GSoC 2020 Week-1

The first week of GSoC has been completed. My goal for this week was to add ParametricRegion class. Many common regions have well known parametric representation. Calculation of Integrals on parametric regions is a common problem of vector calculus.

The PR for adding the ParametricRegion class is #19472.

ParametricRegion

An object of ParametricRegion class represents a region in space defined in terms of parameters. If a user wants to perform integration on a region, they have to create a ParametricRegion object. This object can then be passed to other classes to perform integration.

The implementation part was easy. The difficulty was in deciding the API. I spent much of the week discussing the API, But it was worth it. The ParametricRegion class will most likely be used by many users so it is important for it to be simple and intuitive. I think I should have first completed writing test cases and examples.

The final API of ParametricRegion class:

ParametricRegion(parameters_or_coordsys, definition, limits=None)

Some Examples:

p = ParametricRegion(t, (t, t**3), limits={t: (1, 2)})
halfdisc = ParametricRegion((r, theta), (r*cos(theta), r* sin(theta)), {r: (-2, 2), theta: (0, pi)})
cylinder = ParametricRegion((r, theta, z), (cos(theta), r*sin(theta), C.z), {theta: (0, 2*pi), z: (0, 4)})

parameters_or_coordsys

This argument specifies the parameters of the parametric region. When there are more than one parameter, the user has to pass them as tuple. When a CoordSys3d object is passed, its base scalars are used as parameters.

definition

definiton tuple maps the base scalars to thier parametric representation. It is not necessary for evey base scalar to have a parametric definition.

CoordSys3D

A parametric representation is just defining base scalars in terms of some parameters. If the base scalars are not knowm, the region cannot be completely specified. My intial approach was to include an argument to specify the coordinate system.

Passing a CoordSys3D object can be non intuitive for users. In most cases, the base scalars of the parametric region are the same as that of vector/scalar field. We can avoid passing the CoordSys3D object if we assume that the base scalars of parametric region field are same.

p = ParametricRegion((r, theta), (r*cos(theta), r*sin(theta)), limits={r: (0,1), theta: (0, pi)})
Integral(C.x**2*C.y, p)
# C.x = r*cos(theta),  C.y = r*sin(theta), C.z = C.z
Integral(R.x*R.y*R.z, p)
# R.x = r*cos(theta),  R.y = r*sin(theta), R.z = R.z

There can be a situation where SymPy cannot determine the coordinate system of the field. In such situations, we can raise a ValueError. For example:

Integral(R.x*C.z, p)

limits

Most parametric region have bounds on the parameters. Note that it is not necessary for every parameter to have an upper and lower bound.

There can be many methods for a user to define bounds of parametric region. We decied to use a dictionary which we named limits. The keys of the dictionary are parameters and the values are tuples (lower bound, upper bound). Another approach is to use a nested bound tuple. Every subtuple will represent the bounds of a parameter.

Next week’s goal

My next week’s task is to get started with ParametricIntegral class. This class will represent integral of a scalar/vector field on a parametric surface. I will first write the unit tests and examples.

Comments are closed