What are contexts?

Uses

By default, a database update runs all undeployed changesets in the changelog. If you add a context to a changeset and use a context filter in the CLI, the update only runs changeset that match the context filter. If you don't specify a context filter in the CLI, every undeployed changeset in your changelog runs, even if they have context attached.

For example, if you want to tag your changesets per environment and run only these changesets, set a context on each changeset and specify a matching value for --context-filter in the CLI.

Context logic

You can specify a context using @ in addition to AND, OR, !, and parentheses in the changesets. This requires you to designate a context or label explicitly when Liquibase is run for the changeset to run. The attribute context: @test causes a changeset to not run if Liquibase runs without any contexts provided.

• context="@test"

Without parentheses, the order of operations is !, AND, and then OR. For example:

• context="!test"

• context="v1.0 or map"

• context="!qa and !main"

Using a "," to separate contexts works like an OR operation (a comma is an alias for OR):

• "test, qa " is the same as " test OR qa"

• "test, qa and main " is the same as "

  (test) OR (qa and main)"

Running contexts

You can only specify context filtering logic in a changeset definition, but you can still specify multiple contexts when running Liquibase in the CLI. However, you can only list out all the contexts that apply to the current Liquibase run.

liquibase update --context-filter="test" --changelog-file=example-changelog.xml

If your changelog includes several changesets with complex and simple context filters such as context="qa and main and !dev" for changeset 1 and context="test" for changeset 2, you need to pass the following on the command line to deploy them:

liquibase update --context-filter="test,qa,main" --changelog-file=example-changelog.xml

Example: If you run the liquibase update --context-filter=1.0 command, it will deploy all changesets with the context-filter set to 1.0 and all changesets without any context or with context="". If you run liquibase update, it will deploy all changesets, whether they have contexts or not.

Using contexts for test data

If you manage your test data with Liquibase, it is best practice to have this data in line with all your other changesets, but marked with a "test" context. When you want your test data inserted, run a database update and specify the "test" context in the CLI. When you need to migrate your production database, don't include the "test" context, and your test data will not be included.

Note: If you do not specify any contexts in the CLI at runtime, every changeset will be applied, including those marked with a "test" context.

If you have multiple test environments or test data sets, simply tag them with different contexts, like "min-test" and "integration-test".

Using contexts to control test data is better than having a separate changelogs tree because later Change Types and changes will be applied to existing test data the same as they are applied to production data. If you had a set of test data that was created and simply added after the database is set up, you would be constantly manually updating your test data scripts to keep them in line with the current database schema.

Multi-DBMS changelogs

If you need to use one changelog for multiple databases, and you only want specific changesets to run on each database, it is possible to use the context tag to filter them, and then run liquibase update --context-filter="<dbname>" in your command line. However, this is not recommended. For example:

<changeSet id="1-lawful-evil" author="adrian" context="postgres"> <createTable tableName="my_table"> <column name="id" type="int"/> </createTable> </changeSet>

Instead, it is a best practice to use the dbms tag to differentiate changesets by database type, and then run liquibase update in your command line. This is a clearer use of contexts and decreases the possibility of errors. You can use context and dbms on the same changeset, but only dbms should refer to your database type. For example:

<changeSet id="1-lawful-good" author="adrian" dbms="postgres"> <createTable tableName="my_postgres_table"> <column name="id" type="int"/> </createTable> </changeSet> <changeSet id="2-lawful-good" author="adrian" dbms="oracle"> <createTable tableName="my_oracle_table"> <column name="id" type="int"/> </createTable> </changeSet> <changeSet id="3-lawful-good" author="adrian" dbms="oracle" context="test"> <createTable tableName="my_oracle_test_table"> <column name="id" type="int"/> </createTable> </changeSet>

Default context

You can specify a context attribute in the root DATABASECHANGELOG node to assign that context to all changesets in the changelog by default.

The specified context will have AND with any context specified in changesets within the changelog file.

include and includeAll contexts

You can specify the context attribute in <include> or <includeAll> tags. If specified, the given context is added to all changesets in the included file(s).

Change history