Objects & Base Class - Speckle Docs

The Base Class

Every Speckle object inherits from Base. This is the foundation of Speckle’s object model - it provides identity, serialization, type checking, and dynamic properties.

from specklepy.objects import Base
from specklepy.objects.geometry import Point, Line

# Every geometry object is a Base
point = Point(x=1.0, y=2.0, z=3.0)
print(isinstance(point, Base))  # True

# You can create plain Base objects
obj = Base()
obj.name = "My Object"
obj.value = 42

Core Concepts

1. Static vs Dynamic Properties

Base objects support both typed properties (defined in the class) and dynamic properties (added at runtime):

from specklepy.objects.geometry import Point

point = Point(x=1.0, y=2.0, z=3.0)

# Static (typed) properties - defined in Point class
print(point.x)  # 1.0
print(point.y)  # 2.0
print(point.z)  # 3.0

# Dynamic properties - added at runtime
point.color = "red"
point.timestamp = "2024-01-15"
point.metadata = {"source": "sensor", "accuracy": 0.01}

# Get all property names
print(point.get_member_names())
# ['x', 'y', 'z', 'color', 'timestamp', 'metadata', ...]

# Get only typed properties
print(point.get_typed_member_names())
# ['x', 'y', 'z', 'units', ...]

# Get only dynamic properties
print(point.get_dynamic_member_names())
# ['color', 'timestamp', 'metadata']

Why dynamic properties? They let connectors attach application-specific data without defining custom classes. Revit can add parameters, Rhino can add userData, etc.

2. Type Checking

Base performs runtime type checking on typed properties:

from specklepy.objects.geometry import Point

point = Point(x=1.0, y=2.0, z=3.0)

# This works - correct type
point.x = 5.0

# This works - int converts to float
point.x = 5

# This fails - type checking prevents invalid types
try:
    point.x = "not a number"
except Exception as e:
    print(f"Error: {e}")
    # Error: Cannot set 'Point.x': it expects type 'float', 
    # but received type 'str'

Type checking only validates the top-level type. For List[Point], it checks if it’s a list but doesn’t validate each item’s type (for performance).

3. Identity & Hashing

Every Base object has an id - a unique hash of its content:

from specklepy.objects.geometry import Point

point1 = Point(x=1.0, y=2.0, z=3.0)
point2 = Point(x=1.0, y=2.0, z=3.0)
point3 = Point(x=1.0, y=2.0, z=4.0)

# Same content = same id
print(point1.get_id() == point2.get_id())  # True

# Different content = different id
print(point1.get_id() == point3.get_id())  # False

get_id() serializes the entire object - expensive for large objects! The id property is set during send/receive operations, so use that when available.

4. The speckle_type

Every object has a speckle_type that identifies its class across platforms:

from specklepy.objects.geometry import Point, Line
from specklepy.objects import Base

print(Point().speckle_type)  # "Objects.Geometry.Point"
print(Line().speckle_type)   # "Objects.Geometry.Line"
print(Base().speckle_type)   # "Base"

# speckle_type is protected - you can't change it
point = Point()
point.speckle_type = "Something"  # Silently ignored (by design)
print(point.speckle_type)  # Still "Objects.Geometry.Point"

The speckle_type ensures objects are correctly reconstructed when received in other platforms (C#, TypeScript, etc.).

Working with Properties

Setting Properties

Three ways to set properties:

from specklepy.objects import Base

# 1. Attribute assignment (most common)
obj = Base()
obj.name = "Widget"
obj.count = 5

# 2. Dictionary-style access
obj["description"] = "A useful widget"
obj["tags"] = ["important", "new"]

# 3. During initialization (typed properties only)
from specklepy.objects.geometry import Point
point = Point(x=1.0, y=2.0, z=3.0)

Getting Properties

from specklepy.objects import Base

obj = Base()
obj.name = "Widget"
obj.tags = ["a", "b"]

# Attribute access
print(obj.name)  # "Widget"

# Dictionary-style access
print(obj["tags"])  # ["a", "b"]

# Safe access with hasattr
if hasattr(obj, "description"):
    print(obj.description)

# Get with default
description = obj.__dict__.get("description", "No description")

Validating Property Names

Some property names are invalid:

from specklepy.objects import Base

obj = Base()

# These fail
try:
    obj[""] = "value"  # Empty string
except ValueError as e:
    print(e)  # "Invalid Name: Base member names cannot be empty strings"

try:
    obj["@@invalid"] = "value"  # Multiple @
except ValueError as e:
    print(e)  # "Invalid Name: Base member names cannot start with more than one '@'"

try:
    obj["has.dot"] = "value"  # Contains dot
except ValueError as e:
    print(e)  # "Invalid Name: Base member names cannot contain characters '.' or '/'"

Properties starting with @ (single) are valid and used for detached references like @displayValue.

Nested Objects

Base objects can contain other Base objects:

from specklepy.objects import Base
from specklepy.objects.geometry import Point, Line

# Create a container object
building = Base()
building.name = "Office Building"
building.location = Point(x=0, y=0, z=0)

# Nest objects in lists
building.columns = [
    Line(start=Point(x=0, y=0, z=0), end=Point(x=0, y=0, z=3)),
    Line(start=Point(x=5, y=0, z=0), end=Point(x=5, y=0, z=3)),
    Line(start=Point(x=10, y=0, z=0), end=Point(x=10, y=0, z=3)),
]

# Nest objects in dicts
building.metadata = {
    "origin": Point(x=0, y=0, z=0),
    "designer": Base(name="Jane Smith", company="Acme Corp")
}

# Count all nested objects
print(building.get_children_count())  # Counts all Base objects in the tree

Traversing Nested Objects

from specklepy.objects import Base

def traverse_object(obj, depth=0):
    """Recursively traverse a Base object tree."""
    indent = "  " * depth
    
    if isinstance(obj, Base):
        print(f"{indent}{obj.speckle_type}")
        
        # Traverse all properties
        for name in obj.get_member_names():
            if name.startswith("_"):
                continue
            value = getattr(obj, name, None)
            print(f"{indent}  {name}:")
            traverse_object(value, depth + 2)
    
    elif isinstance(obj, list):
        print(f"{indent}[List with {len(obj)} items]")
        for item in obj[:3]:  # Show first 3
            traverse_object(item, depth + 1)
    
    elif isinstance(obj, dict):
        print(f"{indent}{{Dict with {len(obj)} keys}}")
        for key, value in list(obj.items())[:3]:  # Show first 3
            print(f"{indent}  {key}:")
            traverse_object(value, depth + 2)
    
    else:
        # Primitive value
        print(f"{indent}{repr(obj)}")

# Use it
from specklepy.objects.geometry import Line, Point

line = Line(
    start=Point(x=0, y=0, z=0),
    end=Point(x=10, y=10, z=10)
)
line.color = "red"
line.thickness = 2.5

traverse_object(line)

Creating Custom Objects

You can create custom Base subclasses:

from specklepy.objects import Base
from typing import List, Optional

class Wall(Base, speckle_type="MyApp.Wall"):
    """Custom wall object with typed properties."""
    
    # Typed properties
    height: float
    width: float
    thickness: float
    material: Optional[str] = None
    layers: Optional[List[Base]] = None
    
    def __init__(self, height: float, width: float, thickness: float, **kwargs):
        super().__init__(**kwargs)
        self.height = height
        self.width = width
        self.thickness = thickness

# Use it
wall = Wall(height=3.0, width=5.0, thickness=0.2)
wall.material = "Concrete"
wall.fireRating = "2 hour"  # Dynamic property

print(wall.speckle_type)  # "MyApp.Wall"
print(wall.get_typed_member_names())  # ['height', 'width', 'thickness', 'material', 'layers']
print(wall.get_dynamic_member_names())  # ['fireRating']

Custom types must have unique speckle_type names. Use a namespace prefix like "MyApp.Wall" to avoid conflicts.

Advanced Features

Chunkable Properties

Large arrays can be chunked for efficient serialization:

from specklepy.objects import Base

obj = Base()
obj.vertices = list(range(100000))  # Large list

# Mark vertices as chunkable with chunk size 10000
obj.add_chunkable_attrs(vertices=10000)

# When sent, vertices will be split into 10 chunks

Detachable Properties

Large nested objects can be detached and stored separately:

from specklepy.objects import Base
from specklepy.objects.geometry import Mesh

obj = Base()
obj.displayValue = Mesh(...)  # Large mesh

# Mark displayValue as detachable
obj.add_detachable_attrs({"displayValue"})

# When sent, displayValue is stored separately and referenced by id

Connectors use detachment for displayValue meshes - keeps the main object lightweight while allowing on-demand mesh loading.

The applicationId

Link objects across sends with applicationId:

from specklepy.objects.geometry import Point

# First send
point = Point(x=1, y=2, z=3)
point.applicationId = "wall-column-base-001"
# ... send to Speckle ...

# Later update - same applicationId
point = Point(x=1.5, y=2, z=3)  # Moved slightly
point.applicationId = "wall-column-base-001"  # Same ID
# ... send again ...

# Receiving applications can track that this is an update to the same object

Connectors use applicationId to map Speckle objects back to their native application objects (like Revit element IDs).

Common Patterns

Building Collections

from specklepy.objects import Base
from specklepy.objects.geometry import Point

# Create a collection
collection = Base()
collection.name = "Survey Points"
collection.points = []

# Add items
for i in range(10):
    point = Point(x=i, y=i*2, z=0)
    point.label = f"Point {i}"
    collection.points.append(point)

collection.count = len(collection.points)

Filtering by Type

from specklepy.objects import Base
from specklepy.objects.geometry import Point, Line

def get_objects_by_class(obj, target_class):
    """Recursively find all objects of a specific class (recommended approach)."""
    results = []
    
    if isinstance(obj, target_class):
        results.append(obj)
    
    if isinstance(obj, Base):
        # Search in all properties
        for name in obj.get_member_names():
            if name.startswith("_"):
                continue
            value = getattr(obj, name, None)
            results.extend(get_objects_by_class(value, target_class))
    
    elif isinstance(obj, (list, tuple)):
        for item in obj:
            results.extend(get_objects_by_class(item, target_class))
    
    elif isinstance(obj, dict):
        for value in obj.values():
            results.extend(get_objects_by_class(value, target_class))
    
    return results

# Use it
collection = Base()
collection.items = [
    Point(x=1, y=2, z=3),
    Line(start=Point(x=0, y=0, z=0), end=Point(x=1, y=1, z=1)),
    Point(x=4, y=5, z=6),
]

# Recommended: Use isinstance checks
points = get_objects_by_class(collection, Point)
lines = get_objects_by_class(collection, Line)
print(f"Found {len(points)} points")  # Found 4 points (including nested in Line)

For BIM Data: In v3, most BIM objects are DataObject instances. To differentiate walls from columns, filter by properties (e.g., properties.category == "Walls") rather than by type. See Data Traversal for property-based filtering patterns.

Copying Objects

from specklepy.objects.geometry import Point

# Shallow copy - shares nested objects
point1 = Point(x=1, y=2, z=3)
point2 = Point(**point1.__dict__)

# Deep copy - duplicates nested objects
import copy
point3 = copy.deepcopy(point1)

Best Practices

Use typed properties when possible

Define class properties with types for validation and documentation:

# Good - typed property
class Wall(Base):
    height: float

# Less good - dynamic property
wall = Base()
wall.height = 3.0  # No type checking

Avoid expensive operations in loops

Don’t call get_id() repeatedly:

# Bad - serializes every time
for obj in objects:
    if obj.get_id() in seen:
        continue

# Good - use id property or track differently
for obj in objects:
    if obj.id in seen:  # Set during send/receive
        continue

Use applicationId for tracking

Always set applicationId when creating objects from your application:

from specklepy.objects.geometry import Point

# Link to your app's native object
point = Point(x=1, y=2, z=3)
point.applicationId = f"myapp-{native_object.id}"

Handle unknown types gracefully

When traversing received objects, expect unknown types:

def process_object(obj):
    if isinstance(obj, Base):
        # Check speckle_type string instead of isinstance
        if obj.speckle_type.startswith("Objects.Geometry"):
            process_geometry(obj)
        elif hasattr(obj, "displayValue"):
            process_display_value(obj.displayValue)

Summary

The Base class is powerful because it:

✅ Provides identity - Every object has a unique hash
✅ Supports dynamic properties - Attach any data without custom classes
✅ Type checks typed properties - Catch errors early
✅ Works cross-platform - speckle_type ensures interoperability
✅ Handles nested objects - Build complex hierarchies naturally
✅ Optimizes serialization - Chunking and detachment for large data

Understanding Base is fundamental to working effectively with specklepy!

Next Steps

Data Traversal

Learn how to navigate and extract data from object graphs

Geometry Objects

Explore Points, Lines, Meshes, and other geometry types

Custom Objects

Create your own Base subclasses for domain-specific data

Display Values

Understand how geometry is made visible and interoperable

Python

​The Base Class

​Core Concepts

​1. Static vs Dynamic Properties

​2. Type Checking

​3. Identity & Hashing

​4. The speckle_type

​Working with Properties

​Setting Properties

​Getting Properties

​Validating Property Names

​Nested Objects

​Traversing Nested Objects

​Creating Custom Objects

​Advanced Features

​Chunkable Properties

​Detachable Properties

​The applicationId

​Common Patterns

​Building Collections

​Filtering by Type

​Copying Objects

​Best Practices

​Summary

​Next Steps

Data Traversal

Geometry Objects

Custom Objects

Display Values

The Base Class

Core Concepts

1. Static vs Dynamic Properties

2. Type Checking

3. Identity & Hashing

4. The speckle_type

Working with Properties

Setting Properties

Getting Properties

Validating Property Names

Nested Objects

Traversing Nested Objects

Creating Custom Objects

Advanced Features

Chunkable Properties

Detachable Properties

The applicationId

Common Patterns

Building Collections

Filtering by Type

Copying Objects

Best Practices

Summary

Next Steps