3.4.0

Highlights

• Add *size functions, train and get/set_learning_rate to phi.flow convenience import
• Add join_meshes function
• Add margin_cells to enclosing_grid and support point Tensor inputs
• Implement farthest_points algorithm
• Improved triangle mesh SDF
• Add spline utilities
• Refactor most geometries as dataclasses

pip install phiflow==3.4.0

3.3.0

This version features tons of small improvements to phi.geom and bug fixes. See #189

Highlights

• Support for 3D and periodic unstructured meshes
• Improvements to 3D surface mesh rendering in Plotly
• Compatibility with Python 3.8
• Many common functions (*min, *max, *prod *pack, meshgrid, *cat, p2d) are now available via the convenience import.
• Added squared_length, length now reduces only vector

Breaking changes

• Accessing non-existent dims on geometry or Field now raises an error immediately

pip install phiflow==3.3.0

3.2.0

Highlights

• Field now supports setting variable/value attributes
• New Geometry type: Cylinder
• SDFGrid can now store surface normals
• Add read_stl()
• Unstructured mesh faces now build faster
• Many common functions have been added to the convenience import phi.flow
• Vector-functions moved from PhiML to phi.geom
• Improved plots

Plus tons of bug fixes! See #182

pip install phiflow==3.2.0

3.1.0

3D plotting + triangle meshes #177

Highlights

• Improved Plotly 3D plots and added support for animated plots
• Updated Mesh to handle 3D triangle meshes, support mesh decimation
• Geometry now supports named sets of sample points
• Numerous additions to phi.geom, such as geometry intersections, line tracing, marching cubes, signed distance fields

3.0.1

Bug fixes for 3.0. See #169

3.0.0

Version 3.0 marks a milestone release for Φ-Flow, introducing many new features and simplifying the API. #165

Highlights

• Support for unstructured meshes. This includes many field and physics operations, allowing grid simulations to be ported to FVM with little effort. Meshes can be loaded from .su2 and .gmsh files.
• Major plotting improvements: new plot types, such as bar charts, histograms, streamlines, points clouds with cmap. More flexible arguments, e.g. error bars, color, alpha, same_scale.
• Improved documentation: The GitHub page now lists a collection of examples in the form of Jupyter notebooks.
• Sparse neighborhood search using GPU-enabled hash grids. This enables simulations with interacting particles, such as SPH.
• All linear solves can now use the ILU preconditioner.
• The phi.math package is now stand-along as the phiml library.
• All types of fields have been merge into the Field class which makes a lot of functionality more easily accessible. The legacy constructors still work but now return Field instances.
• Boundaries are now easier to define, e.g. {'x-': 0, 'x+': 1}
• Support for geometries defined by SDF grids.

2.5.3

See #146

2.5.1

See PR #141

pip install phiflow==2.5.1

2.5.0

See #135

pip install phiflow==2.5.0

2.4.0

See #116