheriting its timer overhead.
After:
- Direct timer management without the throttle wrapper
- Single timeoutId tracked explicitly
- Maintains the same semantics (first item processed immediately, rest batched)
3. batcher (line 1235-1259)
Before:
- Created new timer every time queue went from empty to non-empty
- Had unnecessary async in forEach and await r when r was already type U
After:
- Tracks timeoutId to prevent duplicate timer creation
- Only creates timer when queue is truly empty
- Removed unnecessary async/await in result handling
- More efficient error handling (rejects all items at once if length mismatch)
Performance Benefits:
These changes should significantly reduce timer overhead by:
- Fewer timer objects: Only one active timer per throttled/batched function instead of chains
- Less garbage collection: Fewer closure allocations and intermediate objects
- Better memory usage: Explicit timer ID tracking instead of implicit state in closures
Jon Staab
de5695339d