UUIDv7 functions

URL: llms-txt#uuidv7-functions

Contents:

• Examples
• Functions

UUIDv7 is a time-ordered UUID that includes a Unix timestamp (with millisecond precision) in its first 48 bits. Like other UUIDs, it uses 6 bits for version and variant info, and the remaining 74 bits are random.

UUIDv7 is ideal anywhere you create lots of records over time, not only observability. Advantages are:

• No extra column required to partition by time with sortability: you can sort UUIDv7 instances by their value. This is useful for ordering records by creation time without the need for a separate timestamp column.
• Indexing performance: UUIDv7s increase with time, so new rows append near the end of a B-tree instead of This results in fewer page splits, less fragmentation, faster inserts, and efficient time-range scans.
• Easy keyset pagination: WHERE id > :cursor and natural sharding.
• UUID: safe across services, replicas, and unique across distributed systems.

UUIDv7 also increases query speed by reducing the number of chunks scanned during queries. For example, in a database with 25 million rows, the following query runs in 25 seconds:

Using UUIDv7 excludes chunks at startup and reduces the query time to 550ms:

You use UUIDvs for events, orders, messages, uploads, runs, jobs, spans, and more.

• High-rate event logs for observability and metrics:

UUIDv7 gives you globally unique IDs (for traceability) and time windows (“last hour”), without the need for a separate created_at column. UUIDv7 create less churn because inserts land at the end of the index, and you can filter by time using UUIDv7 objects.

• Last hour:

  • Keyset pagination

• Workflow / durable execution runs:

Each run needs a stable ID for joins and retries, and you often ask “what started since X?”. UUIDs help by serving both as the primary key and a time cursor across services. For example:

• Orders / activity feeds / messages (SaaS apps):

Human-readable timestamps are not mandatory in a table. However, you still need time-ordered pages and day/week ranges. UUIDv7 enables clean date windows and cursor pagination with just the ID. For example:

• [generate_uuidv7()][generate_uuidv7]: generate a version 7 UUID based on current time
• [to_uuidv7()][to_uuidv7]: create a version 7 UUID from a PostgreSQL timestamp
• [to_uuidv7_boundary()][to_uuidv7_boundary]: create a version 7 "boundary" UUID from a PostgreSQL timestamp
• [uuid_timestamp()][uuid_timestamp]: extract a PostgreSQL timestamp from a version 7 UUID
• [uuid_timestamp_micros()][uuid_timestamp_micros]: extract a PostgreSQL timestamp with microsecond precision from a version 7 UUID
• [uuid_version()][uuid_version]: extract the version of a UUID

===== PAGE: https://docs.tigerdata.com/api/approximate_row_count/ =====

Examples:

Example 1 (sql):

WITH ref AS (SELECT now() AS t0)
SELECT count(*) AS cnt_ts_filter
FROM events e, ref
WHERE uuid_timestamp(e.event_id) >= ref.t0 - INTERVAL '2 days';

Example 2 (sql):

WITH ref AS (SELECT now() AS t0)
SELECT count(*) AS cnt_boundary_filter
FROM events e, ref
WHERE e.event_id >= to_uuidv7_boundary(ref.t0 - INTERVAL '2 days')

Example 3 (sql):

SELECT count(*) FROM logs WHERE id >= to_uuidv7_boundary(now() - interval '1 hour');

Example 4 (sql):

SELECT * FROM logs WHERE id > to_uuidv7($last_seen'::timestamptz, true) ORDER BY id LIMIT 1000;