MMG Parameters Reference¶

import pyvista as pv
import mmgpy  # noqa: F401  -- registers reader/writer + .mmg accessor
mesh = pv.read("input.mesh")

Complete reference for the keyword arguments accepted by dataset.mmg.remesh(...) (and its variants remesh_optimize, remesh_uniform, remesh_levelset, move). Each call returns a fresh PyVista dataset; the input is not mutated.

Native parameter files¶

All three engines accept MMG's native local-parameter files through the parameter_file keyword. Pass a .mmg3d file for tetrahedral meshes, a .mmg2d file for planar meshes, or a .mmgs file for surface meshes:

from pathlib import Path

parameter_file = Path("local.mmg3d")
remeshed = mesh.mmg.remesh(
    parameter_file=parameter_file,
    verbose=-1,
)

The same keyword is available on Mesh.remesh(...), Mesh.remesh_levelset(...), and the file APIs such as mmgpy.mmg3d.remesh("in.mesh", "out.mesh", parameter_file=parameter_file). Both strings and path-like objects are accepted. Missing or unreadable files raise a RuntimeError before remeshing starts.

mmgpy uses the same parser for all three engines and applies the parsed values through each engine's public MMG setters.

MMG parameter files define local sizing by entity reference. For example, this .mmg3d file targets tetrahedra whose reference is 7:

parameters
1
7 tetrahedra 0.02 0.08 0.01

The final three values are hmin, hmax, and hausd. MMG3D accepts triangle(s) and tetrahedron(s) references, MMG2D accepts edge(s) and triangle(s), and MMGS accepts triangle(s). Native lsreferences and lsbasereferences sections are supported as well.

Parameter files are parsed first. Explicit Python options are applied second, so Python settings take precedence when both configure the same MMG parameter.

Size Parameters¶

hmin¶

Minimum edge length.

Property	Value
Type	`float`
Default	Auto-computed
Range	> 0

remeshed = mesh.mmg.remesh(hmin=0.01)

Edges shorter than hmin will be collapsed or lengthened.

hmax¶

Maximum edge length.

Property	Value
Type	`float`
Default	Auto-computed
Range	> hmin

remeshed = mesh.mmg.remesh(hmax=0.1)

Edges longer than hmax will be split.

hsiz¶

Uniform target edge size.

Property	Value
Type	`float`
Default	None
Range	> 0

remeshed = mesh.mmg.remesh(hsiz=0.05)

When set, overrides hmin and hmax with uniform sizing.

hgrad¶

Gradation parameter controlling size transition.

Property	Value
Type	`float`
Default	1.3
Range	>= 1.0

remeshed = mesh.mmg.remesh(hgrad=1.2)

hgrad=1.0: No size variation allowed
hgrad=1.3: Default, allows 30% size change between adjacent edges
hgrad=2.0: Large size variations allowed

Geometric Parameters¶

hausd¶

Hausdorff distance, maximum distance between input and output geometry.

Property	Value
Type	`float`
Default	0.01 * bounding box diagonal
Range	> 0

remeshed = mesh.mmg.remesh(hausd=0.001)

Smaller values produce a closer geometric approximation at the cost of more elements.

ar¶

Ridge detection angle (degrees).

Property	Value
Type	`float`
Default	45.0
Range	0 - 180

remeshed = mesh.mmg.remesh(ar=30)

Edges with dihedral angle greater than ar are treated as ridges and preserved during remeshing.
ar=180 effectively disables ridge detection. The typed detect_ridges=False option is the explicit equivalent of CLI -nr.

detect_ridges¶

Enable ridge detection in typed options. None leaves MMG's default unchanged; False maps to MMG*_IPARAM_angle=0 (CLI -nr).

from mmgpy import Mmg3DOptions

remeshed = mesh.mmg.remesh(Mmg3DOptions(detect_ridges=False, hmax=0.1))

Control Flags¶

surface_only¶

Restrict level-set discretization to the mesh boundary. This keyword is available only on remesh_levelset(...).

Property	Value
Type	`bool`
Default	`False`
MMG mapping	`-lssurf`; `MMG*_IPARAM_isosurf = 1`

boundary_split = mesh.mmg.remesh_levelset(levelset, surface_only=True)

For MMG3D this splits boundary faces without splitting the interior volume into level-set regions. MMG2D and MMGS apply the mode to boundary or referenced edges.

optim¶

Enable optimization mode (no topology changes).

Property	Value
Type	`int`
Default	0
Values	0 (off), 1 (on)

remeshed = mesh.mmg.remesh(optim=1)

When enabled, only moves vertices to improve quality.

noinsert¶

Disable vertex insertion.

Property	Value
Type	`int`
Default	0
Values	0 (off), 1 (on)

remeshed = mesh.mmg.remesh(noinsert=1)

Prevents adding new vertices during remeshing.

noswap¶

Disable edge/face swapping.

remeshed = mesh.mmg.remesh(noswap=1)

Prevents topology changes via edge/face swaps.

nomove¶

Disable vertex movement.

remeshed = mesh.mmg.remesh(nomove=1)

Keeps vertices at their original positions.

nosurf¶

Preserve surface vertices.

remeshed = mesh.mmg.remesh(nosurf=1)

Prevents modification of surface mesh vertices.

Output Control¶

verbose¶

Verbosity level.

Property	Value
Type	`int`
Default	1
Range	-1 to 10

silent = mesh.mmg.remesh(verbose=-1)  # Silent
errors = mesh.mmg.remesh(verbose=0)   # Errors only
info = mesh.mmg.remesh(verbose=1)     # Standard info
debug = mesh.mmg.remesh(verbose=5)    # Debug output

Common Combinations¶

Quality optimization only¶

optimized = mesh.mmg.remesh(optim=1, noinsert=1)

Or use the convenience method:

optimized = mesh.mmg.remesh_optimize()

Uniform remeshing¶

uniform = mesh.mmg.remesh(hsiz=0.05)

Or use the convenience method:

uniform = mesh.mmg.remesh_uniform(size=0.05)

High-quality surface approximation¶

remeshed = mesh.mmg.remesh(
    hmax=0.1,
    hausd=0.0001,  # Tight geometric tolerance
    hgrad=1.1,     # Smooth size transition
)

Preserve sharp features¶

remeshed = mesh.mmg.remesh(
    hmax=0.1,
    ar=20,         # Detect more ridges
    hausd=0.001,
)

Fast coarse remeshing¶

remeshed = mesh.mmg.remesh(
    hmax=0.5,
    hgrad=2.0,     # Allow large size variations
    verbose=-1,
)

Volume interior only¶

remeshed = mesh.mmg.remesh(
    hmax=0.1,
    nosurf=1,      # Keep surface fixed
)

Parameter Interactions¶

Parameters	Effect
`optim=1, noinsert=1`	Quality optimization only
`hmin=hmax`	Near-uniform sizing
`hausd` small + `hmax` large	More elements on surface
`ar=180`	No ridge preservation
`hgrad=1.0`	No size gradation

Best Practices¶

Start with defaults, MMG auto-computes reasonable values.
Set hmax first, it is the most important parameter.
Add hausd for surfaces, it controls geometric fidelity.
Tune hgrad, lower values give smoother transitions.
Use verbose=-1 for batch processing.
Validate results with dataset.mmg.validate(detailed=True) after remeshing.