- A file without a schema number is refused again; the example output in
RAG/pierzainer_ground now has schema 2, so the temporary path is not needed.
- The legacy fixture is removed; tests cover the missing and null schema
numbers and a pier without a ground object.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
- ground-model.ts: read schema 2 terrain_grid (origin, step, nx, ny, z),
checked the same way as the stored project form; schema 1 is still read.
- view/terrain.ts: the grid as a mesh in pier-scene coordinates
(x = E − pivot E, z = −(N − pivot N), y = EL), same as the pier frame.
Cells with a missing corner are not drawn.
- scene.ts: draws the terrain in place of the 60 m plane when the model has a
grid; the camera fits the members, not the terrain.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
- A file with no `schema` key is read as schema 1, but every pier must have a
`ground` object, so a renamed field is refused instead of loading as empty.
- `schema` null or any other number is still refused.
- The project file keeps the normalised model and checks it with its own
check (checkStoredGroundModel), so a saved model reopens.
- Temporary: remove the no-schema branch once examples/out is regenerated.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
- ground-model.ts: read and check schema 1 of the Python ground model.
- ② 지반·지층 card: import the json, show the DTM value beside the excel
value, and apply it only when the user presses the button (tracked as an
override, revertible).
- groundEl becomes an editable, tracked input; the project file keeps the
imported model and checks it on open.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The card only checked the minimum spacing. A 12000 mm footing with 5 piles per 열 gives 2650 mm, above the maximum. The card
now marks such a row ("최대 2250 mm를 넘습니다. 본수를 늘리세요") and the alert list gets a warning (P02). Tests check 5 and 6
piles on a 12000 mm footing.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The plan has overall dimension lines (length along the top, width at the right) and two chains: along the length, the
distances from the footing edge to the first pile and between piles; across the width, the same for the pile rows. The
chains are read from the pile positions, so they follow the layout.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The side view has dimension lines with 45° ticks: the footing length (mm) above the footing, the footing height h1 at its right,
and the visible pile length (끊음) at its left. The margins are wider (1.5 m) for the labels; the plan uses the same margin so the
two views stay aligned. The old h1 label is replaced by the dimension label.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The dimension labels were 0.28 m, about 6 px on screen. They are 0.5 m now, so the mm values can be read.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The break lines ran across the members at a slant. They are now two short slanted strokes on the column stub and across the
pile group, and the pile label is moved below the break.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Plan and side use the same scale (22 px per m) and share their horizontal extent, so the footing lines up. The side view breaks
the column (3 m above the footing) and the piles (4 m below it) so its height stays near the plan's. The row is three columns
from 1100 px, the side view drops to its own row between 800 and 1100 px, and it stacks below 800 px.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The pile-count card (−/+ and the 20본 total) is on the left and the plan on the right, in one row; the side view is below.
Changing a count is now seen on the plan next to it. The layout stacks below 800 px.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The footing card shows a plan (footing rectangle, pile circles, column, pier axis, mm labels) and an elevation (footing,
piles to their tips, column, h1 and pile length). The drawings are SVG strings built from the model, so they follow the
inputs. θ is noted as not yet applied. Tests check the pile count, the mm labels and the width input.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The footing card takes L, W, h1 (default 2000), the blinding margin on each side (100) and thickness (100). The pile is picked
from the 강관 or PHC list by outer diameter and wall; the inner diameter, area and weight are read from the list. Defaults are
the 1단계 constants, so P1 is unchanged. The model reads them (footing width, height, blinding, pile diameter; the footing
length for grouping). 복합 말뚝 and the rule-pending items stay disabled; the footing rebar is not yet tied to these
inputs (1·2단계 후).
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Single-piece lengths such as the B9 and C7 pieces showed long unrounded decimals (2128.1174619702915). They are rounded to
0.1 mm for display and in the Excel rows; the totals are unchanged.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
D is an input on the column card (default 2000 mm). The 3D column, the piles of a cast-in pier and the column rebar follow it;
n_div stays 46. Three checks run on each regenerate and appear in 알림: D below the Level width (2100), D + 2c within the
coping width, and the 주철근 arc spacing ≥ max(1.5·d_b, 40). Spec §5.2 records the rule (approved 2026-10-23).
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The drawing has two "S(%)" labels; the first one in the file belonged to another table, so the solplate header was missing.
The header is now the S(%) label with the most solplate column names on its row. The full 운곡교 교각도(2) DXF reads all ten
solplate rows; a test adds a stray label.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The solplate's L (A + margin) runs across the 교각축 직각 and its W (B + margin) along the 교축; the top tapers across L with
T1 at the (−) side (T1 = T + L·|S|/200). The 교량받침 본체 is one box per bearing: the solplate plan less 100 mm on each
side, from the mortar top (EL.X) up h3 = 181 mm, turned with the bearing row. It is shown with the bearing group.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
"도면값 채우기" set only EL.X, h1 and h2, so no solplate was drawn. It now also sets A, B and the solplate slope S from the
reference values. The bearing card warns how many bearings lack A, B or S.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Each bearing gets a solplate wedge: tapered along the 교축 with T1 on the 시점 side, underside at the bearing top (block
top + h3). A, B come from the drawing's ELEVATION table, S and T1/T2 from the SOLPLATE table (NO, S(%), L, W, T, T1, T2),
read by position. The 공통 규칙 (T 30, margin 100, h3 181) feed the plan; changing them regenerates the piers. A new
"솔플레이트" part can be shown or hidden. A solplate table is read only when its S values are found; a missing S is reported.
Also commits the option-1 test change (drawing EL.X minus Excel z).
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Per bearing: L = A + margin along the 교축 (s), W = B + margin across it; the underside is the bearing top (block top
EL.X + h3); the top tapers from T1 = T + L·|S|/200 at the 시점 side (−s) to T2 = T − L·|S|/200 at the 종점 side, so
T1 + T2 = 2T and the volume is L·W·T. Corners follow the bearing row (rowSkew). Tests cover L, W, T1, T2, the underside,
the volume and the centring. The model and the 3D do not use it yet.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The drawing has other "EL.X" and "NO" labels. The header is now the EL.X label with the most table column names to its right
on the same row; the No header is the NO label on that row. The full 운곡교 교각도(2) DXF reads all ten bearings (EL.X, h1, h2)
without errors. Tests add the distractor labels.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The reader takes TEXT and MTEXT entities (x, y, text). The EL.X header row names the columns; the No column holds the
bearing numbers; each value is the nearest entry under its header on the same row. On the 운곡교 교각도(2) drawing all ten
bearings read back the reference values (EL.X, h1, h2). The 도면 불러오기 button accepts .dxf as well as sheets. The test
fixture holds only the table's text entries.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The sheet needs the header names No, EL.X, h1 and h2 (case, spaces, dots and units in parentheses are ignored). Rows are
checked: No must be a unique positive integer, EL.X must be in metres (30–100), h1 and h2 must be positive mm up to 1000.
Any error stops the import with the first messages; a valid sheet fills this pier's drawing values and regenerates the
pier. The parser is pure and tested.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The project has one basis: 엑셀 z (default) or 도면 EL.X. On ②, the bearing card lists EL.X, h1 and h2 for each bearing
(per pier), with a button to fill the reference values. With the drawing basis the model takes the drawing EL.X as the
bearing height; the coping tops stay within the 5 mm tolerance, the front block is 164 mm and the back 139 mm (excluding the
pad). The check compares EL.X − h1 − h2 with the Excel-basis coping top. The basis and the drawing values are saved in the
project file (optional fields). Changes regenerate every pier at once.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
With this definition the drawing's coping top matches the model's coping top to within 1 mm on all ten bearings. The earlier
0.211 m offset came from reading EL.X as the solplate top, which was wrong. The reference table now uses EL.X − h1 − h2,
and the check tests the match.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
On the bearing card, four inputs drive E = h1 + h2 + h3 + h4, L = A + margin, W = B + margin and T1/T2 = T ± L·|S|/2. The
defaults are the drawing values, so every row gives back the drawing values; changed rows are marked. The values are screen
state for now and do not change the model or the project file (Q16 stays open).
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The solplate tapers along the bridge axis. T is the centre thickness (30), T1 the thickness on the 교각축 직각 (−) side
(38) and T2 on the (+) side (22); T1 + T2 = 2T. The table's T column had shown the (−) side value, so T is now 30 on every
row. A test checks T1 + T2 = 2T.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The P1 상세도 ELEVATION TABLE and 솔플레이트 TABLE are shown on the bearing card. The model does not use them (Q16). The
ELEVATION table compares the drawing's coping top (EL.X − E) with the model's coping top; the difference is a constant
0.211 m on all ten bearings. The solplate table checks T1 and T2 against T ± L·|S|/2 with the 공통 규칙 T = 30 mm; all
rows match. Tests check E = h1+h2+h3+h4, L and W, the T1/T2 rule, and the constant offset.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The coping middle level is (Level underside + step top)/2. P3's top and P6's upper limit are this level, and e_top is the
step top minus it (P1: 1322, replacing the 1623 reverse-engineered value). On P1 P3 goes from 38 to 39 ties and P6 stays at
8 levels. With taper h2 = 3000 the middle level is 13275: P3 37 ties, P6 8 levels (32 bars), e_top 2072.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
e_P1 is no longer an input: P1 top = C3 (lowest coping top − 127 − 2·57.2 = 43.170) − 100 = 43.070, so e_P1 = step top −
43.070 = 485 mm. The P1 top moves from 15147 to 14862 mm; the top piece shortens by 285 mm. Couplers are placed at k·3000
with at least 2500 mm below the P1 top (a value to confirm). On P1 the coupler count stays at four (3, 6, 9, 12 m). P2 and
the tie top do not change.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Hc = Level underside − footing top, and h_cop = step top at the column centre − Level underside. P1 gives 12703 and 2644,
unchanged. With taper h2 = 3000 the Level underside moves down, so Hc = 11203 and the P2 top (Level underside + l_d)
moves to 13039, while the P1 top stays at the step top (15147). The 3D and the 철근표 use the same heights.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
C12 and C13 come from one grid from the Level underside + 127 at 250 mm, up to top − 400 mm (상면 여유). The fixed count of
seven rows is kept only for callers that give no top. P1 is unchanged (C12 5, C13 2); a taper of 3000 mm now gives C12 11
and C13 2, and the C13 first row is always the last C12 row plus 250.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The 250 grid runs from the Level underside + 127 for C12 and C13 together. The drawing's sixth C12 row at the bottom is the
bottom bar C7·C8 (코핑철근.md §6.1), so no bars are added and the counts stay the same.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The split between C12 and C13 follows the end underside, not a fixed level: C12 rows lie between the Level underside and
the end underside (taper height h2), C13 rows between the end underside and the top (end height h1). The break is the end
underside + 127 mm, the taper line's start at the end face. P1 is unchanged (C12 41.288–42.288, C13 42.538–42.788). Tests
check the band rule for h1 and h2 changes.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The three values are inputs on the coping card, with the P1 values (1000 · 1500 · 2100) when absent, and they are tracked
like the other shape inputs. The geometry and the coping rebar read them from one CopingGeom: the U heights follow the
closed form (h1 − 2c) + h2·min(x_s, x_L0)/x_L0, which reproduces the P1 values. The side rows are assumed to keep their
position relative to the Level underside (명세 확인 필요). P1 results are unchanged; the full suite passes.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The coping card lists d_IP2 and a table of the step segments: range, front and back coping tops, the lateral step, and
the block heights including the 50 mm pad (150 / 200 as in the 삽도). The table is read from the model; nothing new is
computed or entered. Tests check the P1 values against the 삽도.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The four panels are in one stack on the right of the 3D view, 180 px wide, with the ⓘ origin note opening to its left.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The right stack is gone. 색 기준 and the ⓘ origin note now sit under 시점 in the left stack, so every 3D info panel is on the
left. The stack is 180 px wide; the origin note opens to the right of its button.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
On wide screens the pier name moves to the 교각 목록 (name · type), the band is gone and the 3D view takes the full height.
A left stack holds 부재 표시 and then 시점; a right stack holds 색 기준 and then the ⓘ origin note. The active pier is scrolled
into view when it changes. Below 600 px the title is kept above the model and the stacks sit below it.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The title stays. The view buttons become a menu that shows the current view and closes on a choice, Esc or an outside
click. The origin note is an ⓘ toggle. 부재 표시 and 색 기준 are collapsed by default and show their state in the summary
(e.g. 부재 표시 7/7 ▾, 색 기준: 역할 ▾); their open state is remembered per viewer, with guarded storage access.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Between 601 and 1100 px the title stays on the left and the view buttons and origin note sit on the right. Wider windows
keep the current order; below 600 px the band wraps as before.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The title, view buttons and origin sit in one 44 px band on the model. 부재 표시 is a collapsible panel on the left and
색 기준 (③ only) a collapsible panel on the right, both over the canvas, about 170 px wide, with 44 px touch targets.
Below 600 px the band wraps and the panels fall below the model.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
On wide screens the page is one fixed-height frame: the left list, the centre (model band on top, table below) and the
right results each scroll on their own. The model band is 280–560 px (48vh) and the 3D view follows its container
through a ResizeObserver. The table header stays on top. Below 1100 px the panels stack with a 60vh height limit.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Each block takes the skew of its bearing row against the perpendicular of its stretch axis (rowSkewDeg, from the bearing
azimuth). The blocks and pads are turned to that skew, and the S1/S2 bars are drawn in the block frame and turned with
it, so their lengths and embedments do not change. ② shows the row skew (P1: front −4.42°, back −3.11°). Approved
2026-10-23 with the drawing as reference.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The status line and the backup banner share one row (the banner appears only when a backup exists). Results and notices
show as toasts that close after 4 s; problems stay until closed. The status line keeps only the empty-state hint.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
A row can now carry pieces (length, count). C9, C10, C11 and C9·C11 사이 have U and 삿갓 pieces per position (3 loops each);
C12 and C13 have 일자 and ㄷ pieces per bar; P1 and P2 have their splice pieces, each counting all the bars. The
재료표 totals from the pieces, so no bar is left out. 개수 stays the 철근표 count; the piece count is kept beside it. The
삿갓 length h − 100 is an assumption (Q-K5) and is noted on the ④ screen.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The 철근재료표 moves from ③ to a new stage ④ in the header. It shows the checked piers (or the current one), one table
per pier with 강종 → 직경 → 번호 and subtotals, and a total across piers. The Excel export has one sheet per pier and a
합계 sheet. Print uses A4 landscape without the header and buttons. ③ keeps a summary line and a link to ④.
Spec §5, §6.9 and §7 record the stage (approved 2026-10-23).
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
The quantity table sits below the 철근표 for the whole pier. Lines group by grade and nominal diameter, with
subtotals, unit weight 0.00617·d² (kg/m), the total in kg and ton, and a 할증 total. 할증 is an input in 공통 설정,
default 0 as an assumption. Couplers are counted, not weighed. Bars with no length yet (C9, C10, C11, C12, P1, P2) are
listed as missing and kept out of the totals. The table exports to Excel with SheetJS.
Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
