Testing Parsers

Derived classes will need to override RelativeTestDataPath to declare the path to the data and gold files, relative to test\data.

Each test method can call one of the following methods, which allows specifying the base name used to create the input and gold file names:

• DoNamedTest or DoNamedTest2 - will use the name of the test method as the file base name. The DoNamedTest2 method will also strip any Test prefix from the method first.

• DoOneTest will use the passed in string as the file base name.

These methods will specify the base name of the file to add to the in-memory project, but do not specify the file extension. This is added by the test base classes, by appending the value of the Extension property. This property will retrieve the extension to use by looking for attributes on either the test method or the test class. The first attribute that implements ITestFileExtensionProvider is used to get the value. Typically, this is achieved by using the [TestFileExtension] attribute, and specifying the extension (with dot) in the constructor. This is usually taken as a constant value defined in the project file type.

[TestFileExtension(CssProjectFileType.CSS_EXTENSION)]
public class CssParserTest : ParserTestBase<CssLanguage>
{
  protected override string RelativeTestDataPath { get { return @"parsing\css"; } }

  [Test] public void test01() { DoNamedTest(); }
  // ...[snip]...
}

The ParserTestBase<T> base class does not provide any hooks to customise. However, it is possible to use an alternative implementation, which does not require the {caret} special text in the input file, and includes extra asserts.

The default base class requires the input file to include {caret} so that it knows which file to test, even if there is only one file added to the in-memory project. As an alternative, a language can implement a different base class that doesn't have this requirement, something like this:

public abstract class AlternativeParserBaseTest<TLanguage> : BaseTestWithSingleProject
{
  protected abstract TLanguage GetLanguage();

  protected override void DoTest(IProject testProject)
  {
    var projectItem = testProject.GetSubItems().Single();
    string buffer = projectItem.Location.ReadAllText(Encoding.Default);

    var languageService = GetLanguage().LanguageService();
    Assert.IsNotNull(languageService);

    var lexer = languageService.CreateCachingLexer(new StringBuffer(buffer.Trim()));
    var psiModule = Solution.PsiModules().GetPrimaryPsiModule(testProject, TargetFrameworkId.Default);
    Assert.IsNotNull(psiModule);
    var parser = languageService.CreateParser(lexer, psiModule, null);
    var psiFile = parser.ParseFile();
    Assert.IsNotNull(psiFile);
    SandBox.CreateSandBoxFor(psiFile, psiModule);

    ExecuteWithGold(projectItem.Name, writer => DebugUtil.DumpPsi(writer, psiFile));
    Assert.AreEqual(buffer.Trim(), psiFile.GetText(), "Reconstructed text mismatch");
    CheckRange();
  }

  private static void CheckRange (string documentText, ITreeNode node)
  {
    Assert.AreEqual(node.GetText(), documentText.Substring(node.GetTreeStartOffset().Offset, node.GetTextLength()), "node range text mismatch");

    for (var child = node.FirstChild; child != null; child = child.NextSibling)
      CheckRange(documentText, child);
  }
}

This will create a parser for the single file in the project, retrieve the IFile, and compare it against the .gold file. The input file does not require the {caret} identifier.

Furthermore, this test will also check that the text of the file matches the text recreated by getting the text of each node in the tree, as well as validating the range of each tree node against the length of its text, further ensuring the tree is correct.