This page is a quick reference checkpoint for LAG in Databricks SQL: behavior, syntax rules, edge cases, and a minimal example; plus the official vendor documentation.
LAG lets you look back at a previous row's value within the same partition.
Offset defaults to 1; returns NULL when offset exceeds partition unless DEFAULT is provided
If this behavior feels unintuitive, the tutorial below explains the underlying pattern step-by-step.
LAG(expr [, offset [, default ]]) OVER ( [PARTITION BY ...] ORDER BY ... )
SELECT LAG(salary, 1, 0) OVER (PARTITION BY department ORDER BY hire_date) AS prev_salary FROM employees
If you came here to confirm syntax, you’re done. If you came here to get better at window functions, choose your next step.
LAG is part of a bigger window-function pattern. If you want the “why”, start here: Lead Lag
Reading docs is useful. Writing the query correctly under pressure is the skill.
For the authoritative spec, use the vendor docs. This page is the fast “sanity check”.
View Databricks SQL Documentation →Looking for more functions across all SQL dialects? Visit the full SQL Dialects & Window Functions Documentation.