A switch (rozjazd) is a relation on a through niweleta, not a niweleta - the
through track is never split. It marks the theoretical point on the through
centreline; two tangent lines cross there (the through heading and that heading
turned by the crossing mark 1:n), and the internal diverging curve is fitted to
round the corner between them, reusing the same GapFitter the gap fits go
through. So the curve is free: a plain arc, or a łuk koszowy right inside the
switch. The theoretical tangent lengths and the curve length follow from it and
are read off, not set.
The branch niweleta's leaving straight is now locked to the frog: it starts at
the frog and runs along the frog heading every solve, only its length is the
user's. That enforces tangency no matter what is drawn, and because the first
straight is pinned tangent, any fit in the branch's first gap - including a
basket - meets the switch tangentially by construction.
solve_layout solves the niwelety and the switches together in two passes: solve,
read each frog, pin the branches, solve again. Switches serialize as an ejuncs /
ejunc section; older files without it still load.
Typical Polish turnouts (rozjazdy zwyczajne) are offered as templates that set
the crossing mark and radius together; the length each spans is computed from
them (arc = R*atan(1/n)) and shown, so how long the switch is is known.
The flip was my mistake. Measured against the service instead of guessed:
fetching a box and its northern half shows the half matching the top of the
whole, so the images arrive with their first row at the north edge and the
default texture coordinates were right all along. Both the base map and the
orthophoto go back to them.
Imagery of a given piece of ground does not change, so a fetched tile is now
kept under editor/cache/orto and the next session starts from it. Until now
the cache lived only in memory, which meant going back to a slow and flaky
service for ground that had already been fetched once.
Gaps can be fitted with a basket curve as well: the library's compound mode,
with clothoids at both ends and a chain of arcs between them. Arcs can be
added and removed; the closing one takes whatever deflection the earlier
ones leave, so its length is not offered for editing.
The Geoportal images arrive with their first row at the north edge, which is
the opposite of what the default texture coordinates assume - the vertical
ones are now swapped, for the picker's base map and the orthophoto alike.
With the real map carrying its own detail as the view zooms in, the drawn
city markers had nothing to add and are gone, along with the coordinate
table behind them. Only the country's extent remains, to open the picker on.
-editor no longer inherits the scenery named in eu07.ini. That is whatever
was last driven, which has nothing to do with editing; without an explicit
-s the editor opens on empty ground.
The location picker now shows the real topographic map instead of a
hand-drawn outline, fetched as one WMS image for whatever box is on screen;
the schematic border is gone and only the city list remains, to jump the
view somewhere recognisable. Under the plan itself a georeferenced map gets
the orthophoto, tiled through the layout library's own EPSG:2180 grid and
cache.
What a native build was missing is here: the download (WinHTTP, on worker
threads), the decode and the texture. Finished transfers are handed back on
the thread that drains them, so both the library's cache and the graphics
API stay on one thread. The imagery service turns down browser user agents
and fails the odd request even for a plain one, so requests go out as curl
and are retried a few times.
Starting a new map now drops the loaded scenery and comes back on empty
ground, instead of leaving the previous scenery's contents under a plan that
has nothing to do with them. The reload is deferred to the top of the frame:
taking the editor off the mode stack while it is drawing its own panels
would pull the ground out from under them.
"New map" now asks what is being started. A fictional scenery keeps its own
frame, where coordinates mean nothing outside it. A real one pins the
scenery's zero to a point picked off a map of Poland, so the plan is drawn
in EPSG:2180 - the frame the layout library already works in, and the one
Geoportal serves imagery in.
The projection is the standard transverse Mercator series for PUWG 1992 on
GRS80. Checked against known values: the central meridian comes out at
easting 500000 exactly, Warszawa at 637382/486757, Krakow at 567614/244588,
Gdansk at 477038/720937.
The map behind the picker is a schematic border outline plus the larger
cities, projected from geographic coordinates on first use. It is coarse on
purpose - it is there to point at a region, not to measure against - and the
dialog says so. No network is involved.
The plan panel had its own 2D canvas, which was the wrong place for it: the
layout is drawn on the scenery. The canvas is gone. Points are laid by
clicking the ground under the cursor, the solved geometry is painted over
the rendered world through the colour pass camera, and the panel keeps only
the controls - fits, storage and what is being edited.
The left button belongs to the plan tool while it is up, so a click no
longer also picks or places a scenery node. Clicks landing on a panel stay
with the panel.
With no georeference set the plan frame and the world share an origin, so
easting runs along world x and northing against world z.
Opening the Plan tool now switches the whole colour pass to an orthographic
projection and pins the camera looking straight down, so the plan is drawn
against the loaded scenery rather than in a window of its own. Closing the
tool hands the camera's orientation back; leaving the editor clears the view
outright. The wheel changes how much ground fits on screen instead of moving
the camera, which with no perspective would change nothing but the clipping.
Both configured backends follow: the experimental renderer swaps the pass
projection (and fits its shadow cascades to it), the modern GL renderer uses
its existing ortho helpers. Everything reading the pass projection - the
mouse ray, the frustum test - picks the change up on its own.
The experimental renderer also publishes its colour pass camera now, which
it previously left at the interface defaults, so overlays can place world
points on screen.
A "Plan" panel, opened from Track layout in the editor's menu bar, draws a
niweleta the way the web frontend did: click to lay chained straights, drag
a corner to move it, and pick a fit for each gap between them. The geometry
is entirely maj0sted's - the panel gathers the input, hands the document to
solve_project and paints the rails that come back, coloured by whether they
came from a straight, an arc or a clothoid. Scenery nodes are untouched.
A fit the library cannot produce is simply absent from its answer, so the
gap is labelled as not fitting rather than left looking as though its
parameters took effect. Save/load go through the library's own document
format, so a project written here opens in the same state.
Also translates the comments added with -editor into English, matching the
files they sit in.
