Syntax tree analysis instead of raw text analysis in spacing analyzers

[alxru]

Hi all,

I was researching StyleCop.Analyzers code as a reference for implementing my own custom analyzer and there is something that is not obvious to me.

Is there any reason that the complex token-level processing is used in spacing analyzers instead of simple parsing of raw text lines? For example, the analyzing code in SA1028CodeMustNotContainTrailingWhitespace class might look like the following:

SourceText sourceText = context.Tree.GetText(context.CancellationToken);
foreach (TextLine line in sourceText.Lines)
{
    if (line.Span.IsEmpty)
        continue;

    string text = line.ToString();
    int wsStart = FindTrailingWhitespace(text);

    if (wsStart >= 0)
    {
        Location diagnosticLocation = Location.Create(
            context.Tree, TextSpan.FromBounds(line.Start + wsStart, line.End));

        // TODO: report diagnostic for the location above
    }
}

This seems more simple and reliable than the current implementation.

[sharwell]

We don't examine text inside disabled text #if false, or within verbatim string literals. Also many of the spacing rules have special exceptions depending on where you are in the syntax tree (see #1191 for example).

[alxru]

It makes sense to walk the syntax tree then. Thanks for the answer.