Txtmark - Java markdown processor

Copyright (C) 2011-2015 René Jeschke [email protected]
See LICENSE.txt for licensing information.

Txtmark is yet another markdown processor for the JVM.

• It is easy to use:

  String result = txtmark.Processor.process("This is ***TXTMARK***");
  

• It is fast (see below)
  ... well, it is the fastest markdown processor on the JVM right now. (This might be outdated, but txtmark is still flippin' fast.)

• It does not depend on other libraries, so classpathing txtmark.jar is sufficient to use Txtmark in your project.

For an in-depth explanation of markdown have a look at the original Markdown Syntax.

Maven repository

Txtmark is available on maven central.

Txtmark extensions

To enable Txtmark's extended markdown parsing you can use the $PROFILE$ mechanism:

[$PROFILE$]: extended

This seemed to me as the easiest and safest way to enable different behaviours. Just put this line into your Txtmark file like you would use reference links.

Behavior changes when using [$PROFILE$]: extended

• Lists and code blocks end a paragraph

  In normal markdown the following:

  This is a paragraph
  * and this is not a list
  

  Will produce:

  <p>This is a paragraph
  * and this is not a list</p>
  

  When using Txtmark extensions this changes to:

  <p>This is a paragraph</p>
  <ul>
  <li>and this is not a list</li>
  </ul>
  

• Text anchors

  Headlines and list items may recieve an ID which you can refer to using links.

  ## Headline with ID ##     {#headid}
  
  Another headline with ID   {#headid2}
  ------------------------
  
  * List with ID             {#listid}
  
  Links: [Foo] (#headid)
  

  this will produce:

  <h2 id="headid">Headline with ID</h2>
  <h2 id="headid2">Another headline with ID</h2>
  <ul>
  <li id="listid">List with ID</li>
  </ul>
  <p>Links: <a href="#headid">Foo</a></p>
  

  The ID must be the last thing on the first line.

  All spaces before {# get removed, so you can't use an ID and a manual line break in the same line.

• Auto HTML entities

  • (C) becomes © - ©
  • (R) becomes ® - ®
  • (TM) becomes ™ - ™
  • -- becomes – - –
  • --- becomes — - —
  • ... becomes … - …
  • << becomes « - «
  • >> becomes » - »
  • "Hello" becomes “Hello” - “Hello”

• Underscores (Emphasis)

  Underscores in the middle of a word don't result in emphasis.

  Con_cat_this
  

  normally produces this:

  Con<em>cat</em>this
  

• Superscript

  You can use ^ to mark a span as superscript.

  2^2^ = 4
  

  turns into

  2<sup>2</sup> = 4
  

• Abbreviations

  Abbreviations are defined like reference links, but using a * instead of a link and must be single-line only.

  [Git]: * "Fast distributed revision control system"
  

  and used like this

  This is [Git]!
  

  which will produce

  This is <abbr title="Fast distributed revision control system">Git</abbr>!
  

• Fenced code blocks

  ```
  This is code!
  ```
  
  ~~~
  Another code block
  ~~~
  
  ~~~
  You can also mix flavours
  ```
  

  Fenced code block delimiter lines do start with at least three of `` or `~

  It is possible to add meta data to the beginning line. Everything trailing after `` or `~ is then considered meta data. These are all valid meta lines:

  ```python
  ~ ~ ~ ~ ~java
  ``` ``` ``` this is even more meta
  

  The meta information that you provide here can be used with a BlockEmitter to include e.g. syntax highlighted code blocks. Here's an example:

  public class CodeBlockEmitter implements BlockEmitter
  {
      private static void append(StringBuilder out, List<String> lines)
      {
          out.append("<pre class=\"pre_no_hl\">");
          for (final String l : lines)
          {
              Utils.escapedAdd(out, l);
              out.append('\n');
          }
          out.append("</pre>");
      }
  
      @Override
      public void emitBlock(StringBuilder out, List<String> lines, String meta)
      {
          if (Strings.isEmpty(meta))
          {
              append(out, lines);
          }
          else
          {
              try
              {
                  // Utils#highlight(...) is not included with txtmark, it's sole purpose
                  // is to show what the meta can be used for
                  out.append(Utils.highlight(lines, meta));
                  out.append('\n');
              }
              catch (final IOException e)
              {
                  // Ignore or do something, still, pump out the lines
                  append(out, lines);
              }
          }
      }
  }
  

  You can then set the BlockEmitter in the txtmark Configuration using Configuration.Builder#setCodeBlockEmitter(BlockEmitter emitter).

Markdown conformity

Txtmark passes all tests inside MarkdownTest_1.0_2007-05-09 except of two:

1. Images.text

   Fails because Txtmark doesn't produce empty 'title' image attributes.
   (IMHO: Images ... OK)

2. Literal quotes in titles.text

   What the frell ... this test will continue to FAIL.
   Sorry, but using unescaped " in a title which should be surrounded by " is unacceptable for me ;)

   Change:

   Foo [bar](/url/ "Title with "quotes" inside").
   [bar]: /url/ "Title with "quotes" inside"
   

   to:

   Foo [bar](/url/ "Title with \"quotes\" inside").
   [bar]: /url/ "Title with \"quotes\" inside"
   

   and Txtmark will produce the correct result.
   (IMHO: Literal quotes in titles ... OK)

Where Txtmark is not like Markdown

• Txtmark does not produce empty title attributes in link and image tags.

• Unescaped " in link titles starting with " are not recognized and result in unexpected behaviour.

• Due to a different list parsing approach some things get interpreted differently:

  * List
  > Quote
  

  will produce when processed with Markdown:

  <p><ul>
  <li>List</p>
  
  <blockquote>
   <p>Quote</li>
  </ul></p>
  </blockquote>
  

  and this when produced with Txtmark:

  <ul>
  <li>List<blockquote><p>Quote</p>
  </blockquote>
  </li>
  </ul>
  

  Another one:

  * List
  ====
  

  will produce when processed with Markdown:

  <h1>* List</h1>
  

  and this when produced with Txtmark:

  <ul>
  <li><h1>List</h1>
  </li>
  </ul>
  

• List of escapeable characters:

  \   [   ]   (   )   {   }   #
  "   '   .   <   >   +   -   _
  !   `   ^
  

Performance comparison of markdown processors for the JVM

Remarks: These benchmarks are too old to be of any value. I leave them here as a reference, though.

Based on this benchmark suite.

Excerpt from the original post concerning this benchmark suite:

Most of these tests are of course unrealistic: Who would write a text where each word is a link? Yet they serve an important use: It makes it possible for the developer to pinpoint the parts of the parser where there is most room for improvement. Also, it explains why certain texts might render much faster in one Processor than in another.

Benchmark system:

• Ubuntu Linux 10.04 32 Bit
• Intel(R) Core(TM) 2 Duo T7500 @ 2.2GHz
• Java(TM) SE Runtime Environment (build 1.6.0_24-b07)
• Java HotSpot(TM) Server VM (build 19.1-b02, mixed mode)

Benchmarked versions:

Actuarius version: 0.2
PegDown version: 0.8.5.4
Knockoff version: 0.7.3-15

Mentioned/related projects

Markdown is Copyright (C) 2004 by John Gruber
SmartyPants is Copyright (C) 2003 by John Gruber
Actuarius is Copyright (C) 2010 by Christoph Henkelmann
Knockoff is Copyright (C) 2009-2011 by Tristan Juricek
PegDown is Copyright (C) 2010 by Mathias Doenitz
PHP Markdown & Extra is Copyright (C) 2009 Michel Fortin

Project link: https://github.com/rjeschke/txtmark