Database | Versions Tested |
---|---|
Postgres | 9, 9.5, 10, 11, 12, 13, 14 |
Postgres RDS | 9.6, 10, 11, 12, 13 |
MySQL | 5.6, 5.7, 8 |
MariaDB | 10.2, 10.3 , 10.4, 10.5, 10.6, 10.7 |
SQL Server | 2017 , 2019 |
Percona XtraDB | 5.7 , 8.0 |
Oracle | 18.3.0, 18.4.0, 19.9.0, 21.3.0 |
Oracle RDS | 12.1.0, 19.0 |
CockroachDB | 20.1, 20.2, 21.1 |
EDB | 9.5, 9.6, 10, 11, 12, 13, 14 |
H2 | 2.1.210 |
SQLite | 3.34.0 |
Apache Derby | 10.14.2.0 |
Firebird | 3.0, 4.0 |
HSQLDB | 2.3.4 , 2.5 |
The test harness consists of a variety of standard tests to ensure the database-specific interactions within Liquibase work against specific versions and configurations
The test harness logically consists of three parts:
- Test logic
- Configuration files containing inputs for the test logic
- Configuration files containing outputs/expectations for the test logic.
The built-in tests are designed to test an overall functional flow, iterating over all configured connections. For each connection, it will run each applicable input configuration and compare it to the expected output/expectations.
Both the input and output configuration files can be defined in a way that makes them apply to all databases or to specific types and/or specific versions.
The general pattern is that for each directory containing configuration files:
- Files directly in that root apply to all databases. Example:
liquibase/harness/change/changelogs
- Files in a subdirectory named for the database type apply to only that type of database. Example:
liquibase/harness/change/changelogs/mysql
- Files in a subdirectory with a version apply only to this version of the database. Example:
liquibase/harness/change/changelogs/mysql/8
Note: The version folder name should match exactly with the DB version provided in harness-config.yml
file. We do not split this to major/minor/patch subversion folders currently.
At each level in that hierarchy, new configurations can be added and/or can override configurations from a lower level.
Currently, there are four test types defined in the test harness:
- Change Object Tests
- Change Data Tests
- Diff Command Test
- Base level test
This repository is configured to run against databases supported by Liquibase Core.
Extensions that add support for additional databases and/or define additional functionality can add this framework as a dependency and use the existing tests to:
- More easily verify their new functionality works
- And that it also doesn't break existing logic
The test harness will look for a file called harness-config.yml
in the root of your classpath.
That file contains a list of the database connections to test against, as well as an ability to control which subsets of tests to run.
See src/test/resources/harness-config.yml
to see what this repository is configured to use.
For more information on using the test harness in your extension, see [README.extensions.md]
The test-harness validates most of the Change Types as listed on Home Page. The primary focus is on add, create, drop & rename database objects.
The groovy/liquibase/harness/ChangeObjectsTests.groovy
test executes changelogs against the database and validates the SQL generated by them as well as
whether they make the expected changes.
- The test behavior is as follows:
- It reads the changesets from the changelogs provided in
src/main/resources/liquibase/harness/change/changelogs
folders (recursively) - Runs the changeset through the SqlGeneratorFactory to generate SQL
- Compares the generated SQL with the expected SQL (provided in
src/main/resources/liquibase/harness/change/expectedSql
) - If the SQL generation is correct, the test then runs
liquibase update
to deploy the changeset to the DB - The test takes a snapshot of the database after deployment
- The deployed changes are then rolled back by either using
rollbackToDate
(default) orrollback
by tag (test-harness-tag). See-DrollbackStrategy
option below for more information. - Finally, the actual DB snapshot is compared to the expected DB snapshot (provided in
src/main/resources/liquibase/harness/change/expectedSnapshot
)
- It reads the changesets from the changelogs provided in
- The tests work with the 4 types of input files that are supported by liquibase itself - xml, yaml, json, sql. Thus files with extensions 'xml', 'sql', 'json', 'yml', 'yaml' are taken into account, but not all formats together in the same run.
- The default format is xml, so by default only changelogs with xml file extension are executed.
To change it to another format, like 'sql' for instance, specify
-DinputFormat=sql
as the command line argument for Maven or as VM option to your JUnit test run config.
- Go to
src/main/resources/liquibase/harness/change/changelogs
and add the xml (or other) changeset for the change type you want to test.
- The framework tries to rollback changes after deploying them to DB. If liquibase knows how to do a rollback for that particular changeset, it will automatically do that. If not, you will need to provide the rollback by yourself. To learn more about rollbacks read Rolling back changesets article.
- Go to
src/main/resources/liquibase/harness/change/expectedSql
and add the expected generated SQL.
- You will need to add this under the database specific folder.
- NOTE: If your changeSet will generate multiple SQL statements, you should add each SQL statement as a separate line. (See
renameTable.sql
in the postgres folder for an example.) - If you would like to test another DB type, please add the requisite folder.
- Go to
src/main/resources/liquibase/harness/change/expectedSnapshot
and add the expected DB Snapshot results.
- To verify the absence of an object in a snapshot (such as with drop* commands) add
"_noMatch": true,
to that tree level where the missing object should be verified. See dropSequence.json as an example. Additionally the_noMatchField
parameter can be used to define the exact property which should be absent or different for that particular database object (for example Column, Table etc.) see createTableWithNumericColumn.json - You will need to add this under the database specific folder.
- If you would like to test another DB type, please add the requisite folder.
- Go to your IDE and run the test class
ChangeObjectTests.groovy
(You can also choose to runBaseTestHarnessSuite
, orLiquibaseHarnessSuiteTest
-- at present they all work the same).
The primary goal of these tests is to validate change types related to DML (Data Manipulation Language) aspect. Generally it is similar to ChangeObjectTests except it doesn't use Liquibase snapshot to verify data but obtains result set via JDBC.
src/main/resources/liquibase/harness/data/changelogs
- add DML related changelogs here;src/main/resources/liquibase/harness/data/checkingSql
- add select query which will obtain a result set from required table;src/main/resources/liquibase/harness/data/expectedResultSet
- add JSON formatted expected result set from required table where left part of a JSON node is the name of a change type and right part is JSON Array with a result set;src/main/resources/liquibase/harness/data/expectedSql
- add query which is expected to be generated by Liquibase; To run the test go to your IDE and run the test classChangeDataTests.groovy
.
This test executes the following steps:
- Reads
src/test/resources/harness-config.yml
andsrc/main/resources/liquibase/harness/diff/diffDatabases.yml
to locate the databases that need to be compared - Creates a diff based on 2 databases (targetDatabase and referenceDatabase) from
diffDatabases.yml
- Generates the changelog based on diff
- Applies the generated changelog to the targetDatabase
- Checks the diff between the target and reference databases again
- If some diffs still exist, then they are matched with the expected diff from
liquibase/harness/diff/expectedDiff
folder
Warning: This is a destructive test -- it will alter the state of targetDatabase to match the referenceDatabase.
Java 1.8
- Make sure you have a docker container up and running first
- Go to
src/test/resources/docker
and rundocker-compose up -d
. Wait until the databases start up. - Open
src/test/groovy/liquibase/harness/LiquibaseHarnessSuiteTest.groovy
in your IDE of choice and run it
Build the project first by running mvn clean install -DskipTests
Execute mvn test
with the (optional) flags outlined below:
-DinputFormat=xml
or select from the other inputFormats listed in Types of input files-DchangeObjects=createTable,dropTable
flag allows you to run specific changeObjects rather than all. Use comma separated lists.-DchangeData=insert,delete
flag that allows to run specific changeData through ChangeDataTests. Use comma separated list-DconfigFile=customConfigFile.yml
enables to override default config file which is(src/test/resources/harness-config.yml
)-Dprefix=docker
filters database from config file by some common platform identifier. E.g. all AWS based platforms, all Titan managed platforms, all from default docker file.-DdbName=mysql
overrides the database type. This is only a single value property for now.-DdbVersion
overrides the database version. Works in conjunction with-DdbName
flag.-DrollbackStrategy
overrides the default rollback strategy ofrollbackToDate
where we create a timestamp in UTC timezone and then try to rollback to that point in time. But this rollback strategy might not work well in some cases like cloud databases for instance -- cloud databases are often in different timezones than the test-harness runners, so therollback
command can be used instead in conjunction with thetest-harness-tag
tag. To do so, use-DrollbackStrategy=rollbackByTag
.
To run the test suite itself, you can execute mvn -Dtest=LiquibaseHarnessSuiteTest test
When you are done with test execution, run docker-compose down --volumes
to stop the docker containers
gracefully and to allow the tests to start from a clean slate on the next run.