Skip to content

planner-ux: merge develop (the web product) - #3662

Merged
timohueser merged 1702 commits into
planner-uxfrom
claude/planner-ux-merge-develop-2
Nov 19, 2026
Merged

timohueser merged 1702 commits into
planner-uxfrom
claude/planner-ux-merge-develop-2

Conversation

@timohueser

@timohueser timohueser commented Nov 18, 2026 •

Copy link
Copy Markdown
Owner

Mode: skip: merge of develop into the epic branch of #3286
Requirements: none
Public docs: yes

Part of #3286. This pull request merges develop into planner-ux at f06072d91, in two merge commits: f9222000c takes develop at 7207097c6, and edce9c06f takes it at be962056e. Small commits follow each. It adds no feature.

Do not merge without the owner's word. The section "Decisions for the owner" lists the cases where the two branches differ on one element.

The 34 conflicts

File The epic wanted develop wanted Resolution
CompareList.svelte Delete it (no alternatives) The units of the rider, the icon of lib/ui Deleted
CompareProfile.svelte Delete it The units of the rider Deleted
lib/planner/route-options.ts Delete it The units of the rider Deleted
lib/planner/route-options.test.ts Delete it A test with miles Deleted
DayRow.svelte The title "Day 2 · Sun 20"; no place kind in the small line Heights in the units of the rider The title of the epic, the units of develop, no place kind
Itinerary.svelte untrack for the night page The icon of lib/ui Both imports
MapCallout.svelte The prop takeFocus No onRoutes: the Signed routes view is gone takeFocus stays, onRoutes goes
NightNode.svelte No children; the counts follow the stay kinds The icon of lib/ui; the distance in the units of the rider The form of the epic with the units of develop
Planner.svelte.test.ts The type Stay onTestFinished, the type Trip All imports
PlannerApp.svelte No alternatives; route notes; the night page; the trip row; the position on demand; no test route button The planner is a page of the shell: no site header, the library from the page, the router, no Signed routes view, the units, the map settings, the layer legend The file of the epic with each change of develop
PlannerApp.svelte.test.ts The helpers escape and stayBox startApp gives the library Both
PlannerMap.svelte No compare; highlightedLines, showNight, onRiderMove No Signed routes view, no onIdle; the route details in the popup, the route hint, aria-busy The file of develop without compare, with the three additions of the epic
PlannerProfile.svelte No day labels, no distance number, no trip row; the totals in the title row The units of the rider, gradePaths, Segmented and the icon of lib/ui The structure of the epic with the units and the imports of develop
PlannerQuery.svelte No region block The icon of lib/ui The icon of lib/ui; no config import
PlannerSurface.svelte No "Hover to inspect" A legend with the share of each surface The side of develop
QueryResults.svelte A distance column and day groups The units of the rider; no "Signed routes near" row The rows of the epic; the distances in the units of the rider; no routes row
StopCard.svelte One radio group; the header counts follow the stay kinds The units of the rider The card of the epic; stopKm and the day tiles follow the units
StopPicker.svelte No lead sentence; the stay chip row; lists by stay kind The units of the rider The form of the epic; the "No … between" text follows the units
TripBar.svelte The two-row breakpoint at 1210 px; a row gap of 4 px The controls wrap by their room; two rows under 1050 px; the pace select The side of develop (decision 1)
TripRow.svelte StayChip The icon of lib/ui Both imports
plan-session.svelte.ts No alternatives; autoName The library of the product, planSummary, follow The imports of develop without townName, presetName and optionOrder
plan-session.test.ts No alternatives tests; the autoName test Four alternatives tests, a rename test, a pace test The rename test, the pace test and the autoName test stay; the alternatives tests go
route-closures.test.ts closureLabel A miles test of closureNote The tests of the epic; the miles test checks noteDistance
route-difficulty.ts alpineLabel, no alpineNote Each stretch has its hardest grade, T4 to T6 alpineLabel(stretch) names the grade of the stretch (decision 2)
route-difficulty.test.ts The stretches only The note with the hardest grade One test: the stretches and their labels
route-gaps.ts gapLabel for one gap gapNote with noteDistance gapLabel with noteDistance
route-gaps.test.ts gapLabel, at, the loop start A miles test All of the epic; the miles lines check gapLabel
route-legs.ts requestRoute gives one route; no add The chunks of a long plan; a second try after busy and internal The code of develop with the call of the epic; add stays deleted
routing.ts requestRoute without alternatives MAX_REQUEST_POINTS; canOfferAlternatives MAX_REQUEST_POINTS stays; no alternatives
routing.test.ts No requestAlternatives pacedLine, chunks, new types The imports of develop without requestAlternatives
docs/content/software/trips.md The trip values at the top of the day list; a night and a town as pages The pace; what stays after a new guess; 32 days The three paragraphs of the epic; the second has the pace sentences of develop
host/obc-data-steps/src/planner.rs PLANNER_ROUTING at revision 4 PLANNER_ROUTING at revision 19 Revision 20
planner/router-build/README.md No sentence about alternatives hiking/alpine accepts each known grade up to T6 The sentence of develop without the alternatives sentence
specs/vectors/README.md route-answer.json has one route Rows for climbs.json and route-boxes.json; a new text for signed-routes.json The row of the epic for route-answer.json; the other rows of develop

Changed without a conflict

In commits of their own:

  • 323291e52: PlanTotals.svelte, RouteNotes.svelte and StayChip.svelte import the icon from lib/ui/Icon.svelte. develop moved the file; the epic's icons lock, ascent and descent moved with it.
  • 323291e52: PlanTotals.svelte shows the distance, the ascent and the descent with distanceIn, heightIn, heightUnit and settings.units. Its time comes from planView, which develop changed to the time at the pace of the route.
  • 323291e52: RouteNotes.svelte shows the place of a note in the units of the rider: "km 104.6 – 118.3" or "mi 65.0 – 73.5". The length uses noteDistance, as before.
  • 323291e52: stopKm in lib/planner/stops.ts gives "km 42–46" or "mi 26–29". The town rows and the town card use it.
  • 323291e52: the tests of PlanTotals and RouteNotes each have one check with miles.
  • 25b8f1526: routeNotes in lib/planner/route-notes.ts takes AlpineStretch and calls alpineLabel(stretch). Its test gives each stretch a grade.

In the merge commit, in conflicted files but outside the conflict hunks:

  • PlannerMap.svelte: the parts of compare that develop kept are gone: compareData, the compare source and its two layers, the letter markers and the styles .compare-key and .planner-map-letter. syncRouteVisibility only shows or hides the route layers.
  • QueryResults.svelte: the distance column of the epic uses distanceIn and settings.units. The rule "on route" under 50 m stays.
  • StopCard.svelte: the helper at of develop has no user and is gone.
  • TripBar.svelte: the comment above .ride names all three controls.
  • routing.test.ts: the test "offers route options only for a plan that fits one request" of develop is gone. It reads line.primary, which the epic deleted.
  • Planner.svelte.test.ts: one test line of develop set unpavedKm, which the epic deleted.
  • PlannerApp.svelte: the button "Find another route" of develop goes to the Explore page. It keeps the rule of the epic: it does not show on a night page.
  • docs/content/software/trips.md: three sentences of the merged paragraphs are shorter, so the page stays under its limit of 2,500 words. No fact is lost.

Merged without a change, and checked: lib/ui/Icon.svelte, editor.ts, search/presentation.ts, test-support/planner/MapStub.svelte, tests/browser/planner/stubs.js, the route build (source.rs, osm.rs, tests/routing.rs), planner/router and specs/route-api.md. A search for alternatives in the router, the route service, the route build, the route specs and companion-ios finds only PlannerServiceTests.swift, which checks that the request has no such field.

Decisions for the owner

  1. The top row of the planner (TripBar.svelte).
  2. The label of an alpine stretch in the route notes.
    • Option A: each stretch names its hardest grade: "Alpine path (T5)". A T4 stretch and a T5 stretch within 2 km are two rows, because only rows with the same text merge.
    • Option B: one text for all alpine stretches, so that they merge within 2 km, with or without the hardest grade of the merged row. This needs a change of the merge rule.
    • The branch has option A. The epic had the fixed text "Alpine path (T4)", because T4 was the hardest grade of the Alpine hiking preset. develop lets the preset accept T5 and T6.

Cases that the brief named, where both sides hold and no decision is necessary:

  • The surface strip. develop rebuilt the label row with a legend. The text "Hover to inspect" is gone there too.
  • The elevation drawer. develop changed only units and imports in PlannerProfile.svelte. The bottom card of the Route page is another component. The title row with the totals and the body height are those of the epic.
  • "Find another route". The button of develop stays and does not show on a night page.
  • The empty planner text and the callout of the map. develop did not change the text. The callout has takeFocus of the epic and no onRoutes.

Lost or superseded

Changes of the epic that are gone:

Changes of develop that the deletions of the epic removed:

  • The units of the rider in CompareList.svelte, CompareProfile.svelte and route-options.ts, with the test.
  • Alternatives code that develop changed: canOfferAlternatives with its test, firstOptions, LegCache.add, four tests of plan-session.test.ts and the requestAlternatives lines of routing.test.ts.
  • closureNote, gapNote and alpineNote. The labels of the epic replace them. The text "Route misses 2 selected points by up to …" is gone: each gap is one row. The rule for miles stays in noteDistance.
  • In PlannerProfile.svelte: the units of the day labels and of the distance number, and the trip row in the title row. The epic removed these elements.
  • In QueryResults.svelte: the detail text "… along route · … off route". The distance column of the epic shows the two numbers.
  • In StopPicker.svelte and StopCard.svelte: the lead sentence and the header with stop.from to stop.to. The epic counts only the places of the stay kinds of the night.
  • In DayRow.svelte: the place kind after the heights.
  • The handler loadSearchSample and the prop onSample. develop still had the test route button.
  • The sentence "Alternatives keep the difficulty limit of the request." of planner/router-build/README.md, and the two-route text of specs/vectors/README.md.
  • In trips.md: the sentences of develop about the row above the profile, the day labels, "Stay here" and "This night". The planner after the merge has none of these.

Layer revisions

  • PLANNER_ROUTING_SOURCE is at revision 2 (66022373c). develop has 1. The source step makes the closure sets (doubt in osm::source), and the key of a step holds no code. So a store that develop built would keep its source layer and the closures of destination.
  • With the new revision the source step runs again. The roads step reads the source layer, the costs step reads the source and the roads layers, and the package step reads all three. A key holds the digest of each input, so each of these steps runs again where the files of the step before it change. In a region with no destination value the source files are the same, and the later steps are reused. That is correct.
  • PLANNER_ROUTING stays at revision 20. develop has 19 and planner-ux has 4. The package step calls doubt itself for the access overlay, so its output changes also by its own code. The digests of its inputs cover each case that I found, but 20 costs at most one run of the package step, and each side raised this revision.
  • PLANNER_SEARCH stays at revision 3. develop has 2 and did not change planner/search or host/obc-search-bake. The search build has no cached work layer: only the three routing work layers have cache: true. Each search step (model, policy, dump, records, pois, addresses) is a step of planner/search, so revision 3 changes each key: the dump runs with the two kept tags, and pois runs with has_poi. planner/places and the grid steps read the pois files and follow through the digest.
  • This pull request starts no data build.

Public docs

docs/content/software/trips.md changes in the merge commit, because the change is a conflict resolution. It describes the planner after the merge.

Checks

Run on the head of this branch, on a Mac. The builds and checks took about 10 minutes of wall time in total; the longest were the browser journey (2 min 7 s) and obc suites prepare (1 min 46 s).

  • npm run check in builder/web: 1,684 files, 0 errors, 0 warnings.
  • npm test in builder/web: 230 files, 1,743 tests pass. Before that, npx vitest run src/components/planner/ src/lib/planner/: 67 files, 417 tests pass.
  • npm run build:planner: the bundle builds.
  • npm run test:planner --prefix builder/web/tests/browser (the journey with stub hosts): 42 pass, 1 fails. routes.test.js "the Routes page with a selected route" fails at 1440 px: the text of the card is 135 px wide and the test expects 140 px, so the profile and the button move by 2 to 5 px. The fixed boxes agree. The code of the Routes page is that of develop; the merge changes only three icons of lib/ui/Icon.svelte in src outside the two planner folders. I read this as the text width of this Mac. CI must confirm it.
  • obc test -p planner-router-build: 93 pass. obc test -p planner-router: 19 pass. obc test -p planner-service: 4 pass.
  • obc test -p obc-data-steps --no-fail-fast: 77 pass, 2 fail: planner::tests::a_geofabrik_region_reads_its_newest_extract_and_the_tiles_of_its_bounds (the last digit of a latitude) and snow_reference::tests::a_version_makes_the_seasons_of_its_blocks_that_the_version_before_lacks. These are the two tests that fail on this Mac without a change.
  • obc exec cargo clippy -p obc-data-steps --all-targets --no-deps -- -D warnings: clean.
  • python3 docs/build_docs.py --check-links, obc prose --check, python3 tools/cairn/cairn.py check (39 violations, the baseline, 0 new) and obc suites check: all pass.
  • obc test affected --base origin/planner-ux --dry-run selects every suite, because the diff holds all of develop. I did not run the selection: CI is the gate.

Left out:

  • iOS: no Swift build and no Swift test. The merge takes the Swift files of each side as they are; no Swift file had a conflict. The files of develop (PlannerPreviewRoutes.swift, SignedRoutes.swift, PlannerPreviewLayers.swift, PlannerPreviewModelTests.swift) use no symbol that the epic deleted.
  • The search suites of planner/search: develop did not change planner/search or host/obc-search-bake.
  • The other browser suites (test:data, test:trailview, ux-study), the Python suites and the firmware checks.
  • A check with miles in the browser. Unit tests cover miles for the totals, the route notes, the gap label and noteDistance.

A look in a real browser (the dev server against the live services, 1440 × 1000, a two-day plan from Wachseldorn to Altdorf):

  • The top row is one row at 1440 px. At 1150 px the middle column has two lines (decision 1).
  • The side panel shows the notes row, the trip row (date chip, "2 days", stay chip) and the day list with the + between the days.
  • "Choose" opens the night page with no search field. A town opens its card with the radio group and one button. Escape goes back one level each time.
  • The drawer shows the totals in its title row, open and closed. The profile heights are whole hundreds (400 m, 1700 m).
  • The live routing data is from before routing: a road for traffic to a destination is an open road #3375, so a note "Only for destination" still shows.

The screenshots are in the private temp folder of the session, not in the repository: shot-2-two-days-1440.jpg, shot-2b-notes-open-1440.jpg, shot-3-night-page-1440.jpg, shot-4-town-card-1440.jpg, shot-5-drawer-closed-1440.jpg, shot-6-top-row-1150.jpg.

Review round 1

Two reviewers checked the pull request. These changes follow:

  • 66022373c routing: the source layer of the routing bake is at revision 2. See "Layer revisions".
  • 30ef2b972 planner: one Escape closed the route details of the map and also went back one level in the side panel. PlannerMap.svelte now closes the details in the capture phase and prevents the default; the window handler of PlannerApp.svelte leaves an Escape with a prevented default alone. PlannerApp.svelte.test.ts has one test for the page side. In a real browser, with a night page open and the details of a cycling route open, the first Escape closes the details and the second closes the night page.
  • 30ef2b972: route-notes.test.ts has a case for alpine stretches: two T4 stretches within 2 km are one row, and a T5 stretch between them is a row of its own.
  • edce9c06f: a second merge of develop, at be962056e (14 commits: the Routes page sizes, the cover asks of rows, "Open in planner"). It has no conflict and touches no file of the two planner folders. planner-ux did not move.

Checks after these changes, about 75 seconds of wall time:

  • obc test -p obc-data-steps --no-fail-fast: 77 pass, 2 fail, the same two tests as before. clippy with -D warnings is clean.
  • npm run check: 0 errors. npx vitest run src/components/planner/ src/lib/planner/: 67 files, 419 tests pass. npm test: 230 files, 1,751 tests pass.
  • Not run again: the browser journey. The second merge changes the Routes page and its test, where the one local failure of the first run was.

Third merge of develop

008b19a3f merges develop at a954bbd50 (25 pull requests since be962056e). 23 files had a conflict. The rule: keep each change of develop, and keep each removal of the epic. The owner confirmed again that route options, their "?" help and the test route button stay removed. develop had added the "Route options" link and the "?" again in #3695.

Area The epic wanted develop wanted Resolution
Route options (plan-session, PlannerApp, PlannerMap, routing, route-legs, alternatives.rs) No alternatives The "Route options" link and list, the map comparison, primary, unpavedKm Removed. The chunked routing and the gaps of #3719 stay.
Help "?" above the stops No help notes Help behind a "?" Removed, with its 3 tests
Trip settings Dates, day count and stay chip at the top of the day list DayCount above the stops; dates and "Sleep in" in the elevation drawer DayCount above the stops. TripRow has the dates and the stay chip at the top of the day list. Both are hidden on a night page and in an open day. TripRow.svelte.test.ts is deleted: DayCount.svelte.test.ts covers the count.
Elevation profile No day labels, no distance number, no trip row; totals in the title Part selection (#3657), day labels, trip row Part selection stays. No day labels, no distance number, no trip row.
Gaps One note for each gap (gapLabel) legGaps and one combined gapNote routeNotes takes the legs with no legal path as "No legal path" notes. gapNote is deleted.
Popup frame takeFocus in MapCallout The frame moves to Callout.svelte Callout has takeFocus
Escape Town, night page, open day; the map closes its route details first The part, the place list The map selection or mode, then the part, then a town, a night page or the open day. The map closes its route details or its place list in the capture phase. closeList does nothing a second time, so the focus goes back to the mark.
Search box No region block The region block in a row with the "where" chip The epic's file
Route list No "Cumulative from the start" note The note above the list No note
PLANNER_ROUTING Revision 20 Revision 21 Revision 22
trips.md, web-planner.md One paragraph for both; the part section names no route options

Checks on 008b19a3f:

  • npm run check: 0 errors, 0 warnings. npm test: 239 files, 1,845 tests pass.
  • cargo test -p planner-router -p planner-service: all pass.
  • npm run test:planner (browser): 55 of 56 pass. ride-page.test.js fails with the FIT file name "download". It fails the same way on develop at a954bbd50, and the merge changes no file of the ride page.
  • prose --check, cairn check (0 new), build_docs.py --check-links, cargo fmt --check: clean.
  • Not run: obc test -p obc-data-steps (only a revision number changed), the iOS suites (no Mac here).

Generated by Claude Code

timohueser and others added 30 commits November 18, 2026 01:46
…obil

Route sources: SchweizMobil in the registry, and a reader of route source files
…probe read one value

Co-Authored-By: Claude Opus <noreply@anthropic.com>
routing: no French route mark, decided per route; regions over several countries
- The close button of the card moves the focus to the row with no
  scroll: on a phone the list is above the map, and the page jumped up to
  the list.
- The notice of TrailView stands under the zoom buttons, as wide as they
  and the control together. So the room of the credits is the width of
  the control alone, and the open credits are not a narrow column on a
  phone.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
A trail is a record with no `trail` that has two or more official
stages, or takes more than 12 hours of moving time. The routing step
writes one far file for each zoom 7 cell of the release bounds
(`routes/trails/7-X-Y.json`): each trail of the cell with the numbers
of a list row and the pieces of its far line there, simplified at
100 m, or coarser while the file is larger than 400,000 bytes. It
also writes one trail record file for each trail
(`routes/long/ID.json`) with the whole record.

`Kept` keeps the moving time of the plan route. `catalog::write`
returns the trails. The metrics get `trail_hours` and `trail_files`.
`planner/routing` revision 17.

Mode: change #3615

Co-Authored-By: Claude Opus <noreply@anthropic.com>
trailview: the ground probe runs again only when the eye's height moves more than 1 cm
On a phone the card is fixed at the bottom of the screen. A card higher
than the screen had its close button above the screen, where no scroll
of the document reaches it. The place that fixes the card gives it a
height limit (`--card-height`): the rows of the card keep their height,
and the row of the description takes the rest and scrolls. The catalog
allows a description of 200 characters, which gives a card of 248 px, so
a card of today keeps its look; the limit is for a text that is longer.
The TrailView window gives the card the room over its small map.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The grid pack publishes `routes/trails/` and `routes/long/` of the
routing step under `routes/tiles/` with their folders, as the route
catalog contract names them. The step contract lists the new outputs.
Two sentences of the route catalog contract are narrowed to be true for
the far files and the trail record files.

`trail_tiles` takes references and removes the far line of each entry
once, so the rounds copy no unsimplified line. The metrics count the
trails by moving time only (`trails_by_hours`) instead of listing the
time of each trail. `planner/routing` revision 18.

Mode: change #3615

Co-Authored-By: Claude Opus <noreply@anthropic.com>
A row and the card of a signed route read its cover from the live set of
route pictures (specs/route-pictures.md), and draw its line on it in the
colour of the route with a white glow. With no cover, they show the line
on a plain ground as before. The map shows the credits of the cover of
the card.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
LegCache.route cuts the legs that it must request into chunks of at
most 64 points that share their end points, with no chunk end at a
turnaround. It sends the chunks two at a time. A chunk that does not
join the chunk before goes out again, pinned to the end of that chunk.
A chunk that gets busy or internal goes out again after 1 s, at most
twice. joinedPlan has no point limit now.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The device's trip progress record holds 32 days, and a send to it holds
32 waypoints. The web planner and the phone app read one constant each;
the tests read it in place of the old literal.

Co-Authored-By: Claude Opus <noreply@anthropic.com>
Co-Authored-By: Claude Opus <noreply@anthropic.com>
Each chunk is pinned to the end of the chunk before, so each join holds
and no chunk goes out twice. Each gap of an edit goes out by itself,
pinned to its own cached neighbours, and each chunk stays in the cache
as it arrives. Route options need a plan that fits one request. The
saved versions are calculated one after the other, so that they share
the cached legs. The catalog spec says what a client without split
requests offers.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A picture that goes aborts its load, and a row and the card of one
route share one load and one bitmap. A failed manifest read is read
again at most once in a minute, and a failed cell file is not kept.
While a cover shows, the map shows the credits of the set. The unit
test fixes the key that the cover script writes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…er its picture

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A pinned gap that finds no path goes out again with one more leg on
each side, so that a neighbour point can pass the other way. Only when
that finds no path too does the whole run follow.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…nside

A chunk that finds no path under the pin of the chunk before goes out
again from one leg earlier, also in a run with no cached legs. A gap
extends past every turnaround next to it, so that no request starts or
ends at one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The catalogue of a release with trail files has the templates `trails`
and `trail`. The tile Worker serves the far files and the trail record
files. The web reader gives the trails of a far view, with the pieces of
each trail joined over its far files, and the whole record of one trail.
The planner/index layer goes to revision 3, so that each release gets
the new catalogue.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A run without cached legs stops after the widened request finds no
path, so it does not send all its chunks again. Each failing chunk gets
its own widened request, also in the second gap of a run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
planner: route long plans in several route requests
A segment of a simplified far line can cross the view with no vertex in
it, so a trail is in a far view when a segment meets the box. The far
files of the last two views stay in the cache, no more. The optional
config URLs are one list.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A catalog record gets the optional field `credit`. The Explore card shows
it as "Route data: …", and the credits of the Explore map and of the
TrailView window list each credit of their routes once. The map gets its
extra credits also while no card is open.

Co-Authored-By: Claude Opus <noreply@anthropic.com>
Co-Authored-By: Claude Opus <noreply@anthropic.com>
Planner release: publish and read the trail files
The phone route detail shows "Route data: …". The credits of a long route
split at "; ", so each credit shows once. The Explore page escapes the
plain-text record credits before the map joins them as HTML.

Co-Authored-By: Claude Opus <noreply@anthropic.com>
… route; drop empty credit parts

Co-Authored-By: Claude Opus <noreply@anthropic.com>
timohueser and others added 13 commits November 19, 2026 15:01
…m the trail name on a stage page

The page #/trail/<id> shows the totals, a row for each stage in the map data, the stage lines on the map with a dot at
each stage border, and the profile of the whole trail with the stage borders and numbers. It reads the trail record file
and the cell files of the stages, and makes no route request.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…s back

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… the trails test drops a record that tests nothing new

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…s no start mark; the review fixes

The box of the map is the union of the bounds of the stage items, with 48 px padding on the place; the border dots come
from the item lines. A map with border dots shows no start mark. "‹ Explore" always goes through its handler. The left
card counts its stages directly. Tests cover stageNumbers, the taken ticks and the Explore state of the way back.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
router-build: the route catalog stores the moving time of each route and stage
The page shows `moving_s` of the record from the first frame, before and
after its routing answer, so a stage has the same time on its trail page row
and on its own page. A record with no stored time shows the routed time.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…record

The trail page test checks that the Stage 2 row shows the time of its stage page.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
web: the page of a signed route shows the stored catalog time as Time
timohueser and others added 13 commits November 19, 2026 19:05
… from the length rules

A deck way is exempt from deck_min_m and the culvert rule only where a
passage station lies strictly between its A lines; G8 uses the same test.
The deck group merge reads the passages that its tag end keeps. G9 reads an A
line at its tag end the snap length in. The water exception says what it
tests. The storeys test's joint limit says where its count comes from.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…f booked height

points_to_obcr measures the height step from the dead-band reference of the
kept points, so the stored ascent and descent equal those of all points with
fewer points per km.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The route service adds `point` to a `no_snap` error answer and `leg` to a `no_path` error answer,
both counted from 0, so a client can leave out the point or keep the leg as a gap.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…r from a road, and keeps a leg with no legal path as a gap

- The leg cache sends the chunks of a long plan with at most two requests in flight. After the
  answers, a leg that does not join its neighbour goes out again, pinned to both neighbours. A new
  routing package restarts the cache once.
- A `no_snap` point is left out of the line, with the existing note and a Show button. A start, a
  finish or the end of a run beside a manual leg stays; the legs at it become gaps.
- A `no_path` leg becomes a cached gap: a dashed straight line with no time and no climb, and the
  note "No legal path between 2 points".
- The device sync does not send a line with a gap.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…e device

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… climb differs

The test line bends only at its second point, so the last-kept-point rule
books 4/4 and fails it. The spec sentence names the two cases where the stored
ascent and descent differ from a dead-band over the input points.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…-step

obc-route: planned route lines keep a point at each 3 m of booked height
… hangs the page; review fixes

- The leg of a gap finds its stop with a bounded search. The way back from a visit at the end of a
  run belongs to the visit.
- Only the first and the last request point stay when no road is near; a visit at the end of a run
  is left out and keeps its anchor, where the manual leg after it starts. The note names its place.
- An index outside the request goes up as an error, so a bad answer cannot repeat a request.
- A service that names no leg in `no_path` gets the wider request at each pinned end.
- `RoutingLine.leftOut` goes: nothing read it.
- The test of a point with no road near it puts the point into the second request.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…it gets one, before a routed or a manual leg

The stop of a visit is the end of its leg from the anchor to the place, where the line reaches the
place, so `routeGaps` measures only a real miss.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Planner: route long plans two requests at a time, and keep the trip when one point or leg fails
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…-passage

ground plan graph: a bridge keeps the passage under it (#3639 P0)
…merge-develop-2

Third merge of develop into the planner-ux epic (#3286), at a954bbd.
The epic removes route alternatives, the "Route options" list, its "?"
help and the test route button; develop added the "Route options" link
and the "?" again in #3695. They stay removed.

Resolutions that are more than a choice of side:
- Callout takes the takeFocus option of the town card, because develop
  moved the popup frame out of MapCallout.
- The legs with no legal path (develop #3719) are route notes
  ("No legal path") beside the gaps, not one combined line.
- The trip row keeps the dates and the stay chip at the top of the day
  list; the day count is develop's DayCount above the stops, hidden on
  a night page and in an open day, as the trip row is.
- The profile keeps develop's part selection and the epic's title row,
  with no day labels, distance number or trip row.
- Escape: the map selection or mode, then the part, then a town, a
  night page or the open day. The map closes its route details or its
  place list in the capture phase; closeList does nothing a second time.
- PLANNER_ROUTING is at revision 22, above both sides.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvLH9RnCjTsKg1o5SZM
@timohueser
timohueser merged commit 0293452 into planner-ux Nov 19, 2026
24 of 26 checks passed
@timohueser
timohueser deleted the claude/planner-ux-merge-develop-2 branch November 19, 2026 20:22
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.

2 participants