The library was a heap of half-built services - fitting, parallelism, niwelety,
an SVG writer - that never met each other. In their place is one authored
Document and one solver that lays it whole: tracks as element chains, turnouts
placed on them, and a Solution that is rebuilt from scratch every time and never
fed back into the document. What cannot be laid comes back as a diagnostic with
the number in it rather than as silence or a fitted-away error.
Turnouts are dimensioned the way the drawings dimension them. The blade follows
Koc's rys. 5.9 - the curve is theoretically tangent to the opornica at PR, the
blade starts at a with a planed nose of d, and A is where the rail stands u
thick - and a + d derived that way lands on his tablica 5.1. Bending follows
5.3.6: it rotates the triangle OBC about the centre of the turnout, so the
tangent t and the crossing angle stand and both radii follow from them. That is
Koc's (5.7) and (5.8) without coding either, and it puts a symmetrically bent
Rz 49E1-190-1:9 on the catalogue's Rlus 49E1-380,292-1:9. A rozjazd standing on
a curve is bent onto it; one on a track drawn straight can carry a bend of its
own. The szyny laczace take up the difference, because the switch and the frog
are only bent, not re-cut.
The plan panel draws all of it: the catalogue drawing of a type with its
construction points, the theoretical arc dashed where no rail runs, and the
blade nose as the straight it is cut as. Dragging a straight now stops where it
would pull the track out from under a turnout.
File format goes to m0s 6. Versions 3 to 5 still open.
Checkpoint of the editor branch so work isn't carried loose across a branch
switch. Builds.
- maj0sted turnout geometry moved into the domain: domain::lay_turnout lays the
diverging path (pre-blade lead, blade at beta, arc of alfa-beta, frog rail) and
solve_junction is a thin adapter onto it.
- the maj0sted web/ tree renamed to editor/ (namespace and paths) - there is no
WASM here, it's plain native C++ the editor calls.
- click places the switch at its start (początek rozjazdu); the branch's leaving
straight is locked tangent to the frog; switches can be slid along the through
track and flipped L/P.
- Polish turnout templates carry catalogue length (Id-1-consistent lengths), plus
blade-angle and pre-blade fields (left 0 until the catalogue figures are in).
- TURNOUT_SPEC.md: the Bezier-based runtime/model contract.
- also folds in other pending branch changes (scenery, tests) to leave a clean tree.
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 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.
