Skip to content

[WIP][#436] Integrate history endpoints - #492

Draft
Maurier wants to merge 13 commits into
mainfrom
436-integrate-history-endpoints
Draft

[WIP][#436] Integrate history endpoints#492
Maurier wants to merge 13 commits into
mainfrom
436-integrate-history-endpoints

Conversation

@Maurier

@Maurier Maurier commented Aug 3, 2026

Copy link
Copy Markdown
Contributor

Integrates history and session management end to end.

Maurier added 9 commits June 22, 2026 12:53
Start editing opens a chooser of the source's managed datastreams and their
sessions: continue an in-progress session, start a new one over the selected
time range, or create and delete managed datastreams. Managed datastreams are
hidden from the catalog and reached through the chooser. Also adds inline
processing-level creation, the session description captured at commit time,
reuse of the source's loaded series as the working copy, clamping the session
window to the source extent, and session resume that replays saved operations.
@Maurier
Maurier marked this pull request as draft August 3, 2026 22:29
Maurier added 4 commits August 5, 2026 15:50
Plot a QC session's state at any operation as a separate, frozen line so
it can be compared against the session being edited. Buttons sit on every
operation row and on the "Data loaded" row, which plots the state the
session started from.

Snapshots ride in plottedDatastreams under the synthetic id
snap:<sessionId>:<opIndex>, so legend rendering, colour assignment,
visibility and reorder work unchanged. isSnapshotId guards the paths that
would otherwise treat one as real: refreshGraphSeriesArray skips its
fetch, releaseManagedDatastream drops them when the editor closes, and
the share encoder keeps them out of ds in favour of its own snap key.

The base is always the session chain's own window, never the plot's time
range, because operations replay against array indices. Snapshots build
on a detached ObservationRecord since the observation store hands back
one shared instance per datastream and the replay mutates what it is
given.

Also fixes two pre-existing problems in the history panel that this work
surfaced:

- Reloading from a step restored the pre-replay entries wholesale, which
  put the old timings back over the ones the replay had just measured.
  Only the "Data loaded" row updated. The panel now keeps the replayed
  entries and appends only what the replay dropped.
- hasUnsavedChanges compared entries by identity, so a replay while
  viewing a committed session read as pending edits and prompted "You
  have unsaved edits". A committed session cannot be edited, so both it
  and unsavedEditCount now short-circuit on isReadOnly.
Covers the feature in the user guide, the id convention, guards and
detached-fetcher rule in the architecture doc, the axis behaviour in the
plotting doc, and the two new store actions in the API reference.
The Data loaded row is step -1, so a session's starting state is now
reachable the same way any operation is, and it carries the "Showing"
chip when it is the step on screen.

Its reload-from-server button moves to a cloud icon so the row no longer
shows two identical glyphs doing different things: one replays the
history to a step, the other refetches and drops the history.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant