diff --git a/docs/numpy_field_manipulation.rst b/docs/numpy_field_manipulation.rst new file mode 100644 index 0000000..e4d681e --- /dev/null +++ b/docs/numpy_field_manipulation.rst @@ -0,0 +1,156 @@ +NumPy Field Manipulation +================================== + +NumPy-backed fields are a **new feature in CompuCell3D 4.7.0** that allows direct access to scalar or vector field data as NumPy arrays. +In comparison to traditional field manipulation, using NumPy offers easier array manipulation syntax and efficient computational performance. +The array data is shared memory between your Python code and CompuCell3D's C++ backend. + +The demo **shared_numpy_fields** shows this feature in action. + +Creating Shared NumPy Fields +############################################### + +**padding**: adds extra layers around the field. Useful for PDE solvers. + +**precision_type**: a NumPy dtype such as ``"int16"`` or ``"float32"``. The following NumPy data types are supported: + +- ``int8`` (signed char) +- ``uint8`` (unsigned char) +- ``int16`` (short) +- ``uint16`` (unsigned short +- ``int32`` (int) +- ``uint32`` (unsigned int) +- ``int64`` (long long) +- ``uint64`` (unsigned long long) +- ``float32`` (float) +- ``float64`` (double) + +You can create **scalar fields** using: + +.. code-block:: python + + self.create_shared_scalar_numpy_field("exampleScalarNumpy", padding=1) + +.. code-block:: python + + self.create_shared_scalar_numpy_field("exampleInt16FieldPythonNPY", precision_type="int16") + +.. code-block:: python + + self.create_shared_scalar_numpy_field("exampleFloat32FieldPythonNPY", precision_type="float32") + +You can create **vector fields** using: + +.. code-block:: python + + self.create_shared_vector_numpy_field("exampleVectorNumpy") + +.. code-block:: python + + self.create_shared_vector_numpy_field("exampleVectorNumpy", precision_type="float32") + +Accessing Fields +############################################### + +Once created, these fields can be accessed directly: + +.. code-block:: python + + exampleScalarNumpy = self.field.exampleScalarNumpy + exampleScalarNumpy[15, 25, 0] = 5 # Assign scalar + +.. code-block:: python + + exampleVectorNumpy = self.field.exampleVectorNumpy + exampleVectorNumpy[15, 25, 0] = [15, 25, 0] # Assign vector + +Or, to access the field and its padding, use ``raw_field``: + +.. code-block:: python + + cpp_array = self.raw_field.exampleScalar # Includes padding + + cpp_array_user = self.field.exampleScalar # Excludes padding + +********************************************** + +Using Field Manager XML Helper +############################################### + +NumPy fields can be created in XML too. This is just a different way of writing the syntax. + +**Name**: the unique name that will be used to identify the field in XML and Python. + +**Type**: ``"scalar"`` (AKA ``"concentration"``) or ``"vector"``. + +**Precision**: a NumPy dtype such as ``"int16"`` or ``"float32"``. + +**Example**: + +.. code-block:: XML + + + + + + + + +They can then be accessed from Python as before: + +.. code-block:: python + + fibers_fm = self.field.fibers_field_manager + fibers_fm[15, 25, 0, ...] = [120, 120, 0] + +********************************************** + +Key Differences from Standard Field API +--------------------------------------- + +.. raw:: html + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
FeatureNumPy Field API
Access SyntaxNumPy-style slicing: field[x1:x2, y1:y2, z1]
PerformanceSignificantly faster for bulk operations using NumPy
Data TypesMust specify dtype (e.g., int16, float32)
PaddingUse raw_field to access padded data regions
In-place EditingAllows slicing and assignment directly
+ +********************************************** + +Example Use Case: Copying Data from Standard Fields +---------------------------------------------------- + +The demo **shared_numpy_fields_steppables.py** shows that it's possible to edit standard field data into NumPy-backed fields using helper methods: + +.. code-block:: python + + self.copy_cell_attribute_field_values_to("cell_type_field", "type") + self.copy_cell_attribute_field_values_to("cell_volume_field", "id") + +These populate the NumPy fields with values derived from cell attributes. However, if your workflow requires you to transfer data from NumPy arrays to cells, you should write code to do this after `copy_cell_attribute_field_values_to`. diff --git a/docs/vector_field_polarization_plugin.rst b/docs/vector_field_polarization_plugin.rst new file mode 100644 index 0000000..3f2a9fc --- /dev/null +++ b/docs/vector_field_polarization_plugin.rst @@ -0,0 +1,77 @@ +Vector Field Polarization plugin +================================== + +**VectorFieldPolarization** is a CompuCell3D plugin that introduces a custom energy term to bias cellular motion based on alignment with a spatially defined vector field. +This is useful for scenarios where cells follow physical or chemical directional cues, such as fiber alignment in extracellular matrices (i.e. ECM fibers). + +When enabled, the plugin favors cell movement along the local direction specified by the vector field, +thereby increasing the likelihood of copy attempts that align with the vector direction. +Unlike the Chemotaxis plugin, this plugin uses a vector field that exists independently as opposed to a chemical field that cells may interact with. + +Related Demo: ``Demos/PluginDemos/VectorFieldPolarization/vector_field_polarization.cc3d`` + +------------------------ +XML Syntax +------------------------ + +This will register a vector field named ``Fibers``. +CompuCell3D can only support one VectorFieldPolarization plugin at a time. + +.. code-block:: xml + + + + 2.0 + + +------------------------ +XML Attributes +------------------------ + +```` (required): + Specifies the name of the vector field that encodes directional cues for the cells. + This field must be initialized in Python using ``self.field.``. + Each vector at a lattice point defines the preferred direction of movement at that point. + +```` (required): + A non-negative floating-point value controlling the strength of the polarization effect. + Higher values increase the influence of the vector field on cell motility. + Since we take the absolute value, the sign does not matter. + +------------------------ +Python Manipulation +------------------------ + +We can use the ``start`` function to define where the vectors point. +In this example, the cells will migrate in the negative x/y direction at specific positions. + +.. code-block:: python + + class VectorFieldSteppable(SteppableBasePy): + def __init__(self, frequency=10): + SteppableBasePy.__init__(self, frequency) + + def start(self): + fiber_field_cpp = self.field.Fibers + fiber_field_cpp[30, 10, 0] = [-30, -10, 0] + fiber_field_cpp[20, 30, 0] = [-20, -30, 0] + +This field can be updated dynamically during the simulation to reflect changing environments. + +------------------------ +Energy Contribution +------------------------ + +The plugin changes energy according: + +.. math:: + + E = -\left| \lambda \cdot (\vec{v}_{\text{field}} \cdot \vec{d}_{\text{COM (Old cell)}}) \right| - \left| \lambda \cdot (\vec{v}_{\text{field}} \cdot \vec{d}_{\text{COM (New cell)}}) \right| + +Where: + +- :math:`\lambda` is the polarization coefficient (``PolarizationLambda``). +- :math:`\vec{v}_{\text{field}}` is the vector from the field at the cell’s center of mass. +- :math:`\vec{d}_{\text{COM}}` is the displacement vector resulting from a pixel copy attempt. We evaluate this for both the new cell (if it exists) and old cell (if it exists). + +Since the energy contribution is negative, VectorFieldPolarization generally reduces the energy needed for cells to align in the direction specified by the vector field.