You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
TrailView parts P6: the pair of tubes, the opening in rock and the free tube end #3683
Size: M (the builder crate and the ladder values; no new contract)
Problem
Three kinds of tunnel end still draw as a plain tube after P4. A pair of tubes (motorway carriageways, two tracks in separate tubes) got two head walls in the prototype that did not line up or join. Paths and tracks often enter the rock directly, with no head wall. Where the hill never covers a tunnel, or covers it only past the collar limit of its class, the end needs a tube end that stands free of the ground.
Scope
pair: one head wall on one frame line for both openings (mate of P1), one frame per opening, one footprint and boxes for both.
rock: an opening in the rock with no head wall, in the look rock, with a short reveal into the hill.
free: the tube's outer shell for collar metres with a slanted end cut; its shell reaches below the lowest ground under it (foot), so it needs nothing from the ground.
Labelled pictures of each kind, today's app | new roads.
Depends on
This item adds Pair, Rock and Free to BUILT, with no new bake; then no chain keeps today's tunnels() on the new roads (#3681, step 2). A covered end that meets no way is a free end with collar 0 (#3678): its look is a question of this design.
On test box B, 41 of the 65 pair records of P1 have tubes that overlap. Their axes are nearer than both outer half-widths (half + shell_m on each side). Examples: motorway carriageways 9 m to 13 m apart with a half of 6.6 m, and double tracks 3.6 m to 4 m apart with a half of 2.5 m. Rung 2 as designed (the wider half + same_tube_gap_m) does not take them into one tube, so they reach rung 5 as a pair. P1 builds the rule as designed.
The options for P6 (the pair part) and the owner:
The pair part cuts the two tubes where they meet: one wide opening, with no wall between the tubes.
Rung 2 takes two openings into one tube where their tubes overlap (|mate| < the sum of both outer half-widths + same_tube_gap_m). The wider tube widens to the far edge of the other. For two carriageways of the same width, the axis then stands far off the middle.
The owner chose option 1. Where the tubes of a pair overlap, the pair part makes one opening whose outline follows the arches of both tubes. There is no wall between the tubes. Each road stays on the axis of its own arch. Rung 2 of P1 does not change: these ends stay a pair.
A kind draws when it is in BUILT of host/obc-trailview-parts/src/lib.rs. The chord endsnone stays not built.
The tube builds each chain whole along its chords. Where the tubes of a pair overlap (option 1), the cut of the two tubes and the one opening over both arches are P6's work.
The daylight already counts the opening of the mate of a pair (mate to the left on the frame line).
arch.rs has general pieces: Wall, collar and tube::shell (with the face at a step of height). The free end and the rock opening can use them.
#3681 changed its step 5. The bake now writes the cut of each built end (the tag cut of the tunnel-end): where the ground of the ground plan, with the openings of built chains open, lies inside a box of the part. The ground plan leaves a hole in its mesh only there. So:
Adding Pair, Rock and Free to BUILT is not enough. It needs one features bake, so that those ends get their cut and the bake opens their stretches as the device does. "With no new bake" in Depends on no longer holds.
Each new part must fill its boxes as the arch does after TrailView parts P4: the new roads draw the parts #3681: boxes at the middle of the wall, one per straight piece, and an outer shell to the end of the last box with an end face. For a pair whose tubes overlap (option 1), the boxes of both tubes together must hold the one opening. A free end's slanted cut must also lie inside its boxes.
boxes::boxes of obc-trailview-parts gives the boxes of a built end. There is one box for the opening, then one per straight piece of the tube. A box has 12 words: after the spring come the slopes of its two ends, which lie in the planes of the tube's rings.
A pair needs boxes that hold the room of both tubes and of their one opening, each at the middle of a wall. The opening in rock needs a box for its reveal. The free end's boxes must lie inside its slanted shell.
The shell runs to the end of the last box, and a face closes it there (arch::end_face). A free end's shell must also close.
The head wall and the shell go down to boxes::foot.
After the new kinds are in BUILT, one features bake writes their cut.
When Pair, Rock and Free join BUILT, the viewer and the ground plan take them with no more code. The ground plan also opens a pair's second opening (mate_way, mate_shift), and the tunnel line stamps nothing from the way end of each of its openings to the start of the part's cover. The part's ring takes the ground from plants, kind points in it are not drawn, and its boxes go to the terrain (64 at most with the plan, nearest first). One features bake writes the cut. The fault of the sampled cut at the plan's steps (#3684 comment) applies here too.
Data: the records and the measure of the bake on test box B, features revision 36 (all five kinds built; the ground at
each free end from the bake's own measure, printed by a local probe), and the build of branch claude/trailview-3683-pair-rock-free (pictures on the owner's PC: parts/p6/pics/).
1. The free tube end: a square cut with a rim, and the shell's own end face
Facts of box B. 191 free ends. 23 are covered at their opening (see below). 168 have a shell. Beside the shell
(at its outer walls, half + shell_m across), the ground in the first metre stands 2.9 m over the floor at the
median (p75 5.2 m, p90 6.7 m): most free ends lie along a slope, buried at one side. A cut that leans back 1:1 from
the floor at the frame line lies under that ground by more than 0.3 m at 146 of the 168 ends. A cut of the vault
only (from the spring) lies under it at 35. Where the ground stands over the cut, the part has no face there, and the
ground ends at the box of the room: a torn edge.
Rule: the free end is cut square on its frame line (90° to the axis). A rim closes the shell there: the frame
of the opening grown by the shell, 0.95 m (frame_m + shell_m) round the room, from the foot up, with the reveal
and the frame of the arch, but no head wall and no wings. The rim is also the head of rock (look rock).
The far end: no other face. The shell runs to its reach and a face closes it there, as every shell since P4
(arch::end_face). At 53 ends the hill covers the shell before that face (covered past the limit). At 36 ends the
chain is never covered and longer than 2 × free_max_m: there the face shows where the shell ends, 40 m in.
A covered end that meets no way (TrailView parts P1: the bake places the tunnel parts in the feature tiles #3678): 23 ends (17 paths, 6 service ways and tracks); all meet no way. The
ground at the frame line stands 3.7 m to 789 m over the floor (median 18 m): the hill covers the whole tube there.
It gets no opening: a dark face closes the tube on the frame line, no box lies in front of it, and no daylight
comes in. Nothing of it shows outside.
2. The middle wall of a pair: none where the shells overlap
Facts of box B. 55 pairs. At 29 the rooms overlap (by 0.25 m to 6.3 m, median 1.4 m), at 31 the shells. The
floors of the two openings differ by 0.41 m at the median (p90 2.1 m, at most 7 m). Of the 24 pairs whose shells do
not overlap, the narrowest space between the two frames is 0.19 m (a narrow-gauge pair), the next 0.49 m.
Rule (the owner's option 1, made general): each opening of a pair draws the head wall of both (one top over the
pair's frame, the pair's rows) with its own hole, as far as the plane in the middle between the walls of the two
tubes. Where both holes reach that plane, they are one opening whose outline follows both arches. Along the tubes,
where the shells of two built tubes overlap, the inside and the shell of each end at the middle between their walls,
and faces close the steps where their floors, rooms or shells differ there. So a middle wall stands only where the
shells do not overlap, and it is as wide as the space between the two frames (0.19 m at least on box B).
Each tube runs 5 cm past the middle: the vertices are rounded to whole world units (2.5 cm), and two cuts must not
open a gap. Without it, the sky showed through a hairline at the crown of a double track.
A tube finds the other tube by its cells, not by the record of its pair, so a tube in a tile with no record of its
pair ends at the middle too.
The same rule ends a head at a near opening that leads the same way (within pair_max_m and pair_max_deg, at one
level), at most one on each side: the mate of a pair, else the nearest. On box B, at 37 places the head of one end
reached the tube of another that is not its mate: 16 underpasses beside an underpass, 7 arches beside an arch, 6
arches beside a free end, 8 others. Develop draws the wall of one in the opening of the other there.
Two changes of the bake (features revision 36):
12 pairs of revision 35 had a chain with an end of no part (none), which never draws as parts. A pair is one part
of two chains, so the bake now pairs two openings only where both chains have a part at both ends.
The record of a pair takes the heading of the opening farther in, square to its frame line. Its mate was already
measured along that line; at 2 pairs the mate's opening lay 1.1 m and 2.2 m off the line of the old heading.
3. The ragged edge at a free shell: a fault of the ground
In the prototype (place 499), the cover stamp of the tunnel line started at the end of the shell, and the ground
there was ragged.
In this build there is no bake cut, and the part is a closed solid down to its foot (1 m under the lowest ground
over its frame and collar). The tunnel line stamps nothing from the opening to collar + 1 m, and the shell runs
farther, to its reach. So the edge where the shell meets the ground is the line where the ground's triangles cross
the smooth shell. At 499 (free-open-top) that line is ragged along the 40 m of shell, as the terrain's cells are
coarse there; no face of the part ends at it, and no gap opens under the shell. A fault of the ground, if any.
The bake decides it; the device only cuts at the middle, from rings where both tubes are. Design R stays. Nothing
gets thicker than P5: the shell stays shell_m, a wide tube only widens as rung 2 widens it today.
Tests, for tubes A and B with inside half-widths hA ≥ hB, at a point of B's axis: d is its distance to A's
axis, square to A's chord, with the foot between A's two ends; dz the difference of the floors.
level: |dz| ≤ (heightA + heightB) / 2 (the device's test today)
meet: level and d < hA + hB + 2 shell_m (the shells overlap; the device's reach today)
inside: level and d ≤ hA (B's axis lies in A's room)
mostly inside: inside at more than half of B's metres from opening to opening (from the chain's end where an end
has no opening), read every metre along B's axis. A is the wider of the two (larger half, then the lower id).
1. Absorb. A tube that is mostly inside a wider one has no part at either end (rung 2, "same tube"). The wider
tube's half grows to the largest d + way_half(B) + margin(A) over the metres where B is inside. B's ways get the
tag parts = in in where A draws as parts: the viewer then draws no tunnel of B's own, and B's way runs on A's
floor. This replaces the test at one opening, so same_tube_gap_m and same_tube_stagger_m go. An opening of an
absorbed tube that lies outside A's room (B parts from A before it opens) has no part; the bake counts these
(absorbed_openings_outside).
2. Join at an opening. Two openings of different tubes that lead the same way (within pair_max_deg and pair_max_m, as for a pair) are a pair for any kinds where the two tubes meet at the opening farther in. Both use its
frame line: the nearer opening moves in, as for a pair today; where its way cannot reach the line, they stay two
openings and the bake counts them (joins_that_do_not_reach). The head is a wall where either opening's own kind has
one (arch, box), else a rim round both (rock, free; the record then has no face). Each tube keeps its own collar by
its own kind (a free end its 40 m). The section stays the tube's (an underpass stays a box). The pairs of rung 5 stay
as they are.
3. Meet and part in the hill. On two built tubes the bake puts a vertex where their shells start or stop
overlapping between the openings (read every metre, the place found to a centimetre). The device clips a piece only
where both of its rings meet the other tube; there is no halving, no bound plane and no approach. Where a clipped
piece meets one that is not (only at an opening of rule 2 that could not move), a face closes the shell on that ring.
At a fork a chain still ends with no opening (none), and does not draw as parts; the tests read chains, so a
chain that is mostly inside a wider one is absorbed whether it ends at a fork or not. Where two tubes part, their
shells stop overlapping at a vertex; past it each tube has its own walls again, closed, as it is cut only where both
rings meet.
One change to rule 1 before the bake, from box B's data: a footway beside a road in a tunnel lies 3.4 to 5.5 m from the
road's axis, and the road's half is 3.1 to 4.6 m. So "its axis in the wider room" catches few of them; stagger-path
(4.7 m from a road of half 3.6) is not one. Rule 1 now reads: a narrower tube that meets a wider one (their shells
overlap, at one level) at more than half of its metres runs in it. Two tubes of one width are never absorbed, so
pairs such as pair-over stay pairs.
Back to the rule as first posted ("inside": the narrow axis lies in the wide room). The bake with "the shells meet"
absorbed a road whose shell only touched a motorway's, and the motorway tube grew to a half of 16.8 m; on box B it then
stood in front of a footway underpass. With "axis in the room", a wide tube grows at most by the other's way, and a
footway that lies just off a road's room (stagger-path: 4.7 m off a half of 3.6 m) is a pair that shares one opening.
Activity
timohueser commentedon Nov 19, 2026
From P1 (#3678): pairs whose tubes overlap
On test box B, 41 of the 65
pairrecords of P1 have tubes that overlap. Their axes are nearer than both outer half-widths (half+shell_mon each side). Examples: motorway carriageways 9 m to 13 m apart with ahalfof 6.6 m, and double tracks 3.6 m to 4 m apart with ahalfof 2.5 m. Rung 2 as designed (the widerhalf+same_tube_gap_m) does not take them into one tube, so they reach rung 5 as a pair. P1 builds the rule as designed.The options for P6 (the pair part) and the owner:
|mate|< the sum of both outer half-widths +same_tube_gap_m). The wider tube widens to the far edge of the other. For two carriageways of the same width, the axis then stands far off the middle.Generated by Claude Code
timohueser commentedon Nov 20, 2026
Owner decision: pairs whose tubes overlap
The owner chose option 1. Where the tubes of a pair overlap, the pair part makes one opening whose outline follows the arches of both tubes. There is no wall between the tubes. Each road stays on the axis of its own arch. Rung 2 of P1 does not change: these ends stay a
pair.Generated by Claude Code
timohueser commentedon Nov 20, 2026
P3 (#3680, #3754): what P6 gets from the build
BUILTofhost/obc-trailview-parts/src/lib.rs. The chordendsnonestays not built.mateto the left on the frame line).arch.rshas general pieces:Wall,collarandtube::shell(with the face at a step of height). The free end and the rock opening can use them.Generated by Claude Code
timohueser commentedon Nov 20, 2026
P4 (#3681): this item needs a features bake
#3681 changed its step 5. The bake now writes the cut of each built end (the tag
cutof thetunnel-end): where the ground of the ground plan, with the openings of built chains open, lies inside a box of the part. The ground plan leaves a hole in its mesh only there. So:Pair,RockandFreetoBUILTis not enough. It needs one features bake, so that those ends get their cut and the bake opens their stretches as the device does. "With no new bake" in Depends on no longer holds.Generated by Claude Code
timohueser commentedon Nov 20, 2026
P4 (a) (#3761): the boxes and the shell of a part
boxes::boxesofobc-trailview-partsgives the boxes of a built end. There is one box for the opening, then one per straight piece of the tube. A box has 12 words: after the spring come the slopes of its two ends, which lie in the planes of the tube's rings.arch::end_face). A free end's shell must also close.boxes::foot.BUILT, one features bake writes their cut.Generated by Claude Code
timohueser commentedon Nov 20, 2026
P4 (b) (#3763): what changes for P6
When
Pair,RockandFreejoinBUILT, the viewer and the ground plan take them with no more code. The ground plan also opens a pair's second opening (mate_way,mate_shift), and the tunnel line stamps nothing from the way end of each of its openings to the start of the part's cover. The part's ring takes the ground from plants, kind points in it are not drawn, and its boxes go to the terrain (64 at most with the plan, nearest first). One features bake writes the cut. The fault of the sampled cut at the plan's steps (#3684 comment) applies here too.Generated by Claude Code
timohueser commentedon Nov 21, 2026
P6 design: the open questions
Data: the records and the measure of the bake on test box B, features revision 36 (all five kinds built; the ground at
each free end from the bake's own measure, printed by a local probe), and the build of branch
claude/trailview-3683-pair-rock-free(pictures on the owner's PC:parts/p6/pics/).1. The free tube end: a square cut with a rim, and the shell's own end face
Facts of box B. 191 free ends. 23 are covered at their opening (see below). 168 have a shell. Beside the shell
(at its outer walls,
half+shell_macross), the ground in the first metre stands 2.9 m over the floor at themedian (p75 5.2 m, p90 6.7 m): most free ends lie along a slope, buried at one side. A cut that leans back 1:1 from
the floor at the frame line lies under that ground by more than 0.3 m at 146 of the 168 ends. A cut of the vault
only (from the spring) lies under it at 35. Where the ground stands over the cut, the part has no face there, and the
ground ends at the box of the room: a torn edge.
of the opening grown by the shell, 0.95 m (
frame_m+shell_m) round the room, from the foot up, with the revealand the frame of the arch, but no head wall and no wings. The rim is also the head of
rock(lookrock).(
arch::end_face). At 53 ends the hill covers the shell before that face (covered past the limit). At 36 ends thechain is never covered and longer than 2 ×
free_max_m: there the face shows where the shell ends, 40 m in.ground at the frame line stands 3.7 m to 789 m over the floor (median 18 m): the hill covers the whole tube there.
It gets no opening: a dark face closes the tube on the frame line, no box lies in front of it, and no daylight
comes in. Nothing of it shows outside.
2. The middle wall of a pair: none where the shells overlap
Facts of box B. 55 pairs. At 29 the rooms overlap (by 0.25 m to 6.3 m, median 1.4 m), at 31 the shells. The
floors of the two openings differ by 0.41 m at the median (p90 2.1 m, at most 7 m). Of the 24 pairs whose shells do
not overlap, the narrowest space between the two frames is 0.19 m (a narrow-gauge pair), the next 0.49 m.
pair's frame, the pair's rows) with its own hole, as far as the plane in the middle between the walls of the two
tubes. Where both holes reach that plane, they are one opening whose outline follows both arches. Along the tubes,
where the shells of two built tubes overlap, the inside and the shell of each end at the middle between their walls,
and faces close the steps where their floors, rooms or shells differ there. So a middle wall stands only where the
shells do not overlap, and it is as wide as the space between the two frames (0.19 m at least on box B).
open a gap. Without it, the sky showed through a hairline at the crown of a double track.
pair ends at the middle too.
pair_max_mandpair_max_deg, at onelevel), at most one on each side: the mate of a pair, else the nearest. On box B, at 37 places the head of one end
reached the tube of another that is not its mate: 16 underpasses beside an underpass, 7 arches beside an arch, 6
arches beside a free end, 8 others. Develop draws the wall of one in the opening of the other there.
Two changes of the bake (features revision 36):
none), which never draws as parts. A pair is one partof two chains, so the bake now pairs two openings only where both chains have a part at both ends.
matewas alreadymeasured along that line; at 2 pairs the mate's opening lay 1.1 m and 2.2 m off the line of the old heading.
3. The ragged edge at a free shell: a fault of the ground
there was ragged.
foot(1 m under the lowest groundover its frame and collar). The tunnel line stamps nothing from the opening to
collar+ 1 m, and the shell runsfarther, to its reach. So the edge where the shell meets the ground is the line where the ground's triangles cross
the smooth shell. At 499 (
free-open-top) that line is ragged along the 40 m of shell, as the terrain's cells arecoarse there; no face of the part ends at it, and no gap opens under the shell. A fault of the ground, if any.
Generated by Claude Code
timohueser commentedon Nov 21, 2026
Overlapping tubes are one room (round 2 of #3780)
The bake decides it; the device only cuts at the middle, from rings where both tubes are. Design R stays. Nothing
gets thicker than P5: the shell stays
shell_m, a wide tube only widens as rung 2 widens it today.Tests, for tubes A and B with inside half-widths
hA ≥ hB, at a point of B's axis:dis its distance to A'saxis, square to A's chord, with the foot between A's two ends;
dzthe difference of the floors.|dz| ≤ (heightA + heightB) / 2(the device's test today)d < hA + hB + 2 shell_m(the shells overlap; the device's reach today)d ≤ hA(B's axis lies in A's room)has no opening), read every metre along B's axis. A is the wider of the two (larger half, then the lower id).
1. Absorb. A tube that is mostly inside a wider one has no part at either end (rung 2, "same tube"). The wider
tube's
halfgrows to the largestd + way_half(B) + margin(A)over the metres where B is inside. B's ways get thetag
parts=in inwhere A draws as parts: the viewer then draws no tunnel of B's own, and B's way runs on A'sfloor. This replaces the test at one opening, so
same_tube_gap_mandsame_tube_stagger_mgo. An opening of anabsorbed tube that lies outside A's room (B parts from A before it opens) has no part; the bake counts these
(
absorbed_openings_outside).2. Join at an opening. Two openings of different tubes that lead the same way (within
pair_max_degandpair_max_m, as for a pair) are a pair for any kinds where the two tubes meet at the opening farther in. Both use itsframe line: the nearer opening moves in, as for a pair today; where its way cannot reach the line, they stay two
openings and the bake counts them (
joins_that_do_not_reach). The head is a wall where either opening's own kind hasone (arch, box), else a rim round both (rock, free; the record then has no
face). Each tube keeps its own collar byits own kind (a free end its 40 m). The section stays the tube's (an underpass stays a box). The pairs of rung 5 stay
as they are.
3. Meet and part in the hill. On two built tubes the bake puts a vertex where their shells start or stop
overlapping between the openings (read every metre, the place found to a centimetre). The device clips a piece only
where both of its rings meet the other tube; there is no halving, no bound plane and no approach. Where a clipped
piece meets one that is not (only at an opening of rule 2 that could not move), a face closes the shell on that ring.
At a fork a chain still ends with no opening (
none), and does not draw as parts; the tests read chains, so achain that is mostly inside a wider one is absorbed whether it ends at a fork or not. Where two tubes part, their
shells stop overlapping at a vertex; past it each tube has its own walls again, closed, as it is cut only where both
rings meet.
timohueser commentedon Nov 21, 2026
One change to rule 1 before the bake, from box B's data: a footway beside a road in a tunnel lies 3.4 to 5.5 m from the
road's axis, and the road's
halfis 3.1 to 4.6 m. So "its axis in the wider room" catches few of them; stagger-path(4.7 m from a road of half 3.6) is not one. Rule 1 now reads: a narrower tube that meets a wider one (their shells
overlap, at one level) at more than half of its metres runs in it. Two tubes of one width are never absorbed, so
pairs such as pair-over stay pairs.
timohueser commentedon Nov 21, 2026
Back to the rule as first posted ("inside": the narrow axis lies in the wide room). The bake with "the shells meet"
absorbed a road whose shell only touched a motorway's, and the motorway tube grew to a half of 16.8 m; on box B it then
stood in front of a footway underpass. With "axis in the room", a wide tube grows at most by the other's way, and a
footway that lies just off a road's room (stagger-path: 4.7 m off a half of 3.6 m) is a pair that shares one opening.