Line number?

[bobbbob98]

It doesn't seem like there's a way to get the current line number and position in the string within the parser. This would be useful for creating a language where any error message print the line and position of the error.

[zevv]

Yes, you are right: NPeg does not keep track of the line number in the source file; there are a few reasons for this:

• Counting newlines and keeping track of additional state while parsing takes time, counting newlines will slow down the parser quite a bit. (I did a test implementation for this once, I believe the cost was almost 4%. The question if this feature is worth the price, of course.
• More important: NPeg does not know what a "line" is, it just parses a stream of things. (These things are typically characters, but NPeg can parse arrays of any type).
• Even if NPeg would implement this when parsing a string (which is an array of characters), there are encoding ambiguities that complicate things: what even is a newline? there's unix formatting with \n and dos formatting with '\r\n'; and how to handle funny things like \r\r\n, is that one or two newlines?

That was the bad news. The good news is: NPeg actually does keep track of the position in the string, and if parsing failed it will tell you exactly where this happened. Take a look at this part of the documentation: https://github.com/zevv/npeg?tab=readme-ov-file#error-handling

If you want to know the line number and position in the line, there are a few ways you can do this:

• As part of your grammar, you explicitly handle newlines, and you add a bit of code that increments a counter every time you hit a newline.
• If parsing fails, you know the exact byte position in the file; you could write a few lines of Nim that quickly scans through the source file up to that byte and calculates what line/column that position is.

I personally like the second option: the advantage is that you only have to do this work if parsing failed, so there is no run time cost cost to this when parsing properly formatted documents.