如何使用 PEG.js 创建一个简单的解析器

Question

The syntax I would like to parse is of the following kind:

# This is a comment

# This is a block. It starts with \begin{} and ends with \end{}
\begin{document}

# Within the document block other kinds of blocks can exist, and within them yet other kinds.
# Comments can exist anywhere in the code.

This is another block within the block. It is a paragraph, no formal \begin{} and \end{} are needed.
The parser infers its type as a ParagraphBlock. The block ends with the newline.

\end{document}

I am learning how to use PEG, and this is what I have developed so far for the current syntax:

Start
  = (Newline / Comment / DocumentBlock)*
  
Comment
  = '#' value: (!Newline .)* Newline? {
    return {
      type: "comment",
      value: value.map(y => y[1]).join('').trim()
    }
  } 
  
Newline
  = [\n\r\t]
  
DocumentBlock
  = "\\begin\{document\}"
  
    (!"\\end\{document\}" DocumentChildren)*
    
    "\\end\{document\}"
    
DocumentChildren
  = NewlineBlock / ParagraphBlock
    
NewlineBlock
  = value: Newline*
  {
    return {
      type: "newline",
      value: value.length
    }
  }
    
ParagraphBlock
  = (!Newline .)* Newline

I am having some issues with infinite loops. The current code produces this error:

Line 19, column 5: Possible infinite loop when parsing (repetition used with an expression that may not consume any input).

What would be a correct implementation for the simple syntax above?

Answer 1

I think this is due to the NewlineBlock rule using a kleene star on Newline .

In DocumentBlock you have a repeated DocumentChildren . In NewlineBlock you have a repeated Newline which means that it can always return '' , the null string, which would cause an infinite loop.

Changing the * in NewlineBlock to a + would fix the problem. That way it no longer has the option of returning the null string.

NewlineBlock
  = value: Newline+
  {
    return {
      type: "newline",
      value: value.length
    }
  }