AFLR UGRID (.ugrid)

The AFLR UGRID format: a binary/ASCII surface+volume mesh whose byte layout is encoded in the penultimate filename suffix (e.g. foo.lb8.ugrid → flavour lb8).

Format name	ugrid
Extensions	.ugrid
Read / Write	✓ / ✓
Extra dependencies	—

Reading & writing

import meshioplusplus

mesh = meshioplusplus.read("sphere.b8.ugrid")   # flavour taken from the suffix
meshioplusplus.ugrid.write("out.lb8.ugrid", mesh)

No kwargs — the flavour is entirely inferred from the filename.

File-flavour table

suffix	kind	float dtype	int dtype
(none)	ascii	native	native
b8l	C, big-endian	>f8	>i8
b8	C, big-endian	>f8	>i4
b4	C, big-endian	>f4	>i4
lb8l	C, little-endian	<f8	<i8
lb8	C, little-endian	<f8	<i4
lb4	C, little-endian	<f4	<i4
r8	Fortran-record, big-endian	>f8	>i4
r4	Fortran-record, big-endian	>f4	>i4
lr8	Fortran-record, little-endian	<f8	<i4
lr4	Fortran-record, little-endian	<f4	<i4

"C" flavours have no extra framing; "Fortran-record" flavours (r*/lr*) wrap the whole body in Fortran unformatted-record markers: a leading integer byte-count, the payload, then a trailing integer repeating the same byte-count. There are exactly 2 such records total: record 1 is the 7-integer header; record 2 is everything else (points, connectivity, boundary tags) combined.

File structure

Header (always 7 integers): num_points, num_triangle, num_quad, num_tetra, num_pyramid, num_wedge, num_hexahedron. Body, in fixed order:

1. Points: num_points × 3 floats.
2. Triangle connectivity (1-based, -1'd on read) — only if num_triangle > 0.
3. Quad connectivity (1-based) — only if num_quad > 0.
4. Triangle boundary tags — only if num_triangle > 0.
5. Quad boundary tags — only if num_quad > 0.
6. Tetra connectivity.
7. Pyramid connectivity — reordered via [1,0,3,4,2] on read ([1,0,4,2,3] on write) — see below.
8. Wedge connectivity.
9. Hexahedron connectivity.

Volume elements (6-9) get zero-filled boundary tags — UGRID has no per-volume-element tag concept, so meshio++ synthesizes zeros for uniformity with the surface tags.

Cell types

triangle, quad (surface); tetra, pyramid, wedge, hexahedron (volume). Node ordering matches meshio++'s convention for every type except pyramids, which need the permutation above — this is the one place in the format where getting the order wrong would silently produce inverted-volume elements rather than an outright error, so it's specifically covered by a signed-volume regression test.

Data mapping

• cell_data["ugrid:ref"] — one array per written cell block, in the fixed type order (triangle, quad get real boundary tags; tetra/pyramid/wedge/ hexahedron get all-zero tags).

Quirks & limitations

• The writer enforces at most one cell block per known type (ValueError otherwise) and skips unknown types with a warning.
• Fortran-record byte-count markers are written but not validated on re-read — a corrupted record-length byte wouldn't be caught by the reader.
• Boundary tags on write come from the first integer cell_data array found (warns if more than one candidate exists); defaults to all-1 if none.
• 4-byte-float variants (b4/lb4/r4/lr4) inherently have lower round-trip precision than the 8-byte variants — reflected in the test suite's looser tolerance for those flavours.

Notes

• tests/meshes/ugrid/pyra_cube.ugrid (ascii) — a unit cube split into 6 pyramids; used to verify the pyramid permutation via a signed-volume sum (≈1.0).
• tests/meshes/ugrid/sphere_mixed.1.lb8.ugrid (little-endian binary) — 3270 points, 864 triangles, 3024 wedges, 9072 tets, boundary tag counts {1:432, 2:216, 3:216}.
• tests/meshes/ugrid/hch_strct.4.lb8.ugrid — 306 points, 12 tri, 178 quad, 96 wedge, 144 hex; also used to check total surface area, catching any connectivity-order regression.
• Fully handled by the C++ core across every flavour (host-relative byte swapping).