AVS-UCD#

.avs read + write eager

Summary of the specification#

AVS Unstructured Cell Data begins with a five-integer counts line: numbers of nodes, cells, node data values, cell data values and model data values. Node records follow - an id and three coordinates each - then one line per cell giving its id, material id, a cell type keyword (tri, quad, tet, hex, prism, pyr, line, pt) and its node ids. Optional data sections at the end declare component counts and labels before the per-node or per-cell values.

Specification at a glance#

counts line	nnodes ncells ndata_node ndata_cell ndata_model
node line	id x y z
cell line	id mat_id type n1 n2 …
cell types	pt line tri quad tet pyr prism hex
data sections	component counts, then labelled value blocks

Read the full AVS-UCD specification ↗

Reading#

import polyxios as px

mesh = px.read("model.avs")
mesh.vertices          # (n, 3)
mesh.element_types     # one type code per element

Writing#

px.write(mesh, "out.avs")

This codec takes no format-specific options.

Quirks worth knowing#

The per-cell material id is read into element_attrs["mat_id"] and written back from there, so material groups survive a round trip; a mesh without one is written with material 1 throughout.
Only the geometry is read: the node, cell and model data sections are skipped, and the writer declares none.
Node ids are remapped to 0-based indices and the originals are not kept; the writer numbers the nodes from 1 again.
pt cells are not read: a file holding one is refused with CodecError, as is any other cell type outside line, tri, quad, tet, pyr, prism and hex. On write, an element with no AVS type - a vertex, a polygon, a quadratic cell - is skipped with a warning, and only the vertices the written cells use go out.