PLOT3D — multi-block structured grids (.p3d, .x, .g)
The NASA PLOT3D format: multi-block structured grids, with .q solution and .f function files, what OVERFLOW, CFL3D and the NASA Turbulence Modeling Resource exchange (roadmap 2.4.2). Both engines read and write it (formats/plot3d.cpp, meshioplusplus.plot3d).
| Format name | plot3d |
| Extensions | .p3d, .p3dfmt, .x, .g (.xyz stays the point cloud xyz; pass file_format="plot3d") |
| Read / Write | ✓ / ✓ (lattice blocks only; see conformance) |
| Extra dependencies | — |
Reading
import meshioplusplus
grid = meshioplusplus.read("wing.x")
grid = meshioplusplus.plot3d.read("wing.x", solution="wing.q", function="wing.f")
grid = meshioplusplus.plot3d.read("odd.x", precision="single", multiblock=False)A PLOT3D file does not say which variant it is, so the reader infers it: ASCII, Fortran sequential unformatted (record markers) or raw binary (a C stream); single or multi-block (a leading block count); 2-D (ni nj, x y) or 3-D (ni nj nk, x y z); single or double precision; with or without an iblank array; either byte order. Every combination is checked against the file's exact size (a Fortran file's record sizes). When the size fits several readings, the one whose coordinates are finite and of sane magnitude (no subnormals, which a misread precision produces) wins. A file that still reads several ways is a ReadError naming each reading; the read options precision ("single"/"double"), multiblock, iblank and dimensions (2/3) pick one (Plot3dReadOptions in C++).
- Each grid block is one cell block of hexahedra (quads in 2-D, lines when only one dimension exceeds one), its points in PLOT3D's order (
ifastest) after the previous block's, and a cell regionblock_<b>. - The block dimensions are
field_data["plot3d:dims"]((nblocks, 3),nk = 1in 2-D); a 2-D grid hasz = 0. iblankis the Int32 point dataplot3d:iblank(overset grids use negative values, so it is not auint8).solution=reads a.qfile ([nblocks] dims, then per block the four reference values and thedim + 2conserved variables):density,momentum(a 3-vector,0forzin 2-D) andenergypoint data, and the reference values (Mach number, angle of attack, Reynolds number, time) per block infield_data["plot3d:q_reference"]. Its encoding is inferred the same way and its dimensions must match the grid's.function=reads a.ffile ([nblocks], per block its dimensions andnvar, then the values) asplot3d:f,(npoints, nvar).
Writing
meshioplusplus.write("out.x", grid)The writer emits the Fortran unformatted multi-block double-precision 3-D form (little-endian, 4-byte record markers), with iblank when the mesh carries plot3d:iblank. It needs field_data["plot3d:dims"] whose blocks hold exactly the mesh's points, and refuses any other mesh by name: a general mesh is not a set of lattices. A grid read from that form writes back byte-equal; a 2-D grid is written as 3-D with nk = 1. Solutions and function files are not written.
Verification
The variants are generated in the tests (tests/python/test_plot3d.py, tests/cpp/test_plot3d.cpp) from the layouts above, in both engines; no NASA-written file has been checked yet (the TMR grids are the planned fixture).