jentic-openapi-traverse

A Python library for traversing OpenAPI documents. This package is part of the Jentic OpenAPI Tools ecosystem and provides two types of traversal:

1. Datamodel Traversal - OpenAPI-aware semantic traversal with visitor pattern
2. JSON Traversal - Generic depth-first traversal of JSON-like structures

Installation

pip install jentic-openapi-traverse

Prerequisites:

• Python 3.11+

Datamodel Traversal

OpenAPI-aware semantic traversal using the visitor pattern. Works with low-level datamodels from jentic-openapi-datamodels package, preserving source location information and providing structured access to OpenAPI nodes.

Quick Start

from jentic.apitools.openapi.parser.core import OpenAPIParser
from jentic.apitools.openapi.traverse.datamodels.low import traverse, DataModelLowVisitor

# Parse OpenAPI document
parser = OpenAPIParser("datamodel-low")
doc = parser.parse("file:///path/to/openapi.yaml")

# Create visitor
class OperationCollector(DataModelLowVisitor):
    def __init__(self):
        self.operations = []

    def visit_Operation(self, path):
        self.operations.append({
            "path": path.format_path(path_format="jsonpointer"),
            "operation_id": path.node.operation_id.value if path.node.operation_id else None
        })

# Traverse
visitor = OperationCollector()
traverse(doc, visitor)

print(f"Found {len(visitor.operations)} operations")

Visitor Pattern

The datamodel traversal uses a flexible visitor pattern with multiple hook types:

Hook Methods

Generic hooks (fire for ALL nodes):

• visit_enter(path) - Called before processing any node
• visit_leave(path) - Called after processing any node and its children

Class-specific hooks (fire for matching node types):

• visit_ClassName(path) - Main visitor for specific node type
• visit_enter_ClassName(path) - Called before visit_ClassName
• visit_leave_ClassName(path) - Called after children are visited

Dispatch Order

For each node, hooks are called in this order:

1. visit_enter(path) - generic enter
2. visit_enter_ClassName(path) - specific enter
3. visit_ClassName(path) - main visitor
4. [child traversal - automatic unless skipped]
5. visit_leave_ClassName(path) - specific leave
6. visit_leave(path) - generic leave

Control Flow

Visitor methods control traversal by their return value:

• None (or no return) - Continue normally (children visited automatically)
• False - Skip children of this node (but continue to siblings)
• BREAK - Stop entire traversal immediately

from jentic.apitools.openapi.traverse.datamodels.low import traverse, BREAK

class ControlFlowExample:
    def visit_PathItem(self, path):
        if path.parent_key == "/internal":
            return False  # Skip internal paths and their children

    def visit_Operation(self, path):
        if some_error_condition:
            return BREAK  # Stop entire traversal

NodePath Context

Every visitor method receives a NodePath object with context about the current node:

class PathInspector:
    def visit_Operation(self, path):
        # Current node
        print(f"Node: {path.node.__class__.__name__}")

        # Parent information
        print(f"Parent field: {path.parent_field}")  # e.g., "get", "post"
        print(f"Parent key: {path.parent_key}")      # e.g., "/users" (for path items)

        # Ancestry (computed properties)
        print(f"Parent: {path.parent.__class__.__name__}")
        print(f"Ancestors: {len(path.ancestors)}")
        root = path.get_root()

        # Complete path formatting (RFC 6901 JSONPointer / RFC 9535 JSONPath)
        print(f"JSONPointer: {path.format_path()}")
        # Output: /paths/~1users/get

        print(f"JSONPath: {path.format_path(path_format='jsonpath')}")
        # Output: $['paths']['/users']['get']

Path Reconstruction

NodePath uses a linked chain structure (parent_path) internally to preserve complete path information from root to current node. This enables accurate JSONPointer and JSONPath reconstruction:

class PathFormatter:
    def visit_Response(self, path):
        # Complete paths from root to current node
        pointer = path.format_path()
        # /paths/~1users/get/responses/200

        jsonpath = path.format_path(path_format='jsonpath')
        # $['paths']['/users']['get']['responses']['200']

Special handling for patterned fields:

• Patterned fields like Paths.paths don't duplicate in paths: /paths/{key} (not /paths/paths/{key})
• Fixed dict fields like webhooks, callbacks, schemas include their field name: /webhooks/{key}, /components/schemas/{key}

Computed properties:

• path.parent - Returns parent node (computed from parent_path chain)
• path.ancestors - Returns tuple of ancestor nodes from root to parent (computed on access)

Enter/Leave Hooks

Use enter/leave hooks for pre/post processing logic:

class DepthTracker(DataModelLowVisitor):
    def __init__(self):
        self.current_depth = 0
        self.max_depth = 0

    def visit_enter(self, path):
        self.current_depth += 1
        self.max_depth = max(self.max_depth, self.current_depth)
        print("  " * self.current_depth + f"Entering {path.node.__class__.__name__}")

    def visit_leave(self, path):
        print("  " * self.current_depth + f"Leaving {path.node.__class__.__name__}")
        self.current_depth -= 1

Examples

Collecting All Schemas

class SchemaCollector(DataModelLowVisitor):
    def __init__(self):
        self.schemas = {}

    def visit_Schema(self, path):
        schema_name = path.parent_key if path.parent_field == "schemas" else None
        if schema_name:
            self.schemas[schema_name] = path.node

visitor = SchemaCollector()
traverse(doc, visitor)
print(f"Found {len(visitor.schemas)} schemas")

Validating Security Requirements

class SecurityValidator(DataModelLowVisitor):
    def __init__(self):
        self.errors = []

    def visit_Operation(self, path):
        if not path.node.security:
            self.errors.append(f"Missing security at {path.format_path()}")

    def visit_SecurityRequirement(self, path):
        # Validate security requirement
        if not path.node.schemes:
            self.errors.append(f"Empty security requirement at {path.format_path()}")

visitor = SecurityValidator()
traverse(doc, visitor)
if visitor.errors:
    for error in visitor.errors:
        print(f"Security error: {error}")

Finding Deprecated Operations

class DeprecatedFinder:
    def __init__(self):
        self.deprecated_ops = []

    def visit_Operation(self, path):
        if path.node.deprecated and path.node.deprecated.value:
            self.deprecated_ops.append({
                "path": path.format_path(),
                "operation_id": path.node.operation_id.value if path.node.operation_id else None,
                "method": path.parent_field
            })
        return False  # Skip children (we don't need to go deeper)

visitor = DeprecatedFinder()
traverse(doc, visitor)

Early Exit on Error

class ErrorDetector(DataModelLowVisitor):
    def __init__(self):
        self.error_found = False
        self.error_location = None

    def visit_Operation(self, path):
        if not path.node.responses:
            self.error_found = True
            self.error_location = path.format_path()
            return BREAK  # Stop traversal immediately

Merging Multiple Visitors

Run multiple visitors in a single traversal pass (parallel visitation) using merge_visitors:

from jentic.apitools.openapi.traverse.datamodels.low import merge_visitors

# Create separate visitors
schema_collector = SchemaCollector()
security_validator = SecurityValidator()
deprecated_finder = DeprecatedFinder()

# Merge and traverse once
merged = merge_visitors(schema_collector, security_validator, deprecated_finder)
traverse(doc, merged)

# Each visitor maintains independent state
print(f"Schemas: {len(schema_collector.schemas)}")
print(f"Security errors: {len(security_validator.errors)}")
print(f"Deprecated: {len(deprecated_finder.deprecated_ops)}")

Per-Visitor Control Flow:

• Each visitor can independently skip subtrees or break
• If visitor1 returns False, only visitor1 skips children
• Other visitors continue normally
• This follows ApiDOM's per-visitor semantics

Duck Typing

You don't need to inherit from DataModelLowVisitor - duck typing works:

class SimpleCounter:  # No inheritance
    def __init__(self):
        self.count = 0

    def visit_Operation(self, path):
        self.count += 1

visitor = SimpleCounter()
traverse(doc, visitor)

The DataModelLowVisitor base class is optional and provides no functionality - it's purely for organizational purposes.

API Reference

traverse(root, visitor) -> None

Traverse OpenAPI datamodel tree using visitor pattern.

Parameters:

• root - Root datamodel object (OpenAPI30, OpenAPI31, or any datamodel node)
• visitor - Object with visit_* methods (duck typing)

Returns:

• None (traversal is side-effect based)

BREAK

Sentinel value to stop traversal immediately. Return this from any visitor method.

from jentic.apitools.openapi.traverse.datamodels.low import BREAK

def visit_Operation(self, path):
    if should_stop:
        return BREAK

merge_visitors(*visitors) -> object

Merge multiple visitors into one composite visitor.

Parameters:

• *visitors - Variable number of visitor objects

Returns:

• Composite visitor object with per-visitor state tracking

JSON Traversal

Generic depth-first traversal of any JSON-like structure (dicts, lists, scalars). Works with raw parsed OpenAPI documents or any other JSON data.

Quick Start

from jentic.apitools.openapi.traverse.json import traverse

# Traverse a nested structure
data = {
    "openapi": "3.1.0",
    "info": {"title": "My API", "version": "1.0.0"},
    "paths": {
        "/users": {
            "get": {"summary": "List users"}
        }
    }
}

# Walk all nodes
for node in traverse(data):
    print(f"{node.format_path()}: {node.value}")

Output:

openapi: 3.1.0
info: {'title': 'My API', 'version': '1.0.0'}
info.title: My API
info.version: 1.0.0
paths: {'/users': {'get': {'summary': 'List users'}}}
paths./users: {'get': {'summary': 'List users'}}
paths./users.get: {'summary': 'List users'}
paths./users.get.summary: List users

Working with Paths

from jentic.apitools.openapi.traverse.json import traverse

data = {
    "users": [
        {"name": "Alice", "email": "alice@example.com"},
        {"name": "Bob", "email": "bob@example.com"}
    ]
}

for node in traverse(data):
    # Access path information
    print(f"Path: {node.path}")
    print(f"Segment: {node.segment}")
    print(f"Full path: {node.full_path}")
    print(f"Formatted: {node.format_path()}")
    print(f"Depth: {len(node.ancestors)}")
    print()

Custom Path Formatting

for node in traverse(data):
    # Default dot separator
    print(node.format_path())  # e.g., "paths./users.get.summary"

    # Custom separator
    print(node.format_path(separator="/"))  # e.g., "paths//users/get/summary"

Finding Specific Nodes

# Find all $ref references in a document
refs = [
    node.value["$ref"]
    for node in traverse(openapi_doc)
    if isinstance(node.value, dict) and "$ref" in node.value
]

# Find all nodes at a specific path segment
schemas = [
    node.value
    for node in traverse(openapi_doc)
    if node.segment == "schema"
]

# Find deeply nested values
response_descriptions = [
    node.value
    for node in traverse(openapi_doc)
    if node.segment == "description" and "responses" in node.path
]

API Reference

traverse(root: JSONValue) -> Iterator[TraversalNode]

Performs depth-first traversal of a JSON-like structure.

Parameters:

• root: The data structure to traverse (dict, list, or scalar)

Returns:

• Iterator of TraversalNode objects

Yields:

• For dicts: one node per key-value pair
• For lists: one node per index-item pair
• Scalars at root don't yield nodes (but are accessible via parent nodes)

TraversalNode

Immutable dataclass representing a node encountered during traversal.

Attributes:

• path: JSONPath - Path from root to the parent container (tuple of segments)
• parent: JSONContainer - The parent container (dict or list)
• segment: PathSeg - The key (for dicts) or index (for lists) within parent
• value: JSONValue - The actual value at parent[segment]
• ancestors: tuple[JSONValue, ...] - Ordered tuple of values from root down to (but not including) parent

Properties:

• full_path: JSONPath - Complete path from root to this value (path + (segment,))

Methods:

• format_path(separator: str = ".") -> str - Format the full path as a human-readable string

Usage Examples

Collecting All Schemas

from jentic.apitools.openapi.traverse.json import traverse

def collect_schemas(openapi_doc):
    """Collect all schema objects from an OpenAPI document."""
    schemas = []

    for node in traverse(openapi_doc):
        if node.segment == "schema" and isinstance(node.value, dict):
            schemas.append({
                "path": node.format_path(),
                "schema": node.value
            })

    return schemas

Analyzing Document Structure

def analyze_depth(data):
    """Analyze the depth distribution of a document."""
    max_depth = 0
    depth_counts = {}

    for node in traverse(data):
        depth = len(node.ancestors)
        max_depth = max(max_depth, depth)
        depth_counts[depth] = depth_counts.get(depth, 0) + 1

    return {
        "max_depth": max_depth,
        "depth_distribution": depth_counts
    }

Testing

The package includes comprehensive test coverage for JSON traversal:

uv run --package jentic-openapi-traverse pytest packages/jentic-openapi-traverse/tests -v