kcartlidge / markdowner

(ALPHA) Fast-enough and efficient-enough Markdown parsing in pure C# for .Net Core and .Net Framework.

Markdowner

Fast-enough and efficient-enough Markdown parsing in pure C# for .Net Core and .Net Framework.

This is written as a personal challenge. I've done Markdown parsers before and have used parser generators. I've even written my own equivalents to yacc, bison, etc (not just the grammars).

This was different, a challenge to do something by hand without using techniques such as recursive descent or compiler compilers.

It is not feature complete, but is perfectly fine for many simpler cases, supporting:

• Inline (nested) formatting
  • Bold, Italics, Underline, Code
• Headers from 1 to 6
• Paragraphs
  • Adjacent lines are merged into single paragraphs
• Ordered lists
• Unordered lists
• Preformatted blocks
• Block quotes
• Horizontal lines
• Tracking line numbers
  • Line number from the original Markdown
  • Line number in the parsed lines
• Unit tests

Sample usage

Check out the example project in the Example folder.

Progress

It's at an ALPHA stage and is a working proof of concept.

Missing:

• Packaging for Nuget
• Nested lists
• Tables
• Links
• Images

Methodology

There is a 2-step process

1. Convert the input text into a MarkdownDocument containing a collection of lines.
   • Each line is assigned a type, such as Paragraph or UnorderedList.
   • Each line has a Token collection holding all it's text and/or formatting.
2. Generate an output from that document.
   • An HTML output generator is included.

The caller does not need to know it is a 2-step process; they just feed the output generator the original source. The reason for splitting into 2 steps is that parsing into lines and tokens is the computationally expensive bit so doing that separately (a) provides a form of advance compilation and (b) means the document can be fed through multiple output generators without the overhead of parsing again each time.

Parsing follows a simple flow

• Ignore leading empty lines
• Treat runs of empty lines as one single line
• Derive the line types from the line start characters
• Merge consecutive paragraphs (where not separated by empty lines)
• Generate a set of tokens to represent the line content
• Add line-level flags for start/stop markers in quote/pre blocks