Passing arguments to low level operator

[ilGriso92]

Hello, first and foremost I wanted to thank all the maintainers for the time an effort that is going into curating this package. Working with it so far is something I have really enjoyed.

I wanted to open an issue to ask about operators (e.g., gradient, laplace...). I know it is possible to pass the time as argument, when passing an expression as the boundary condition. For example here I am passing the time t:

def boundary(value, dx, coords, t):
    return np.sin(t)

field.laplace(bc=[{"value_expression": boundary}, {"derivative": 0.0}], args=dict(t=0.0))

However, I could not find a way of passing time as an argument to the low level operators.

operator = field.grid.make_operator("laplace", bc=[{"value_expression": boundary}, {"derivative": 0.0}])
operator(field.data)

Could someone confirm if this can be achieved and perhaps provide a short working example?

Many thanks

[david-zwicker]

Thank you for using the package! I don't really understand what you're trying to achieve. The Laplace operator does not depend on time, so passing time doesn't make much sense. I suppose you're trying to achieve something else, so could you please describe what you are trying to do and what outcome you expect?

[ilGriso92]

Thanks for the quick acknowledgement. I am referring to a situation where I would like to impose a time-dependent boundary condition to a field. I can easily pass the time as part of the args when using the high level operator, e.g.

def boundary(value, dx, coords, t):
    return np.sin(t)

field.gradient(bc=[{"value_expression": boundary}, {"derivative": 0.0}], args=dict(t=0.0))

but I am not sure I can achieve the same with a low level operator, e.g.

def boundary(value, dx, coords, t):
    return np.sin(t)

operator = field.grid.make_operator("gradient", bc=[{"value_expression": boundary}, {"derivative": 0.0}])
operator(field.data)

thanks for your support on this.

[david-zwicker]

This can be done by supplying a numba dictionary when calling the operator:

import pde
from pde.tools.numba import numba_dict

grid = pde.UnitGrid([4, 4])
field = pde.ScalarField.random_uniform(grid)

def boundary(value, dx, x, y, t):
    return np.sin(t)

operator = grid.make_operator("gradient", bc=[{"value_expression": boundary}, {"derivative": 0.0}])
operator(field.data, args=numba_dict({"t": 1}))

I'll look into whether we could call numba_dict automatically in future versions of the package, so one could simply provide a python dictionary instead.

[ilGriso92]

Apologies about the delay, got dragged into other projects and was able to test this only now. It worked great! Thanks lots for your time and again for maintaining this great project 🙌

[david-zwicker]

Good to hear that it helps! With newer versions of the package, you should also be able to use numba_dict(t=1) for convenience.