edwh-uuid7

A Python library that implements UUIDv7 generation and parsing.
It supports conversion to and from datetime values, ensuring full sorting compatibility with UUIDs produced by PostgreSQL’s pg_uuidv7 extension, PostgreSQL 18’s built-in uuid7, and Python 3.14’s uuid.uuid7.
Unlike the standard uuid.uuid7, this library also allows generating UUIDs for specific timestamps.

Overview

UUIDv7 is a time-sortable UUID format composed of a timestamp and random bits. Our implementation follows the UUIDv7 specification, which provides flexibility in ensuring monotonicity. We utilize microseconds for sub-millisecond precision within the timestamp; this acts as a stateless monotonic counter for UUIDs generated within the same millisecond and leaves more room for entropy compared to using nanoseconds.

UUIDv7 Layout (128 bits)

Bits	Size (bits)	Field	Description
0–47	48	timestamp (ms)	Unix timestamp in milliseconds (sortable)
48–51	4	version	UUID version (0b0111 for v7)
52–71	20	sub-ms (μs)	Sub-millisecond fraction (0–999,999)
72–127	56	random	Random bits (includes UUID variant information)

This design makes UUIDv7 globally sortable by creation time (1 μs resolution) while retaining enough entropy for distributed uniqueness.

Key Features of edwh-uuid7:

• Compatible with pg_uuidv7’s uuid_generate_v7(), uuid_v7_to_timestamptz(), uuid_timestamptz_to_v7() functions.
• Sorting compatible with Python 3.14's uuid.uuid7()
• Provides conversion between UUIDv7 and datetime.
• Supports timezone-aware and naive datetime conversions.

Why this package?

The widely-used uuid7 Python package uses a different timestamp scheme, which is not compatible with sorting behavior as implemented in the PostgreSQL pg_uuidv7 extension. This package ensures compatibility, ensuring that UUIDv7s generated will sort correctly alongside those produced by pg_uuidv7.

Comparison with Python 3.14's uuid7 functionality

Key differences between this library and Python 3.14's built-in uuid7:

• Timestamp Implementation: This library works with μs (microseconds), while Python 3.14+ uses a counter mechanism
• Conversion Utilities: This library provides uuid7 ↔ datetime conversion, which isn't available in the standard library implementation (at the time of writing)
• PostgreSQL Compatibility: Both implementations are sorting compatible with the pg_uuidv7 PostgreSQL extension

Installation

pip install edwh-uuid7

Usage

Here are examples of how to use edwh-uuid7:

Generating a UUIDv7

from edwh_uuid7 import uuid7  # or `uuid_generate_v7`

uuid = uuid7()
print(f"Generated UUIDv7: {uuid}")

Converting a UUIDv7 to a datetime

from edwh_uuid7 import uuid7, uuid7_to_datetime       # or uuid_generate_v7, uuid_v7_to_timestamptz

uuid = uuid7()                                        # UUID('0196905b-4434-74ef-aa27-442be7069beb')
datetime_value = uuid7_to_datetime(uuid)              # datetime.datetime(2025, 5, 2, 9, 37, 2, 516000, tzinfo=datetime.timezone.utc)
print(f"Corresponding datetime: {datetime_value}")    # 2025-05-02 09:37:02.516000+00:00

High-precision mode

This library provides enhanced timestamp precision capabilities:

• Default Precision: By default, uuid7 supports millisecond precision
• Enhanced Recovery: When working with UUIDs created by this library (uuid7() or uuid7(timestamp_ns=...)), you can enable high_precision=True to recover datetime values with microsecond accuracy

Example:

now = datetime.fromtimestamp(time.time())     # datetime.datetime(2025, 5, 9, 17, 53, 50, 514106)
uuid = datetime_to_uuid7(now)                 # UUID('0196b5c0-c0b2-719e-802c-9964c4992a66')
uuid7_to_datetime(uuid)                       # datetime(2025, 5, 9, 15, 53, 50, 514000, tzinfo=datetime.timezone.utc)
uuid7_to_datetime(uuid, high_precision=True)  # datetime(2025, 5, 9, 15, 53, 50, 514106, tzinfo=datetime.timezone.utc)

Converting a datetime to a UUIDv7

from edwh_uuid7.core import datetime_to_uuid7                      # or uuid_timestamptz_to_v7
from datetime import datetime, timezone

dt = datetime(2025, 5, 2, 9, 37, 2, 516000, tzinfo=timezone.utc)
uuid_from_dt = datetime_to_uuid7(dt)                               # UUID('0196905b-4434-7000-8012-620314c8b88b')
print(f"UUIDv7 from datetime: {uuid_from_dt}")                     # 0196905b-4434-7000-8012-620314c8b88b

Extracting a timezone-aware datetime

from edwh_uuid7 import uuid7, uuid7_to_datetime                            # or uuid_generate_v7, uuid_v7_to_timestamptz
from zoneinfo import ZoneInfo

uuid = uuid7()                                                             # UUID('0196905f-0339-769a-9f0b-8fedc645a688')
datetime_value = uuid7_to_datetime(uuid, tz=ZoneInfo("Europe/Amsterdam"))  # datetime.datetime(2025, 5, 2, 11, 41, 8, 25000, tzinfo=zoneinfo.ZoneInfo(key='Europe/Amsterdam'))
print(f"Generated UUID: {uuid}")                                           # 0196905f-0339-769a-9f0b-8fedc645a688
print(f"Timezone-aware datetime: {datetime_value}")                        # 2025-05-02 11:41:08.025000+02:00

Caveats

While microseconds are used to ensure correct sorting, accuracy is limited (truncated) when converting back to datetime, consistent with the behavior of the pg_uuidv7 function uuid_timestamptz_to_v7.

For example:

Input: 2025-05-02 08:46:48.307329+00:00
UUID conversion: 0196902d-45f3-7505-8038-7795a5ec9a1b (example)
Converted Back: 2025-05-02 08:46:48.307000+00:00

This truncation is by design (in favor of entropy) and aligns with the PostgreSQL pg_uuidv7 implementation.