ParserRuleContext.Start.StartIndex > .Stop.StopIndex?

[n9]

Is it possible that ParserRuleContext.Start.StartIndex is greater than ParserRuleContext.Stop.StopIndex? If so, what are the cases?

Or is it a bug?

[kaby76]

That's ok because it means a zero lenthed token sequence (that subtree of the parse tree should have no tokens). When start and end index are the same it means a single token. It's confusing and, frankly, rather counter intuitive.

So, you would see this if you have an empty rule.

grammars-v4/pascal/pascal.g4

Line 433 in a0e876b

empty_

The start/end indices are the same counter intuitive definition for char streams in the lexer.

[n9]

Ok, what is the recommended way to get a text from a ParseRuleContext (e.g. to read a name of a identifier)?

Such function not:

public static string GetRawText(this ParserRuleContext context) => context.Start.InputStream.GetText(new(context.Start.StartIndex, context.Stop.StopIndex));

[kaby76]

I think it's just node.getText() in Java or node.Text in C#. It computes it from scratch if not memoized. In addition if you want the text with all that off-channel token stuff like whitespace those tokens have to be channel(HIDDEN) instead of skip cause skip tokens aren't on the token stream. If skip then Text won't have whitespe separating the tokens.

[n9]

Is the following function correct for reading all/raw text?

public static string GetRawText(this ParserRuleContext context)
{
    var start = context.Start.StartIndex;
    var stop = context.Stop.StopIndex;
    if (start >= stop)
        return string.Empty;
    return context.Start.InputStream.GetText(new(start, stop));
}

[kaby76]

Looks about right.