When a cancel removes journal entries numbered below a recording's
start, the entries after them are numbered down. The recording's start
and the journal's hold kept the old number, so Stop Recording left out
as many of the recorded steps as there were entries removed, and the
hold no longer covered them. `Journal::remove` now moves the hold down
with the entries, and Stop Recording reads the start from the hold.
`Journal::replace` puts an entry in place of another and counts it at
its new size. A cancel uses it for the `document.activate` that takes
the place of a canceled group's entry that switched documents, so the
journal's size estimate does not keep the old entry's size.
A cancel cut the journal back to where the first canceled group began.
That also removed the commands run in other documents after it began,
which the cancel does not undo, so a replay of the journal lost them.
Each undo group now keeps the numbers of its own journal entries: the
commands that start and end in its document and the commands that edit
it. A cancel removes only those entries, and the later entries are
numbered down to close the gaps (`Journal::remove`). An entry of a
canceled group that also switched to another document, such as a batch
that draws and then activates another document, is replaced by a
`document.activate` of that document, so the commands after it replay
in the same document.
History was limited by step count only; a whole-selection move of 50k paths keeps ~40 MB per
step. Steps now carry a size estimate (only unshared subtrees are visited) and the oldest are
dropped beyond a byte budget, keeping the last 8. Documents entering history drop their id
index, selections are shared between steps, the command journal is bounded (held while an
Action records), and document resources are copy-on-write. Saved files are byte-identical.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VmMmc1J8XeNLxhRihWv9cc
dotCooCoo