GSoC Final Report

31 August 2020 - 10 mins read time
Tags: Sympy GSoC

This report summarizes the work done in my GSoC 2020 project, Implementation of Vector Integration with SymPy. Blog posts with step by step development of the project are available at friyaz.github.io.

About Me

I am Faisal Riyaz and I have completed my third year of B.Tech in Computer Engineering student from Aligarh Muslim University.

Overview

The goal of the project is to add functions and class structure to support the integration of scalar and vector fields over curves, surfaces, and volumes. My mentors were Francesco Bonazzi (Primary) and Divyanshu Thakur.

Work Completed

Here is a list of PRs which were opened during the span of GSoC:

We had an important discussion regarding our initial approach and possible problems on issue #19320

Defining Regions with their parametric equations

The ParametricRegion class is used to represent a parametric region in space.

>>> from sympy.vector import CoordSys3D, ParametricRegion, vector_integrate
>>> from sympy.abc import r, theta, phi, t, x, y, z
>>> from sympy import pi, sin, cos, Eq

>>> C = CoordSys3D('C')

>>> circle = ParametricRegion((4*cos(theta), 4*sin(theta)), (theta, 0, 2*pi))
>>> disc = ParametricRegion((r*cos(theta), r*sin(theta)), (r, 0, 4), (theta, 0, 2*pi))
>>> box = ParametricRegion((x, y, z), (x, 0, 1), (y, -1, 1), (z, 0, 2))
>>> cone = ParametricRegion((r*cos(theta), r*sin(theta), r), (theta, 0, 2*pi), (r, 0, 3))

>>> vector_integrate(1, circle)
8*pi
>>> vector_integrate(C.x*C.x - 8*C.y*C.x, disc)
64*pi
>>> vector_integrate(C.i + C.j, box)
(Integral(1, (x, 0, 1), (y, -1, 1), (z, 0, 2)))*C.i + (Integral(1, (x, 0, 1), (y, -1, 1), (z, 0, 2)))*C.j
>>> _.doit()
4*C.i + 4*C.j
>>> vector_integrate(C.x*C.i - C.z*C.j, cone)
-9*pi

Integration over objects of geometry module

Many regions like circle are commonly encountered in problems of vector calculus. the geometry module of SymPy already had classes for some of these regions. The functionparametric_region_list was added to determine the parametric representation of such classes. vector_integrate can directly integrate objects of Point, Curve, Ellipse, Segment and Polygon class of geometry module.

>>> from sympy.geometry import Point, Segment, Polygon, Segment
>>> s = Segment(Point(4, -1, 9), Point(1, 5, 7))
>>> triangle = Polygon((0, 0), (1, 0), (1, 1))
>>> circle2 = Circle(Point(-2, 3), 6)
>>> vector_integrate(-6, s)
-42
>>> vector_integrate(C.x*C.y, circle2)
-72*pi
>>> vector_integrate(C.y*C.z*C.i + C.x*C.x*C.k, triangle)
-C.z/2

Implictly defined Regions

In many cases, it is difficult to determine the parametric representation of a region. Instead, the region is defined using its implicit equation. To represent implicitly defined regions, we implemented the ImplicitRegion class.

But to integrate over a region, we need to determine its parametric representation. the rationl_parametrization method was added to ImplicitRegion.

>>> from sympy.vector import ImplicitRegion
>>> circle3 = ImplicitRegion((x, y), (x-4)**2 + (y+3)**2 - 16)
>>> parabola = ImplicitRegion((x, y), (y - 1)**2 - 4*(x + 6))
>>> ellipse = ImplicitRegion((x, y), (x**2/4 + y**2/16 - 1))
>>> rect_hyperbola = ImplicitRegion((x, y), x*y - 1)
>>> sphere = ImplicitRegion((x, y, z), Eq(x**2 + y**2 + z**2, 2*x))

>>> circle3.rational_parametrization()
(8*t/(t**2 + 1) + 4, 8*t**2/(t**2 + 1) - 7)
>>> parabola.rational_parametrization()
(-6 + 4/t**2, 1 + 4/t)
>>> rect_hyperbola.rational_parametrization(t)
(-1 + (t + 1)/t, t)
>>> ellipse.rational_parametrization()
(t/(2*(t**2/16 + 1/4)), t**2/(2*(t**2/16 + 1/4)) - 4)
>>> sphere.rational_parametrization()
(2/(s**2 + t**2 + 1), 2*s/(s**2 + t**2 + 1), 2*t/(s**2 + t**2 + 1))

After PR #19883 gets merged, vector_integrate can directly work upon ImplicitRegion objects.

Future Work

Conclusion

This summer has been a great learning experience. Writing a functionality, testing, debugging it has given me good experience in test-driven development. I think I realized many of the goals that I described in my proposal, though there were some major differences in the plan and actual implementation. Personally, I hoped I could have done better. I could not work much on the project during the last phase due to the start of the academic year.

I would like to thank Francesco for his valuable suggestions and being readily available for discussions. He was always very friendly and helpful. I would love to work more with him. S.Y Lee also provided useful suggestions.

Special thanks to Johann Mitteramskogler, Professor Sonia Perez Diaz, and Professor Sonia L.Rueda. They personally helped me in finding the algorithm for determining the rational parametrization of conics.

SymPy is an amazing project with a great community. I would mostly stick around after the GSoC period as well and continue contributing to SymPy, hopefully, exploring the other modules as well.

Comments are closed