COMB UUIDs

COMB (Combined UUID-Timestamp) is a UUID v4 variant where the last 6 bytes encode a millisecond timestamp. This produces UUIDs that are mostly sequential when used as clustered primary keys in SQL Server, dramatically reducing index fragmentation.

Usage

import { newCOMB } from "@tsdrivers/mssql";

const id = newCOMB();
// e.g. "a3f1b2c4-d5e6-4f7a-8b9c-019437a2b1c0"

Why COMB?

SQL Server's NEWID() generates fully random GUIDs that cause heavy page splits on clustered indexes. NEWSEQUENTIALID() is sequential but only works as a column default — you can't generate them client-side.

COMB gives you:

• Client-side generation — no database round-trip needed
• Sequential ordering — last 48 bits encode current time, so inserts are mostly append-only
• Uniqueness — first 80 bits are random (same as UUID v4)
• Standard format — valid UUID v4, works with uniqueidentifier columns

With Bulk Insert

const rows = users.map((u) => [newCOMB(), u.name, u.email]);

await cn.bulk("Users")
  .columns([
    { name: "Id", type: "uniqueidentifier" },
    { name: "Name", type: "nvarchar" },
    { name: "Email", type: "nvarchar" },
  ])
  .rows(rows)
  .execute();