具有“复杂”参数的函数的C＃单元测试结构

Question

I am working in C# and trying to figure out which unit test framework (MSTest, Nunit, Xunit) to use in order to best approach a data driven testing methodology. I have some experience with unit testing, but only with very simple functions. The function that I would like to apply unit tests to is defined below, as well as my attempted solution, and the form of the desired solution.

Function to Unit Test

Point3D GetIntersectionPoint(List<Point3D> line1, List<Point3D> line2)

Which returns a 3D point (x,y,z) from two lines of arbitrary length and segments (assuming they intersect at a single point). One could say that all you need to test is the simpler function:

Point3D GetIntersectionPointSimple(LineSegment segment1, LineSegment segment2)

Where a "LineSegment" is a single straight line segment. However I want to trap the case with multiple intersections, and am interested in the solution as it is fairly general.

Attempted Solution

In MSTest it is easy to grab a row of data from a data source, and so I had the following in a CSV file:

LineA            LineB                IntersectionPoint
0,0;1,0;2,0      1.5,1;1.5,0;1.5,-1   0,1.5

Where the individual points are delimited by ';' and the (x,y) coordinates are delimited by ','. This example is in 2D just for brevity.

With this you can read in a each string like:

TestContext.DataRow["LineA"].ToString()

And then parse it to get out the desired list of points. This could then be checked like:

Assert.AreEqual(parsedIntersectionPoint, GetIntersectionPoint(parsedLine1, parsedLine2))

This however seems quite messy.

Desired Solution

Ideally the solution to this problem would satisfy the following:

1. Simple to add the data (for instance a serialized object, ease of manual entry as in the CSV approach)
2. Intuitive (in the sense that storing a list of points within a CSV file is not)
3. Preferably sticking to simple data sources such as CSV, XML, or JSON.

Really I am just looking for the "right" way to approach this, as the direction I took seems "wrong".

Any help would be appreciated!!

Answer 1

It's the right approach, obviously for 3D datasets it's inconvenient but serializing objects would not be simpler.

It might be easier to write 3 nested loops that boundary test the function and print those results to a csv. Then unit test the function with 3 nested loops (x,y,z) checking anything you change in the function doesn't change the expected output.