Introduce MicrotaskTimer as a TimeoutTimer-style helper that schedules a
single runner via queueMicrotask, with set and cancel but no cancelAndSet.
Use it in DecorationLineCache to batch insert/delete line-index updates after
marker line adjustments, replacing ad-hoc queueMicrotask scheduling.
Co-authored-by: Cursor <cursoragent@cursor.com>
Extract the per-line lookup index from DecorationService into an exported
DecorationLineCache class in the same module, with DecorationService listed
first for readability.
Optimize buffer insert/delete maintenance to shift map keys in O(unique
indexed lines) instead of full re-indexing every affected decoration. Only
multi-line decorations that span the mutation point get a full re-index.
Add a direct DecorationLineCache unit test in DecorationService.test.ts.
Co-authored-by: Cursor <cursoragent@cursor.com>
Refs #5176.
forEachDecorationAtCell and getDecorationsAtCell previously scanned every
registered decoration on each call. With thousands of search highlights this
becomes O(total decorations) per cell and blocks rendering during viewport
refreshes.
Maintain a Map of buffer line -> decorations that cover that line. Multi-line
decorations are indexed on every line in their span, so mixed single-line and
multi-line workloads stay fast without falling back to a full scan.
Keep the index aligned with buffer mutations:
- Register and unregister decorations in the line buckets on add/dispose.
- On trim, shift line keys synchronously (before marker handlers run).
- On insert/delete, re-index affected decorations in a microtask after marker
line updates are applied.
Inject IBufferService so line listeners follow the active buffer. Add unit
tests for trim, dispose-on-trim, insert, and dense single-line plus multi-line
lookup. Add DecorationService.benchmark.ts to measure sparse hits and
viewport-sized grid scans with 20k decorations.
Co-authored-by: Cursor <cursoragent@cursor.com>
Per kovidgoyal/kitty#9999, enabling "Report event types" without
"Disambiguate escape codes" doesn't really make sense, and terminals may
choose whatever behavior they think makes sense. However, keys shouldn't
get swallowed (which, after #5791, happens when the !useCsiU branch
cannot handle the event), so instead make "Report event types" imply
"Disambiguate escape codes".
According to spec:
- The Enter, Tab and Backspace keys will not have release events unless
"Report all keys as escape codes" is also set.
- Key events that result in text are reported as plain UTF-8 text, so
events are not supported for them, unless the application requests
"key report" mode.
I checked kitty's behavior. While it does report release events for keys
that result in text as CSI sequences, press and repeat events for these
are reported as plain UTF-8 text. This commit changes Xterm.js to match
this behavior.
Add a parallel array container for execute handlers (codes < 0x18)
alongside the existing object. Setters populate both, the EXE
fast-path reads from the array with nullish coalescing to avoid
branching. ExecuteHandlerType updated to accept optional ident param
to allow the ?? pattern with the fallback handler.
Replaces 10 separate reset()+addParam(0) pairs across the parser with
a single resetZdm() method that inlines both operations, avoiding two
calls to a foreign object per sequence reset.
Remove intermediate byte handling from the fast-path, bail to table
lookup instead. Remove redundant per-step length checks by requiring
i + 2 < length upfront and letting the for-loop guard the rest.
Replace three !== checks (OSC_STRING, DCS_PASSTHROUGH, APC_STRING)
with one < OSC_STRING comparison. States below OSC_STRING are all
safe for the CSI fast-path; states above need the normal table path
for proper subparser abort propagation.
Address reviewer feedback:
- CSI fast-path is now a condition at the top of the codepoint loop
instead of inside CLEAR, avoiding table lookups for the common case
- EXE bytes < 0x18 get a fast-path before the CSI check so control
bytes don't pay the penalty
- PRINT loop unrolling replaced with while-based version that uses
trueish eval and spares individual length checks in the unrolled case
- Removed misplaced PRINT scan-ahead from CSI dispatch
Daniel Imms
zeertzjq
Jörg Breitbart
Ismail Pelaseyed
Sridhar Ratnakumar