Structured Document Parser

A PHP library for turning semi-structured text into a tree you can work with. It gives you the pieces for each stage and leaves the grammar to you:

Lexing. A Lexer runs your matchers against the text and produces a TokenStream. Matchers are grouped into states, and a matcher's callback can push or pop states. Lookahead matches that consume nothing are allowed when they switch state. Each token records its line and column.
Parsing. You implement ParserInterface to turn tokens into a tree of your own node classes, built on AbstractNode. A node holds attributes, and a series of child nodes (such as a document's blocks) goes in an array attribute.
Transforming. A NodeTraverser walks the tree and calls your NodeVisitorInterface visitors, which can replace or remove nodes.
Rendering. VisitorInterface, VisitableInterface and RendererInterface are extension points for double-dispatch visitors, for example a renderer with a visitHeading() method per node type.

Requirements

PHP 8.2 or later with the mbstring extension.

Installation

composer require chrisjenkinson/structured-document-parser

Usage

This example parses a small Markdown-like document with headings and paragraphs, then gives each heading an id.

require __DIR__ . '/vendor/autoload.php';

use chrisjenkinson\StructuredDocumentParser\Finder\RegexFinder;
use chrisjenkinson\StructuredDocumentParser\Lexer\Lexer;
use chrisjenkinson\StructuredDocumentParser\Matcher\AbstractMatcher;
use chrisjenkinson\StructuredDocumentParser\Matcher\MatchedText;
use chrisjenkinson\StructuredDocumentParser\Node\AbstractNode;
use chrisjenkinson\StructuredDocumentParser\Node\NodeInterface;
use chrisjenkinson\StructuredDocumentParser\NodeTraverser\NodeTraverser;
use chrisjenkinson\StructuredDocumentParser\NodeVisitor\AbstractNodeVisitor;
use chrisjenkinson\StructuredDocumentParser\Parser\ParserInterface;
use chrisjenkinson\StructuredDocumentParser\State\InitialState;
use chrisjenkinson\StructuredDocumentParser\Token\TokenStream;

// 1. Matchers recognise pieces of text. The token type is the matcher name without "Matcher".
final class RegexMatcher extends AbstractMatcher
{
    private readonly RegexFinder $finder;

    public function __construct(private readonly string $type, string $pattern)
    {
        $this->finder = new RegexFinder($pattern);
    }

    public function getName(): string
    {
        return $this->type . 'Matcher';
    }

    public function match(string $text): ?MatchedText
    {
        if (!$this->finder->find($text)) {
            return null;
        }

        return new MatchedText($this->finder->getMatches(['all', 'level', 'text']));
    }
}

// 2. A state holds the matchers that apply at a point in the document.
$state = new InitialState();
$state->registerMatcher(new RegexMatcher('Heading', '/(?<all>(?<level>#{1,6}) (?<text>[^\n]*)\n?)/A'));
$state->registerMatcher(new RegexMatcher('Paragraph', '/(?<all>(?<text>[^#\n][^\n]*)\n?)/A'));
$state->registerMatcher(new RegexMatcher('BlankLine', '/(?<all>\n)/A'));

$tokens = (new Lexer($state))->tokenise("# Title\nSome text.\n\n## Section\nMore text.\n");

echo $tokens, "\n\n";

// 3. Your own node classes and parser turn tokens into a tree.
final class Document extends AbstractNode {}
final class Heading extends AbstractNode {}
final class Paragraph extends AbstractNode {}

final class DocumentParser implements ParserInterface
{
    public function parse(TokenStream $tokens): NodeInterface
    {
        $blocks = [];

        while (null !== $token = $tokens->getCurrentToken()) {
            $tokens->consumeToken();

            $block = match ($token->getType()) {
                'Heading'   => new Heading(),
                'Paragraph' => new Paragraph(),
                default     => null,
            };

            if (null === $block) {
                continue;
            }

            $block->setAttribute('text', $token->getValue('text'));

            if ($token->hasKey('level')) {
                $block->setAttribute('level', mb_strlen($token->getValue('level')));
            }

            $blocks[] = $block;
        }

        $document = new Document();
        $document->setAttribute('blocks', $blocks);

        return $document;
    }
}

$document = (new DocumentParser())->parse($tokens);

// 4. Visitors transform the tree: here, giving each heading an id.
final class HeadingIdVisitor extends AbstractNodeVisitor
{
    public function enterNode(NodeInterface $node): ?NodeInterface
    {
        if ($node instanceof Heading) {
            $node->setAttribute('id', mb_strtolower(str_replace(' ', '-', $node->getAttribute('text'))));
        }

        return null;
    }
}

$traverser = new NodeTraverser();
$traverser->addVisitor(new HeadingIdVisitor());

echo $traverser->traverse($document), "\n";

The lexer produces these tokens:

Heading (# Title)
Paragraph (Some text.)
BlankLine ()
Heading (## Section)
Paragraph (More text.)

and the traversed document contains the headings with their ids:

{
    "attributes": {
        "blocks": [
            {
                "attributes": {
                    "text": "Title",
                    "level": 1,
                    "id": "title"
                },
                "nodes": []
            },
            ...
        ]
    },
    "nodes": []
}

If no matcher matches, or more than one does, the lexer throws NoTokenFoundException or AmbiguousTokenFoundException, with the line, column and the text at that point.

Development

composer install
composer test     # phpspec, PHPUnit, PHPStan and the code style check
composer cs-fix   # fix code style

Licence

GPL-3.0-or-later. See LICENSE.

FilesExpand file tree

README.md

Latest commit

History