Book Trolley

A library book truck: two arched end panels standing in skids on four swivel casters, a flat bottom deck, and on each side two sloped troughs under a turned push rail. Each trough is a shelf tilted 14° back toward the centre and a backrest square to it, both housed in the end panels. Forty-seven hardbacks stand in the troughs spines out, leaning back into the corner; five more lie in two loose stacks on the deck. Every book is a rounded-spine cover in cloth or leather, a page block glued into it, and two gilt bands tooled round the spine. A showcase piece, not an example — it witnesses no API contract. It asserts that generated geometry meets declared asset budgets, recomputed from the finished mesh.

What it composes#

The budget that matters#

A book on a flat shelf stands on one support. A book in a sloped trough is carried at two: its tail on the shelf and its fore-edge against the backrest. Seated on the shelf and standing off the backrest, it would topple forward on a real trolley, and nothing else sees it:

• the book still sits its seat into the shelf, so the seat budget passes;
• it stays inside the carcass, so the bounding box does not move;
• the triangle count is the same, because only a position changed;
• the book-overlap budget has to exempt both designed contacts, the shelf and the backrest, so it cannot see how deep either goes.

The piece pairs each tilted shelf with the one backrest whose shell overlaps it, finds each standing book's shelf by the top plane its lowest vertex sits nearest (read off the mesh, on the book's own side of the trolley), and takes the plane of that backrest's face that looks at the book. It asserts how far the book's deepest cover vertex sits behind that plane, as a band of 0.3–2.0 mm, next to the shelf seat's own band. All forty-seven sit at 0.80 mm.

A plane, not the backrest solid, for the same reason as bookshelf's lean contact: a fore-edge driven through the 18 mm board would come out behind it and read as no contact.

--off-back and --deep-back are the falsifiers built for exactly this. One stands every trough book 3 mm off its backrest, the other drives it 4.5 mm in, both along the slope so the shelf seat is untouched. Both exit 21 while every seat, the envelope and the book-overlap budget still pass.

Budgets#

Declared in the script as named constants, recomputed from the generated mesh. Measured values are from Blender 5.2.1; see the cross-version table below.

Real-world size: 0.77 m long over the skids, 0.53 m deep and 1.00 m tall over the arched ends, on 70 mm casters. Each trough is 170 mm deep and its backrest 170 mm tall; the books run from a 150 mm pocket book to a 245 mm octavo.

Construction#

• End panels and skids. Each end is one prism: a flat-sided outline whose top is a 16-segment arc, extruded 22 mm and chamfered 2.5 mm, standing 12 mm deep in a 70 × 35 mm skid that runs 15 mm past it front and back. The panels are the carcass's only bearers; the casters are the named supports.
• Cross members. The deck and the four tilted shelves and backrests are each housed DADO (8 mm) into both ends and stop 14 mm short of the outer faces. The push rail is a 16-sided dowel whose ends sit 12 mm blind in both panels, 10 mm short of the outer faces.
• Troughs. Built in a trough frame: s runs down the slope into the backrest, t up the shelf normal, the corner at the origin. The shelf runs from 170 mm in front of the corner to 6 mm past the backrest's back face, so the two backs never share a plane; the backrest stands 6 mm into the shelf. The two sides are mirror images, and the two tiers of a side stand 320 mm apart, so each tier's books clear the shelf above.
• Casters. Stacked by a named bite at each joint: a 56 mm plate 1 mm into the skid, a swivel race 1 mm into the plate, a fork (one U-section prism, chamfered 0.8 mm) 1 mm into the race, an axle through both fork legs and a 70 mm rubber wheel whose lowest vertex is the floor. The wheel's rim is rounded with a two-segment bevel and shaded smooth; the iron is near-black with a little rust in the noise, not chrome.
• Books. Copied from bookshelf: one U-section cover per book with a rounded spine, a page block 0.6 mm into both boards, two gilt bands on the spine's own chain.
• Trough books. Built in their own frame, turned so the height runs up the shelf normal and the fore-edge down the slope, and yawed about the shelf normal in a four-step cycle (−2.4°, −0.8°, 0.8°, 2.4°) so no two neighbours share a plane. Each is then slid up the normal to a stepped 1.0–3.2 mm seat into the shelf, down the slope until its deepest vertex is BACK_BITE (0.8 mm) into the backrest, and along the trough with a 2.5–4.5 mm gap. A trough that overflows its span is a build error.
• Loose stacks. Books laid flat on the deck, spine out, yawed ±3° alternately and shifted a few millimetres; each book above the first takes its own 1.0–1.4 mm seat into the board below it.

Findings the budgets forced#

• A 45° plane shared by accident. Dropping the caster plates 2 mm for the plate-seat falsifier put each plate's lower chamfer and the fork's upper chamfer on the same 45° plane — the plate's half-width minus the fork's is 13 mm, and so was the drop between them — and the run exited 15 on eight coplanar pairs instead of 18. The drop is 2.5 mm.
• An edge falsifier the mirror budget saw first. Leaving one front backrest square broke the carcass's front/back symmetry, so the run exited 19 before the edge budget. The falsifier now leaves the deck square: one centred member, symmetric in both axes, 12 right-angle edges.
• The bake cage reached the books. At 10 mm the cage from each end panel's inner face reached the high-poly covers 6 mm away, and the books' edges baked onto the panel as pale streaks. Cutting the cage to 3 mm cleared the panels but not the deck: the loose stacks sit 1.0–1.4 mm into it, so rays from round each stack still found the spines and baked every footprint in as a dished groove. Only the wood takes a baked map, so the high-poly bake source is now the carcass and casters alone; the cage stays at 3 mm.

Conventions walked#

Every convention in showcase/README.md, and whether it applies here.

Falsifiers#

Each breaks one pipeline stage so a named budget fails. All fifteen were run on 4.5.11, 5.1.2 and 5.2.1 and exited the same declared code on all three.

Exit codes#

File-local and sequential. 9 is a valid check code. 1 is the FATAL wrapper — a crash, never a named check.

Run it#

# Budget check, no render.
blender --background --python book_trolley.py --

# Falsifier: every trough book stands off its backrest. Must exit 21.
blender --background --python book_trolley.py -- --off-back

# Falsifier: every trough book cuts into its backrest. Must exit 21.
blender --background --python book_trolley.py -- --deep-back

# Render the gallery still (EEVEE; --engine cycles on a GPU-less host).
blender --background --python book_trolley.py -- --output book-trolley.webp

Smoke runs the check-only path. It does not pass --output or any falsifier.

Cross-version measurements#

The glTF size differs by 8 bytes on 5.2 (exporter metadata, not a budget). Every other value, including the DECIMATE COLLAPSE LOD counts, is identical on all three.