How to manage schema migrations in Google BigQuery

Question

How to manage schema migrations for Google BigQuery, we have used Liquibase and Flyway in the past. What kind of tools can we use to manage schema modifications and the like (e.g. adding a new column) across dev/staging environments.

Answer 1

Found open source framework for BigQuery schema migration

https://github.com/medjed/bigquery_migration

One more solution

https://robertsahlin.com/automatic-builds-and-version-control-of-your-bigquery-views/

PS

In flyway someone opened the ticket to support BigQuery.

Answer 2

Flyway, a very popular database migration tool, now offers support for BigQuery as a beta, while pending certification.

You can get access to the beta version here: https://flywaydb.org/documentation/database/big-query after answering a short survey.

I've tested it from the command line and it works great! Took me about an hour to get familiar with Flyway's configuration, and now calling it with a yarn command.

Here's an example for a NodeJS project with the following files structure:

package.json
fireway/
    <SERVICE_ACCOUNT_JSON_FILE>
    flyway.conf
    migrations/
        V1_<YOUR_MIGRATION>.sql

package.json

{
  ...
  "scripts": {
    ...
    "migrate": "flyway -configFiles=flyway/flyway.conf migrate"
  },
  ...
}

and flyway.conf:

flyway.url=jdbc:bigquery://https://www.googleapis.com/bigquery/v2:443;ProjectId=<YOUR_PROJECT_ID>;OAuthType=0;OAuthServiceAcctEmail=<SERVICE_ACCOUNT_NAME>;OAuthPvtKeyPath=flyway/<SERVICE_ACCOUNT_JSON_FILE>;

flyway.schemas=<YOUR_DATASET_NAME>
flyway.user=
flyway.password=

flyway.locations=filesystem:./flyway/migrations
flyway.baselineOnMigrate=true

Then you can just call yarn migrate any time you have new migrations to apply.

Answer 3

According to the BQ docs, you can add a row to the schema without any additional process.

For more complex transformations, if it can be resolved in a SQL query, you can just run that query setting the destination table as the source table (although I would suggest creating a backup of the table in case something goes wrong).

Example

Let's say I have a table with a column that is a integer (column d), but at the insertion time it was written as a string. I can modify the table by setting itself as a destination table and running a query like:

SELECT
  a,
  b,
  c,
  CAST(d AS INT64) AS d,
  e,
  f
FROM
  `example.dataset.table`

This is an example for changing the schema, but this can be applied as long as you can get the result with a BQ query.