Files
core/docs/internal/ast-README.md
T

341 lines
8.1 KiB
Markdown
Raw Normal View History

# AST Module - DSL Compiler for Zanzibar-Graph
This module provides a complete Abstract Syntax Tree (AST) system for parsing and compiling the Evidence DSL into rule configurations that interface with the zanzibar-graph authorization system.
## Overview
The AST module consists of three main components:
1. **Parser** - Converts DSL text into AST nodes
2. **Generator** - Converts AST nodes into rule configurations
3. **Compiler** - Orchestrates the parsing and generation process
## Architecture
```
DSL Text → Parser → AST Nodes → Generator → Rule Configs → setRelationConfig()
```
## Features
- **Complete DSL Support** - Supports all DSL constructs from the specification
- **Modular Design** - Clean separation of concerns with pluggable components
- **Error Handling** - Comprehensive error reporting and validation
- **Program Management** - Support for multiple compiled programs
- **Rule Generation** - Automatic conversion to existing rule system
- **Extensible** - Easy to add new node types and generators
## Quick Start
```javascript
import { DSLCompiler } from './src/ast/index.js';
// Create compiler with arbiter instance
const compiler = new DSLCompiler(arbiter);
// Compile DSL text
const result = compiler.compile(dslText, 'my-program');
if (result.success) {
console.log(`Generated ${result.generatedCount} rules`);
} else {
console.error('Compilation errors:', result.errors);
}
```
## Core Components
### 1. AST Nodes (`/nodes/`)
The AST nodes represent the parsed structure of the DSL:
- **BaseNode** - Base class for all AST nodes
- **ProgramNode** - Root node containing all definitions
- **DefinitionNode** - Type definitions with fields and behaviors
- **FactNode** - Fact definitions with parameters and properties
- **EvidenceNode** - Evidence definitions with bodies
- **MeasureNode** - Measure definitions for value collection
- **ExpressionNode** - Expressions, variables, and literals
- **And many more...**
### 2. Parser (`/parser/`)
The DSL parser converts text into AST nodes:
```javascript
import { DSLParser } from './parser/DSLParser.js';
const parser = new DSLParser();
const program = parser.parse(dslText);
```
### 3. Generator (`/generator/`)
The rule generator converts AST nodes into rule configurations:
```javascript
import { RuleGenerator } from './generator/RuleGenerator.js';
const generator = new RuleGenerator(arbiter);
const result = generator.generateRules(program);
```
### 4. Compiler (`DSLCompiler.js`)
The main compiler orchestrates the entire process:
```javascript
import { DSLCompiler } from './DSLCompiler.js';
const compiler = new DSLCompiler(arbiter);
const result = compiler.compile(dslText, 'program-name');
```
## Usage Examples
### Basic Compilation
```javascript
const dsl = `
definition User {
role: string
isActive: boolean
}
fact hasRole(user: User, role: string) CACHE eager
evidence canRead(user: User, doc: Document) {
hasRole(user, 'admin')
}
`;
const result = compiler.compile(dsl, 'auth');
```
### Complex DSL with Defeasible Logic
```javascript
const complexDSL = `
evidence canAccessCritical(user: User, resource: Resource) {
// Strict requirement
ALWAYS user.isActive
// Defeasible access
WHEN hasRole(user, 'admin') UNLESS isSuspended(user)
// Requirements
REQUIRES hasClearance(user, resource.level)
// Fusion evidence
fusion majority {
user.isTrusted
user.hasRecentActivity
}
}
`;
const result = compiler.compile(complexDSL, 'critical-access');
```
### Multiple Programs
```javascript
const programs = {
'auth': `
fact hasRole(user: User, role: string)
evidence canRead(user: User, doc: Document) {
hasRole(user, 'admin')
}
`,
'finance': `
fact hasBalance(user: User, amount: number)
evidence canWithdraw(user: User, amount: number) {
hasBalance(user, amount)
}
`
};
const result = compiler.compileMultiple(programs);
```
## API Reference
### DSLCompiler
#### Methods
- `compile(dslText, programName)` - Compile DSL text into rules
- `compileMultiple(programs)` - Compile multiple programs
- `validate(dslText)` - Validate DSL without compilation
- `getCompiledProgram(name)` - Get compiled program by name
- `getAllCompiledPrograms()` - Get all compiled programs
- `removeCompiledProgram(name)` - Remove compiled program
- `clearCompiledPrograms()` - Clear all programs
- `getCompilationStats()` - Get compilation statistics
#### Properties
- `arbiter` - The arbiter instance
- `parser` - The DSL parser
- `generator` - The rule generator
### DSLParser
#### Methods
- `parse(dslText)` - Parse DSL text into AST
- `getErrors()` - Get parser errors
### RuleGenerator
#### Methods
- `generateRules(program)` - Generate rules from AST
- `getErrors()` - Get generator errors
- `getGeneratedRules()` - Get generated rules
## Supported DSL Constructs
### Type Definitions
```typescript
definition User {
role: string
isActive: boolean
lastActive: timestamp BEHAVES {
decaying down hourly
} CACHE lazy
}
```
### Facts
```typescript
fact hasRole(user: User, role: string) CACHE eager
fact isMember(user: User, group: Group) transitive CACHE lazy
```
### Evidence
```typescript
evidence canRead(user: User, doc: Document) {
// Direct evidence
owns(user, doc)
// Pattern matching
isMember(user, *group) {
canRead(group, doc)
} limit 5
// Defeasible logic
WHEN hasRole(user, 'admin') UNLESS isSuspended(user)
}
```
### Measures
```typescript
measure userBalance(user: User) {
user.balance
} PROVIDES number
measure userPermissions(user: User) {
fusion max {
user.role.permissions
user.group.permissions
}
} PROVIDES Permission[]
```
## Error Handling
The compiler provides comprehensive error handling:
```javascript
const result = compiler.compile(dslText);
if (!result.success) {
console.error('Compilation failed:');
result.errors.forEach(error => console.error(` - ${error}`));
}
if (result.warnings.length > 0) {
console.warn('Warnings:');
result.warnings.forEach(warning => console.warn(` - ${warning}`));
}
```
## Testing
Run the test suite:
```javascript
import { runDSLCompilerTests } from './tests/DSLCompiler.test.js';
const testResults = runDSLCompilerTests(arbiter);
console.log('Test Results:', testResults);
```
## Examples
See the examples directory for comprehensive usage examples:
- `examples/DSLExample.js` - Complete usage examples
- `tests/DSLCompiler.test.js` - Test suite
## Integration with Existing System
The AST module integrates seamlessly with the existing zanzibar-graph system:
1. **Parser** converts DSL text to AST nodes
2. **Generator** converts AST nodes to rule configurations
3. **Compiler** applies rules via `arbiter.setRelationConfig()`
The generated rules are compatible with all existing rule types:
- Direct rules
- Computed rules
- Parent rules
- Tuple-to-userset rules
- Similarity rules
- Multi-hop rules
- Logical operators
- Defeasible logic
## Extensibility
The AST system is designed to be extensible:
1. **Add new node types** by extending `BaseNode`
2. **Add new parsers** by extending `DSLParser`
3. **Add new generators** by extending `RuleGenerator`
4. **Add new compilers** by extending `DSLCompiler`
## Performance Considerations
- **Lazy evaluation** - AST nodes are created on demand
- **Efficient parsing** - Token-based parsing with minimal memory usage
- **Rule caching** - Generated rules are cached for reuse
- **Batch processing** - Support for compiling multiple programs at once
## Future Enhancements
- **Incremental compilation** - Only recompile changed parts
- **Parallel processing** - Compile multiple programs in parallel
- **Advanced optimizations** - Rule optimization and simplification
- **IDE support** - Language server protocol support
- **Visualization** - AST visualization tools
## Contributing
When contributing to the AST module:
1. Follow the existing code structure
2. Add comprehensive tests for new features
3. Update documentation
4. Ensure backward compatibility
5. Follow the established patterns
## License
This module is part of the zanzibar-graph project and follows the same license terms.