Introduction
Test Driven Development is the
practice of (1) writing tests, (2) writing code that passes those tests, and (3)
then refactoring. This concept is becoming very popular in the .NET community
due to the quality assurance that it adds. While it is easy to test public
methods, the natural question emerges "How do I test protected and private
methods?"
This article will:
- Summarize some key points of the "Should-you-even-test private methods
debate"?
- Make the case that it is still useful to at least know how to test private
and protected methods - regardless of which side of the debate you're on.
- Provide the theory and downloadable code sample to demonstrate these testing
techniques.
Background Theory
Should You Test Private Methods?
A Google
search will show you that there's a lot of debate about using private
methods, let alone testing them. The table below summarizes some of the common
views of the pro and con for both issues.
|
Pro |
Con |
Use private methods |
- Encapsulation - Private methods provide encapsulation, which makes
the code easier for the end client to use
- Refactoring - It is easier to refactor private methods because they
are never directly called by external clients, therefore changing the signature
won't break any method calls.
- Validation - Unlike public methods that must validate all input
because they are called by external clients, private methods are called safely
within the class and don't require the same rigor of validation - the inputs
should have been already validated in the public methods.
- Test Scope - Exposing every method as public greatly increases the
test scope. Private methods are used only how the developer intended them,
whereas public methods need to be tested for every possible case which required
a broader test scope.
|
- Didn't Refactor - If a class is complicated enough to merit private
members, then it should be refactored.
- Hides Functionality - Private methods (if designed correctly) provide
useful features that clients may want access to, so any private method worth
testing should really be public.
|
Test private methods |
- Test Control - Private methods can contain complex logic and it
increases test control to be able to directly access the method and test it,
instead of needing to indirectly accessing it through a public method.
- Principle - Unit Testing is about testing the smallest functional
piece of code; private methods are functional pieces of code, therefore on
principle private methods should be testable.
|
- Already Covered - Only the public interface should be tested. Private
methods should already have thorough test coverage from being called by the
public methods that are tested.
- Brittle Code - If you refactor the code, and juggle around the
private methods, and if you had tests linked to those private methods, then you
need to juggle around your tests too.
|
There are bright and experienced people on both sides of the issue. So while
I have no intention or expectation of ending the "should I test private method"
debate, there is still value for both sides to know how to test them. Even if
you think that private methods should not be tested:
- Your opinion will be more influential if you can at least show that you are
able to test them, but choose not to (i.e. you're not saying "Don't test private
methods" simply because you don't know how).
- Having the option of testing non-public methods lets you see what really
works best for your team over time.
- As long as there are still some valid situations, it's worth having a
convenient way to test them.
Good Criteria and Inadequate Techniques
Andrew Hunt and David Thomas explain in their book, Pragmatic
Unit Testing in C# with NUnit, that good unit tests are ATRIP:
- Automatic
- Thorough
- Repeatable
- Independent
- Professional
There are three additional criteria that any testing overhead for
private/protected methods should meet:
- Transparency - Don't alter the System Under Test (SUT), such as
adding wrapper methods in the production code.
- Scope - Be able to run in both Debug and Release mode.
- Simplicity - Have minimal overhead, and hence be easy to change and
simple enough to introduce minimal risk.
Keeping these criteria in mind, there are several strategies that fall
short:
Strategy |
Problem |
Don't have any private methods. |
|
Use the directives #if DEBUG ... #endif
to wrap a public method which in turns wraps the private method. The unit tests
can now indirectly access that private method through the public wrapper. (This
is a method that I myself have used many times, and found it to be tedious and
non-object oriented). |
- This only works in Debug mode.
- This is procedural and not object-oriented. We would need to wrap the
individual methods in both the production code and unit tests.
- This alters the SUT by adding the public method
wrappers.
|
Use the [Conditional("DEBUG")] attribute
on public methods that wrap the private methods. |
- This only works in Debug mode.
|
Create internal methods to access the private method;
then have a public test class elsewhere in the assembly that wraps those
internal methods with public ones. |
- This alters the release code by adding the internal hooks, ultimately making
the private methods accessible in production.
- This requires a lot of extra coding, and is hence
brittle.
|
Testing Protected Methods
A protected method is visible only to derived classes, therefore it is not
immediately available to a test suite. For example, suppose we wanted to test
the method from ClassLibrary1.MyObject
:
protected string MyProtectedMethod(string strInput, int i32Value)
{
return this.Name + ": " + strInput + ", " +
i32Value.ToString();
}
The book Pragmatic
Unit Testing in C# with NUnit explains one solution: make a derived class
MyObjectTester
that inherits class MyObject
, and then
create a public method TestMyProtectedMethod
that wraps the
protected one. For example:
public new string TestMyProtectedMethod(string strInput, int i32Value)
{
return base.MyProtectedMethod(strInput,
i32Value);
}
This approach is simple, yet meets all the criteria:
Criteria |
Fulfillment |
Transparency |
By using inheritance and putting the
MyObjectTester class in the UnitTests assembly, it
doesn't add any new code to the production assembly. |
Scope |
Nothing in this approach depends on Debug-only
techniques. |
Simplicity |
Although this approach requires a new derived class,
and an additional public wrapper method for each protected method, it is
object-oriented and type safe. |
Testing Private Methods
Testing private methods is a little more involved; but we can still do it
using System.Reflection.
You can use Reflection to dynamically access methods of a type, including both
instance and static private methods. Note that accessing private methods does
require the ReflectionPermission,
but that is not a problem for Unit Tests running on a development machine or
build server.
Suppose we wanted to test the private method MyPrivateMethod
from ClassLibrary1.MyObject
:
private string MyPrivateMethod(string strInput, DateTime dt, double
dbl)
{
return this.Name + ": " + strInput + ", " +
dt.ToString() + ", " + dbl.ToString();
}
One solution is to create a UnitTestUtilities project with a helper class to
call the test method via reflection. For example, the download solution has the
following methods in UnitTestUtilities.Helper
:
public static object RunStaticMethod(System.Type t, string strMethod,
object [] aobjParams)
{
BindingFlags eFlags =
BindingFlags.Static | BindingFlags.Public |
BindingFlags.NonPublic;
return RunMethod(t, strMethod,
null, aobjParams, eFlags);
}
public static object RunInstanceMethod(System.Type t, string strMethod,
object objInstance, object [] aobjParams)
{
BindingFlags eFlags = BindingFlags.Instance | BindingFlags.Public |
BindingFlags.NonPublic;
return RunMethod(t, strMethod,
objInstance, aobjParams, eFlags);
}
private static object RunMethod(System.Type t, string
strMethod, object objInstance, object [] aobjParams, BindingFlags eFlags)
{
MethodInfo m;
try
{
m = t.GetMethod(strMethod, eFlags);
if (m == null)
{
throw new ArgumentException("There is no method '" +
strMethod + "' for type '" + t.ToString() + "'.");
}
object objRet = m.Invoke(objInstance, aobjParams);
return objRet;
}
catch
{
throw;
}
}
Private method RunMethod
takes in the necessary parameters that
Reflection needs to invoke a method, and then returns the value. It has two
public methods that wrap this: RunStaticMethod
and
RunInstanceMethod
for static and instance methods respectively.
Walking through RunMethod
, it first gets the MethodInfo
from a type. Because we expect this to only be called for existing
methods, a null method triggers an Exception
. Once we have the
MethodInfo
, we can invoke the method given the instantiated object
(null
for static methods) and the parameter array.
We could use this Utility in an NUnit test like so:
[Test] public void TestPrivateInstanceMethod()
{
string strExpected = "MyName: Hello, 5/24/2004
12:00:00 AM, 2.1";
ClassLibrary1.MyObject objInstance
= new MyObject("MyName");
object obj =
UnitTestUtilities.Helper.RunInstanceMethod(
typeof(ClassLibrary1.MyObject), "MyPrivateMethod",
objInstance, new object[3] {"Hello",
new DateTime(2004,05,24), 2.1});
string strActual = Convert.ToString(obj);
Assert.AreEqual(strExpected,strActual);
}
Criteria |
Fulfillment |
Transparency |
The only extra code we created -
UnitTestUtilities , is not shipped in production. |
Scope |
Nothing in this approach depends on Debug-only
techniques. |
Simplicity |
This approach can call any method with a single call.
Once you have the UnitTestUtilities , the only complication is
creating the correct parameters (method name, data types, etc...) for
RunInstanceMethod or RunStaticMethod . Because the
method is being dynamically called, the parameters aren't checked at compile
time. |
Conclusion
While there is a debate on whether or not to test private methods, at least
we have the ability to do so. We can test protected methods using inheritance to
create a derived TesterClass
that wraps the base protected methods
with public ones. We can test private methods using Reflection, which can be
abstracted to a UnitTestUtility
helper class. Both of these
techniques can help to improve test coverage.