HighlighterEngine::GetLine copied the span vector on every call (twice
on a miss) and did a dynamic_cast per call; it now returns the cached
entry (valid until the next call) and dispatches on Stateful().
PieceTable and HighlighterEngine took a mutex in their read paths while
their writes took none, and PieceTable.h claimed concurrent reads were
safe; they are not (reads update caches). Every Buffer is used from the
main thread only (the swap writer thread receives copies; the regex
large-stack thread runs while its caller waits), so the mutexes are
removed and the headers say so. ASan and the threaded tests under TSan
are clean.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015rcVQiiWyTzoNrsvLSaAfB
- HighlighterEngine caches spans only for rows near the requested row
and trims the cache around it past 16k entries; per-row end states are
still kept to resume lexing. Jumping to the end of a 78 MB C++ file
cached every row's spans (854 MB RSS, now 197 MB), and the first edit
near the top spent 270 ms freeing them.
- Plain and regex search read lines as views into the materialized
buffer instead of Buffer::Rows(), which built (and kept) a string per
line, O(lines x pieces) after edits. Plain search scans the whole text
once and tracks rows by newline.
- transform_rows() applies a per-row rewrite of a row range as one edit
(replace_rows_text). Indent/unindent region and visual-line insert,
backspace, delete, newline and yank used one piece-table edit per row,
each linear in the buffer: 1.1 s per keystroke with 5,000 rows selected
in a 26 MB file (now 4 ms), and minutes to indent a large region.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015rcVQiiWyTzoNrsvLSaAfB
Every edit bumped the buffer version (through SetDirty) and called
InvalidateFrom(0), so all cached spans and line states were discarded
and stateful highlighters re-lexed from row 0 to the viewport on each
keystroke.
Buffer mutators now report the row they changed (edited_at_), which
bumps the version and tells the engine to drop only rows from there
down; a row's highlighting depends only on rows above it. A version
change that arrives without that notice still clears everything, so a
missed path cannot show stale spans. The engine keeps states as a
contiguous vector and caches spans for every row it lexes. SetDirty no
longer touches the version: the edits themselves do, including undo
and redo back to the saved state (D9 still holds; its test now checks
the highlighting itself).
Typing near the end of a 100k-line C++ file: 46.3 ms -> 0.14 ms of
highlighting work per keystroke.
Audit finding P3.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015rcVQiiWyTzoNrsvLSaAfB
The worker read the Buffer (PieceTable pieces and add buffer) while the
main thread edited it with no shared lock, and kept a raw Buffer* that
dangled once the buffer vector reallocated or shifted. PrefetchViewport
now only highlights the visible rows synchronously.
Audit finding M2.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015rcVQiiWyTzoNrsvLSaAfB
Claude
kyle