AGTree Integration

This document describes the integration of @adguard/agtree into the adblock-compiler project.

Overview

AGTree is AdGuard's official tool set for working with adblock filter lists. It provides:

Adblock rule parser - Parses rules into Abstract Syntax Trees (AST)
Rule converter - Converts rules between different adblock syntaxes
Rule validator - Validates rules against known modifier definitions
Compatibility tables - Maps modifiers/features across different ad blockers

Why AGTree?

Before AGTree

The compiler used custom regex-based parsing in RuleUtils.ts:

Limited to basic pattern matching
No formal grammar or AST representation
Manual modifier validation
No syntax detection for different ad blockers
Prone to edge-case parsing errors

After AGTree

Feature	Before	After
Rule Parsing	Custom regex	Full AST with location info
Syntax Support	Basic adblock	AdGuard, uBlock Origin, Adblock Plus
Modifier Validation	Hardcoded list	Compatibility tables
Error Handling	String matching	Structured errors with positions
Rule Types	Network + hosts	All cosmetic, network, comments
Maintainability	Manual updates	Upstream library updates

Architecture

Module Structure

src/utils/
├── AGTreeParser.ts    # Wrapper module for AGTree
├── RuleUtils.ts       # Refactored to use AGTreeParser
└── index.ts           # Exports AGTreeParser types

AGTreeParser Wrapper

The AGTreeParser class provides a simplified interface to AGTree:

import { AGTreeParser } from '@/utils/AGTreeParser.ts';

// Parse a single rule
const result = AGTreeParser.parse('||example.com^$third-party');
if (result.success && AGTreeParser.isNetworkRule(result.ast!)) {
    const props = AGTreeParser.extractNetworkRuleProperties(result.ast);
    console.log(props.pattern);    // '||example.com^'
    console.log(props.modifiers);  // [{ name: 'third-party', value: null, exception: false }]
}

// Parse an entire filter list
const filterList = AGTreeParser.parseFilterList(rawFilterListText);
for (const rule of filterList.children) {
    if (AGTreeParser.isNetworkRule(rule)) {
        // Process network rule
    }
}

// Detect syntax
const syntax = AGTreeParser.detectSyntax('example.com##+js(aopr, ads)');
// Returns: AdblockSyntax.Ubo

Key Features

1. Type Guards

AGTreeParser provides comprehensive type guards for all rule types:

AGTreeParser.isEmpty(rule)           // Empty lines
AGTreeParser.isComment(rule)         // All comment types
AGTreeParser.isSimpleComment(rule)   // ! or # comments
AGTreeParser.isMetadataComment(rule) // ! Title: ...
AGTreeParser.isHintComment(rule)     // !+ NOT_OPTIMIZED
AGTreeParser.isPreProcessorComment(rule) // !#if, !#include
AGTreeParser.isNetworkRule(rule)     // ||domain^ style
AGTreeParser.isHostRule(rule)        // /etc/hosts style
AGTreeParser.isCosmeticRule(rule)    // ##, #@#, etc.
AGTreeParser.isElementHidingRule(rule)
AGTreeParser.isCssInjectionRule(rule)
AGTreeParser.isScriptletRule(rule)
AGTreeParser.isExceptionRule(rule)   // @@ or #@# rules

2. Property Extraction

Extract structured data from parsed rules:

// Network rules
const props = AGTreeParser.extractNetworkRuleProperties(networkRule);
// Returns: { pattern, isException, modifiers, syntax, ruleText }

// Host rules
const hostProps = AGTreeParser.extractHostRuleProperties(hostRule);
// Returns: { ip, hostnames, comment, ruleText }

// Cosmetic rules
const cosmeticProps = AGTreeParser.extractCosmeticRuleProperties(cosmeticRule);
// Returns: { domains, separator, isException, body, type, syntax, ruleText }

3. Modifier Utilities

Work with network rule modifiers:

// Find a specific modifier
const mod = AGTreeParser.findModifier(rule, 'domain');

// Check if modifier exists
const hasThirdParty = AGTreeParser.hasModifier(rule, 'third-party');

// Get modifier value
const domainValue = AGTreeParser.getModifierValue(rule, 'domain');
// Returns: 'example.com|~example.org' or null

4. Validation

Validate rules and modifiers:

// Validate a single modifier
const result = AGTreeParser.validateModifier('important', undefined, AdblockSyntax.Adg);
// Returns: { valid: boolean, errors: string[] }

// Validate all modifiers in a network rule
const validation = AGTreeParser.validateNetworkRuleModifiers(rule);
if (!validation.valid) {
    console.log(validation.errors);
}

5. Syntax Detection

Automatically detect which ad blocker syntax a rule uses:

const syntax = AGTreeParser.detectSyntax(ruleText);
// Returns: AdblockSyntax.Adg | Ubo | Abp | Common

// Check specific syntax
AGTreeParser.isAdGuardSyntax(rule)   // AdGuard-specific
AGTreeParser.isUBlockSyntax(rule)    // uBlock Origin-specific
AGTreeParser.isAbpSyntax(rule)       // Adblock Plus-specific

Integration Points

RuleUtils

RuleUtils now uses AGTree internally while maintaining the same public API:

// These methods now use AGTree parsing internally:
RuleUtils.isComment(ruleText)
RuleUtils.isAllowRule(ruleText)
RuleUtils.isEtcHostsRule(ruleText)
RuleUtils.loadAdblockRuleProperties(ruleText)
RuleUtils.loadEtcHostsRuleProperties(ruleText)

// New AGTree-powered methods:
RuleUtils.parseToAST(ruleText)       // Get raw AST
RuleUtils.isValidRule(ruleText)      // Check parseability
RuleUtils.isNetworkRule(ruleText)    // Network rule check
RuleUtils.isCosmeticRule(ruleText)   // Cosmetic rule check
RuleUtils.detectSyntax(ruleText)     // Syntax detection

ValidateTransformation

The validation transformation uses AGTree for robust rule validation:

Parses rules once and reuses the AST
Uses structured type checking instead of regex
Validates modifiers against AGTree's compatibility tables
Properly handles all rule categories (network, host, cosmetic, comment)
Provides better error messages with context

// Before: String-based validation
if (RuleUtils.isEtcHostsRule(ruleText)) {
    return this.validateEtcHostsRule(ruleText);
}

// After: AST-based validation  
if (AGTreeParser.isHostRule(ast)) {
    return this.validateHostRule(ast as HostRule, ruleText);
}

Configuration

AGTree is configured in deno.json:

{
    "imports": {
        "@adguard/agtree": "npm:@adguard/agtree@^3.4.3"
    }
}

Performance Considerations

Parsing Once: Parse each rule once and pass the AST to multiple validation functions
Tolerant Mode: Use tolerant: true to get InvalidRule nodes instead of exceptions
Include Raws: Use includeRaws: true to preserve original rule text in AST

const DEFAULT_PARSER_OPTIONS: ParserOptions = {
    parseHostRules: true,
    includeRaws: true,
    tolerant: true,
};

Error Handling

AGTree provides structured error information:

const result = AGTreeParser.parse(ruleText);

if (!result.success) {
    console.log(result.error);    // Error message
    console.log(result.ruleText); // Original rule
    
    // In tolerant mode, ast may be an InvalidRule
    if (result.ast?.category === RuleCategory.Invalid) {
        // Access error details from the InvalidRule node
    }
}

Supported Rule Types

AGTree supports parsing all major adblock rule types:

Network Rules

Basic blocking: ||example.com^
Exception: @@||example.com^
With modifiers: ||example.com^$third-party,script

Host Rules

Standard: 127.0.0.1 example.com
Multiple hosts: 0.0.0.0 ad1.com ad2.com
With comments: 127.0.0.1 example.com # block ads

Cosmetic Rules

Element hiding: example.com##.ad-banner
Extended CSS: example.com#?#.ad:has(> .text)
CSS injection: example.com#$#.ad { display: none !important; }
Scriptlet injection: example.com#%#//scriptlet('abort-on-property-read', 'ads')

AdBlock Compiler