This is the documentation for the unsupported version 2.2. Please consider upgrading your code to the latest stable version

Block Parsing

At a high level, block parsing is a two-step process:

  1. Using a BlockStartParserInterface to identify if/where a block start exists on the given line
  2. Using a BlockContinueParserInterface to perform additional processing of the identified block

So to implement a custom block parser you will actually need to implement both of these classes.


Instances of this interface have a single tryStart() method:

 * Check whether we should handle the block at the current position
 * @param Cursor                       $cursor
 * @param MarkdownParserStateInterface $parserState
 * @return BlockStart|null
public function tryStart(Cursor $cursor, MarkdownParserStateInterface $parserState): ?BlockStart;

Given a Cursor at the current position, plus some extra information about the state of the parser, this method is responsible for determining whether a particular type of block seems to exist at the given position. You don’t actually parse the block here - that’s the job of a BlockContinueParserInterface. Your only job here is to return whether or not a particular type of block does exist here, and if so which block parser should parse it.

If you find that you cannot parse the given block, you should return BlockStart::none(); from this function.

However, if the Markdown at the current position does indeed seem to be the type of block you’re looking for, you should return a BlockStart instance using the following static constructor pattern:

use League\CommonMark\Parser\Block\BlockStart;

return BlockStart::of(new MyCustomParser())->at($cursor);

Unlike in 1.x, the Cursor state is no longer shared between parsers. You must therefore explicitly provide the BlockStart object with a copy of your cursor at the correct, post-parsing position.

NOTE: If your custom block starts with a letter character you’ll need to add your parser to the environment with a priority of 250 or higher. This is due to a performance optimization where such lines are usually skipped.


The previous interface only helps the engine identify where a block starts. Additional information about the block, as well as the ability to parse additional lines of input, is all handled by the BlockContinueParserInterface.

This interface has several methods, so it’s usually easier to extend from AbstractBlockContinueParser instead, which sets most of the methods to use typical defaults you can override as needed.


public function getBlock(): AbstractBlock;

Each instance of a BlockContinueParserInterface is associated with a new block that is being parsed. This method here returns that block.


public function isContainer(): bool;

This method returns whether or not the block is a “container” capable of containing other blocks as children.


public function canContain(AbstractBlock $childBlock): bool;

This method returns whether the current block being parsed can contain the given child block.


public function canHaveLazyContinuationLines(): bool;

This method returns whether or not this parser should also receive subsequent lines of Markdown input. This is primarily used when a block can span multiple lines, like code blocks do.


public function addLine(string $line): void;

If canHaveLazyContinuationLines() returned true, this method will be called with the additional lines of content.


public function tryContinue(Cursor $cursor, BlockContinueParserInterface $activeBlockParser): ?BlockContinue;

This method allows you to try and parse an additional line of Markdown.


public function closeBlock(): void;

This method is called when the block is done being parsed. Any final adjustments to the block should be made at this time.


public function parseInlines(InlineParserEngineInterface $inlineParser): void;

This method is called when the engine is ready to parse any inline child elements.

Note: For performance reasons, this method is not part of BlockContinueParserInterface. If your block may contain inlines, you should make sure that your “continue parser” also implements BlockContinueParserWithInlinesInterface.


Here are some additional tips to consider when writing your own custom parsers:

Combining both into one file

Although parsing requires two classes, you can use the anonymous class feature of PHP to combine both into a single file! Here’s an example:

use League\CommonMark\Parser\Block\AbstractBlockContinueParser;
use League\CommonMark\Parser\Block\BlockStartParserInterface;

final class MyCustomBlockParser extends AbstractBlockContinueParser
    // TODO: implement your continuation parsing methods here

    public static function createBlockStartParser(): BlockStartParserInterface
        return new class implements BlockStartParserInterface
            // TODO: implement the tryStart() method here


The BlockStartParserInterface::tryStart() and BlockContinueParserInterface::tryContinue() methods may be called hundreds or thousands of times during execution. For best performance, have your methods return as early as possible, and make sure your code is highly optimized.

Block Elements

In addition to creating a block parser, you may also want to have it return a custom “block element” - this is a class that extends from AbstractBlock and represents that particular block within the AST.

If your block contains literal strings/text within the block (and not as part of a child block), you should have your custom block type also implement StringContainerInterface.

Edit this page