Overview¶

fVTKHDF is a modern Fortran library for writing VTKHDF files readable by VTK and ParaView. VTKHDF is an HDF5-based file format designed for scalable scientific data storage and visualization.

fVTKHDF is designed primarily for MPI-parallel high-performance computing (HPC) applications. A serial build is also available, preserving the same high-level API.

Dataset Types and Writers¶

The VTKHDF format distinguishes simple VTK dataset types, such as ImageData and UnstructuredGrid, from composite types, such as MultiBlockDataSet and PartitionedDataSetCollection. Composite types are containers for simple VTK datasets. Each simple VTK dataset has its own mesh and associated point, cell, and field data.

Currently supported VTK dataset types:

ImageData, RectilinearGrid, and StructuredGrid: logically rectangular grids. ImageData uses uniform spacing; RectilinearGrid is defined by the tensor product of three coordinate vectors; StructuredGrid gives each point explicit physical coordinates. RectilinearGrid and StructuredGrid also support moving geometry with fixed logical grids.
UnstructuredGrid (UG): meshes with arbitrary cell connectivity, supporting moving geometry and changing topology.
MultiBlockDataSet (MB): a composite type containing a flat collection of UnstructuredGrid blocks.

fVTKHDF provides a file writer for each supported VTK dataset type. Each writer is a Fortran derived type exported by its corresponding module. Writers for simple VTK datasets provide methods for defining the mesh and writing point, cell, and field data. Composite writers manage collections of simple VTK datasets.

VTK dataset type	File writer type	Module
ImageData	`vtkhdf_image_file`	`vtkhdf_image_file_type`
RectilinearGrid	`vtkhdf_rectilinear_file`	`vtkhdf_rectilinear_file_type`
StructuredGrid	`vtkhdf_structured_file`	`vtkhdf_structured_file_type`
UnstructuredGrid	`vtkhdf_ug_file`	`vtkhdf_ug_file_type`
MultiBlockDataSet	`vtkhdf_mb_file`	`vtkhdf_mb_file_type`

The vtkhdf_vtk_cell_types module additionally provides VTK cell type integer codes relevant to UnstructuredGrid.

Data and Writer Modes¶

For simple VTK datasets, point and cell data associate scalar or multicomponent values with mesh entities; field data associates values with the VTK dataset as a whole. Their writers share conventions for static writes by name and temporal registration followed by writes through handles. Field data uses a common interface and array interpretation across simple VTK dataset types. Structured writers additionally support rectangular region placement and incremental point/cell writes.

Writers for simple VTK datasets support subsets of the shared modes VTKHDF_STATIC, VTKHDF_FIXED_MESH, VTKHDF_MOVING_MESH, and VTKHDF_DYNAMIC_MESH. The last three modes produce temporal VTK datasets. See Common Conventions for mode definitions and shared conventions, and Field Data for the field data interface.

Reader Compatibility¶

This library targets VTKHDF Version 2.8. Unless otherwise indicated, reader support targets ParaView 5.13 and later. Older releases may work, but compatibility is not guaranteed. Additional version requirements and known reader limitations are described in the individual VTK dataset chapters.

Parallel and Serial I/O¶

In MPI builds, parallel HDF5 uses MPI-IO to write mesh and data collectively to a single shared file. The library manages HDF5 operations and MPI coordination internally. Writer methods are collective over the communicator supplied at file creation; all participating ranks must make the same sequence of calls. See Common Conventions for the collective-call requirements.

A serial build uses a standard HDF5 installation without MPI. The API is the same, except that file creation omits the communicator argument.

Data Placement¶

UnstructuredGrid stores each MPI rank’s mesh contribution as a separate partition, with its own points and cell connectivity. Each structured writer assembles caller-owned rectangular regions into one global structured VTK dataset without recording rank partitions. The library performs no repartitioning or data redistribution.