fastscapelib.create_flow_kernel

class fastscapelib.create_flow_kernel(flow_graph, kernel_func, spec, *, apply_dir=<FlowGraphTraversalDir.DEPTH_UPSTREAM: 2>, outputs=(), n_threads=1, get_data_at_receivers=True, set_data_at_receivers=True, get_data_at_donors=True, set_data_at_donors=True, auto_resize=False, print_stats=False)

Creates a numba flow kernel.

Parameters:

  • flow_graph (FlowGraph) – A flow graph object.

  • kernel_func (callable) – A Python function to apply to each node of the graph. It must take one input argument (numba jit-compiled class) that holds or references input and output kernel data of one graph node. It may return an integer.

  • spec (dict) – Dictionary where keys are kernel input and output variable names and values are either variable types (i.e. numba scalar or array types) or variable (type, value) tuples (only for scalar variables).

  • apply_dir (FlowGraphTraversalDir, optional) – The direction and order in which the flow kernel will be applied along the graph (default: downstream to upstream, depth-first search).

  • outputs (iterable) – The names of the kernel output variables. All names given here must be also present in spec. There must be at least one output.

  • n_threads (int, optional) – Number of threads to use for applying the kernel function in parallel along the flow graph (default: 1, the kernel will be applied sequentially). A value > 1 may be useful for kernel functions that are computationally expensive. For trivial kernel functions it is generally more efficient to apply the kernel sequentially (i.e., n_threads = 1).

  • get_data_at_receivers (bool, optional) – If True (default), kernel input and output (array) variables will have their values accessible at the node .receivers from within the kernel function (i.e., values are copied from the kernel data prior to calling the kernel function). You can set it to False if this is not needed, it may speed-up the application of the kernel.

  • set_data_at_receivers (bool, optional) – If True (default), allow setting the values of output variables at the node .receivers from within the kernel function (i.e., output values are copied back to the kernel data after calling the kernel function). You can set it to False if this is not needed, it may speed-up the application of the kernel. Note: when the kernel function is evaluated at a base level node, any value set via .receivers within the function won’t be copied in the output variable.

  • get_data_at_donors (bool, optional) – Same than get_data_at_receivers but for the node .donors.

  • set_data_at_donors (bool, optional) – Same than set_data_at_receivers but for the node .donors.

  • auto_resize (bool, optional) – If True, dynamically resize the temporary containers used to store data at flow receivers and donors. Otherwise (default), set fixed sizes according to the flow graph properties. Default settings are generally recommended.

  • print_stats (bool) – If True, prints a small report on kernel creation performance (default: False).

Returns:

__init__(**kwargs)