MicroShard UUID (Python Implementation)

A zero-dependency, Partition-Aware UUIDv8 library for Python.

Unlike standard UUIDs (v4/v7), MicroShard embeds a 32-bit Shard ID directly into the 128-bit identifier. This enables Zero-Lookup Routing, allowing your application to determine the database shard, tenant, or region of a record simply by decoding its Primary Key.

📦 Features

• Zero-Lookup Routing: Extract Shard/Tenant IDs instantly from the UUID.
• Microsecond Precision: 54-bit timestamp ensures strict chronological sorting.
• Massive Scale: Supports 4.29 Billion unique Shards/Tenants.
• Collision Resistant: 36 bits of randomness per microsecond per shard.
• Stateless & Stateful: Supports both functional usage and class-based generators.
• Standard Library Only: No external dependencies (uses os, struct, uuid).

🛠 Installation

Requires Python 3.6+.

📐 Architecture: The 54/32/36 Split

MicroShard utilizes the UUIDv8 (Custom) format defined in IETF RFC 9562.

🚀 Usage

1. Stateless Generation (Functional)

Best for simple scripts or when the Shard ID changes dynamically per request.

from microshard_uuid import generate, get_shard_id

# 1. Generate an ID for Shard #500
uid = generate(shard_id=500)
print(f"Generated: {uid}")

# 2. Extract Shard ID (Routing)
target_shard = uid.get_shard_id()
assert target_shard == 500

2. Stateful Generation (Class-Based)

Best for application configuration or dependency injection where the Shard ID is fixed for the lifecycle of the service.

from microshard_uuid import Generator

# Configure once at startup
id_gen = Generator(default_shard_id=101)

# Generate anywhere in your app without passing the ID
uid = id_gen.new_id()

3. Backfilling Data (Explicit Time)

If you are migrating legacy data, you can generate UUIDs for past timestamps while maintaining correct sorting order.

from datetime import datetime, timezone
from microshard_uuid import from_timestamp

# Create an ID for a specific past event (e.g., Jan 1, 2023)
past_date = datetime(2023, 1, 1, 12, 0, 0, tzinfo=timezone.utc)

legacy_uid = from_timestamp(past_date, shard_id=50)

# Supports integer microseconds as well
legacy_uid_int = from_timestamp(1672574400000000, shard_id=50)

4. Extracting Metadata

You can decode any MicroShard UUID to retrieve its creation time and origin shard.

from microshard_uuid import get_timestamp, get_shard_id

# Option A: Get Python Datetime object
dt = uid.get_timestamp()
# datetime.datetime(2025, 12, 12, 10, 0, 0, 123456, tzinfo=datetime.timezone.utc)

# Option B: Get ISO String (Strict format with Z)
iso = uid.get_iso_timestamp()
# "2025-12-12T10:00:00.123456Z"

# Get Origin Shard
shard = uid.get_shard_id()
print(f"Origin Shard: {shard}")

🛡️ Safety & Performance

Collision Probability

Uniqueness is guaranteed by the combination of: (Timestamp + Shard ID + Randomness).

• Global Uniqueness: Guaranteed if Shard IDs are unique.
• Per-Shard Uniqueness: A single shard has 36 bits of randomness per microsecond.
• Risk: A collision only occurs if a single specific shard generates billions of IDs within a single microsecond. This is physically impossible on current hardware.

Performance

• Generation: ~2-3 microseconds per ID (comparable to uuid.uuid4()).
• Sorting: UUIDs are generated with the Timestamp in the highest bits, ensuring that standard string/byte sorting yields strict chronological order.

🧪 Running Tests

This library includes a comprehensive test suite using unittest.

# Run tests from the implementation folder
make test