lakemigrate

Ordered SQL migrations with a version table for Delta lakehouse storage.

lakemigrate discovers SQL files in a directory, applies any that have not yet run — in version order — and records each in a Delta history table. Checksums guard against silent edits after a migration has been applied.

Installation

Managed Spark environment (Databricks, EMR, GCP Dataproc — Spark is already present):

pip install lakemigrate

Local development (installs PySpark and delta-spark):

pip install "lakemigrate[spark]"

Migration files

Name each file NNN__description.sql where NNN is a zero-padded three-digit version number:

migrations/
  001__create_events_table.sql
  002__add_user_id_column.sql
  003__backfill_legacy_data.sql

Files are applied in version order. Once applied, the file content is checksummed — modifying a file after it runs raises ChecksumMismatchError.

CLI

# apply all pending migrations
lakemigrate migrate --migrations-dir migrations/

# custom history table
lakemigrate migrate --migrations-dir migrations/ --history-table mydb.schema_history

Python API

from lakemigrate import migrate

# uses default history table: default.lakemigrate_history
migrate("migrations/")

# custom history table
migrate("migrations/", history_table="mydb.schema_history")

A SparkSession must be active before calling migrate. The history table is created automatically on first run.

Error handling

All exceptions inherit from LakeMigrateError.

Links

• PyPI
• Source
• Issues