Groovy Liquibase

A pluggable parser for Liquibase that allows the creation of changelogs in a Groovy DSL, rather than hurtful XML. If this DSL isn't reason enough to adopt Liquibase, then there is no hope for you. This project was started once upon a time by Tim Berglund, and is currently maintained by Steve Saliman.

News

July 14, 2018

We're proud to announce the release of version 2.0.0 of the Groovy DSL, which now fully supports Liquibase 3.6.1. This release will decouple the DSL from any particular version of Liquibase, making it easier for users to take advantage of new releases of Liquibase as soon as they are released. Note that the DSL itself is tested against a specific version of Liquibase to ensure complete compatibility with a specific version, so YMMV with regard to newer releases.

Release 2.0.0 has been tested against all the valid elements and attributes of Liquibase 3.6.1

This release has breaking changes, so make sure to read all the information in this section before upgrading.

There are several breaking changes with this version of the DSL:

1. The Groovy DSL no longer includes Liquibase itself as a dependency. Users must make sure the desired version of Liquibase is on the classpath.

2. There was a bug introduced in version 1.2.2 of the DSL regarding filenames and the includeAll change. Version 1.2.2 was incorrectly converting all changeset filenames to absolute paths, a bug that was fixed in version 2.0.0. If you are updating from version 1.2.1 or earlier, this change should not effect you, but if you've run changes with version 1.2.2, you will need to fix some or all of the paths in the DATABASECHANGELOG table before running the 2.0.0 parser. Failing to do this wil result in Liquibase trying to run the changes again.

3. Liquibase made a change to the checksum logic in version 3.6.0. According to the Liquibase documentation, Liquibase will just fix the checksums of each change when you run the first update command, but it won't detect changes to any changes that were marked with the runOnChange. If you have any changes that use runOnChange, you should run an update once with your old version, then run it again with the new version to fix the checksums.

4. Liquibase changed the resourceFilter attribute of the includeAll element to just filter. Since the 2.0.0 version of the Groovy DSL was built for Liquibase 3.6.x, it will throw an error if it finds the old resourceFilter attribute, so you will need to convert any effected change sets. Note that includeAll is one of the few things handled by the DSL itself, so filter will still work even if you're using an older version of Liquibase.

5. The alterSequence change used to have a willCycle attribute. That attribute is now called cycle

###February 23, 2017 Release 1.2.2 of the Groovy DSL is a minor release that resolves a few bugs. See the CHANGELOG for more details. Note that if you use this DSL via the Gradle Liquibase plugin, you will need version 1.2.3 of that plugin NOT 1.2.2, as that version has a broken dependency on the SNAPSHOT version of the DSL.

May 16, 2015

We are proud to announce that the Liquibase Groovy DSL is now a part of the Liquibase organization. I will continue maintain the code, but bringing this project into the Liquibase organization will help keep all things Liquibase together in one place. This will help promote Liquibase adoption by making it easier for more people to use, and it will help people stay up to date with the latest releases. As part of that move, the artifact name has changed from net.saliman:groovy-liquibase-dsl to org.liquibase:liquibase-groovy-dsl to be consistent with the rest of the Liquibase artifacts. A special thank you to Nathan Voxland for his help and support in bringing the Liquibase project and the Groovy DSL together into one home.

Usage

Simply include this project's jar file in your class path, along with a version of Liquibase, and Liquibase can parse elegant Groovy changelogs instead of ugly XML ones. The DSL syntax is intended to mirror the [Liquibase XML] (http://www.liquibase.org/documentation/databasechangelog.html) syntax directly, such that mapping elements and attributes from the Liquibase documentation to Groovy builder syntax will result in a valid changelog. Hence this DSL is not documented separately from the Liquibase XML format. We will, however let you know about the minor differences or enhancements to the XML format, and help out with a couple of the gaping holes in Liquibase's documentation of the XML.

Note that wile the Groovy DSL fully supports using absolute paths for changelogs, we strongly recommend using relative paths instead. When Liquibase sees an absolute path for a changelog, all changes included by that changelog will also have absolute path names, even if the include or includeAll element used the relativeToChangeLog attribute. This will cause problems in multi-developer environments because the difference in the users' directories will cause Liquibase to think that the changes are new, and it will try to run them again.

Deprecated and Unsupported Items

• Liquibase has a whereParam element for changes like the update change. It isn't documented in the Liquibase documentation, and I don't see any benefits of using it over the simpler where element, so it has been left out of the Groovy DSL.
• In the Liquibase XML, you can set a sql attribute in a sqlFile change, but that doesn't make a lot sense, so this has been disabled in the Groovy DSL.
• The documentation mentions a referencesUniqueColumn attribute of the addForeignKeyConstraint change, but what it doesn't tell you is that it is ignored. In the code, Liquibase has marked this item as being deprecated, so we've deprecated it as well, and we let you know about it.
• If you were using the DSL prior to version 1.0.0, a changeSet could have an alwaysRun property. This is inconsistent with Liquibase and has been replaced in 1.0.0 with runAlways
• Prior to 1.0.0, the DSL allowed a path attribute in an include. This is no longer allowed. includeAll should be used instead.
• Prior to 1.0.0, the DSL allowed createStoredProcedure changes. This has been replaced with createProcedure.
• Prior to 1.0.0, the DSL allowed a File object to be passed as an attribute to loadData and loadUpdateData changes. This is no longer supported, the path to the file should be used instead.
• Prior to 1.0.0, the DSL allowed constraint attributes to be set as methods in a constraint closure. This is inconsistent with the rest of the DSL and has been removed.
• Prior to 2.0.0, the DSL used the resourceFilter attribute of the includeAll element to filter the changelogs included in an directory. This has been changed to filter to remain consistent with Liquibase itself.

Additions to the XML format:

• In general, boolean attributes can be specified as either strings or booleans. For example, changeSet(runAlways: 'true') can also be written as changeSet(runAlways: true).
• The Groovy DSL supports a simplified means of passing arguments to the executeCommand change. Instead of:

execute {
  arg(value: 'somevalue')
}

You can use this the simpler form:

execute {
  arg 'somevalue'
}

• The sql change does not require a closure for the actual SQL. You can just pass the string like this: sql 'select some_stuff from some_table' If you want to use the comments element of a sql change, you need to use the closure form, and the comment must be in the closure BEFORE the SQL, like this:

sql {
  comment('we should not have added this...')
  'delete from my_table'
}

• The stop change can take a message as an argument as well as an attribute. In other words, stop 'message' works as well as the more XMLish stop(message: 'message')
• A customPrecondition can take parameters. the XMLish way to pass them is with param(name: 'myParam', value: 'myValue') statements in the customPrecondition's closure. In the Groovy DSL, you can also have myParam('myValue')
• The validChecksum element of a change set is not well documented. Basically you can use this when changeSet's current checksum will not match what is stored in the database. This might happen if you, for example want to reformat a changeSet to add white space. This doesn't change the functionality of the changeset, but it will cause Liquibase to generate new checksums for it. The validateChecksum element tells Liquibase to consider the checksums in the validChecksum element to be valid, even if it doesn't match what is in the database.
• The Liquibase documentation tells you how to set a property for a databaseChangeLog by using the property element. What it doesn't tell you is that you can also set properties by loading a property file. To do this, you can have property(file: 'my_file.properties') in the closure for the databaseChangeLog.
• Liquibase has an includeAll element in the databaseChangeLog that includes all the files in the given directory. The Groovy DSL implementation only includes groovy files, and it makes sure they are included in alphabetical order. This is really handy for keeping changes in a different file for each release. As long as the file names are named with the release numbers in mind, Liquibase will apply changes in the correct order.
• Remember, the Groovy DSL is basically just Groovy closures, so you can use groovy code to do things you could never do in XML, such as this:

sql { """
  insert into some_table(data_column, date_inserted)
  values('some_data', '${new Date().toString()}')
"""
}

Items that were left out of the XML documentation

• The createIndex and dropIndex changes have an undocumented associatedWith attribute. From an old Liquibase forum, it appears to be an attempt to solve the problem that occurs because some databases automatically create indexes on primary keys and foreign keys, and others don't. The idea is that you would have a change to create the primary key or foreign key, and another to create the index for it. The index change would use the associatedWith attribute to let Liquibase know that this index will already exist for some databases so that Liquibase can skip the change if we are in one of those databases. The Liquibase authors do say it is experimental, so use at your own risk...
• The executeCommand change has an undocumented os attribute. The os attribute is a string with a list of operating systems under which the command should execute. If present, the os.name system property will be checked against this list, and the command will only run if the operating system is in the list.
• The column element has some undocumented attributes that are pretty significant. They include:
  • valueSequenceNext, valueSequenceCurrent, and defaultValueSequenceNext, which appear to link values for a column to database sequences.
  • A column can be set auto-number if it the autoIncrement attribute is set to true, but did you know that you can also control the starting number and the increment interval with the startWith and incrementBy attributes?
  • Since Liquibase 3.6, you can specify a defaultValueConstraintName.
• The constraints element also has some hidden gems:
  • Some databases automatically create indexes for primary keys. The primaryKeyTablespace can be used to control the tablespace.
  • A foreign key can be made by using the references attribute like this: references: 'monkey(id)', It can also be done like this: referencedTableName: 'monkey', referencedColumnNames: 'id' for those who prefer to separate out the table from the column. Since Liquibase 3.5, this second form also has referencedTableCatalogName and referencedTableSchemaName attributes.
  • There is also a checkConstraint attribute, that appears to be useful for defining a check constraint, but I could not determine the proper syntax for it yet. For now, it may be best to stick to custom sql changes to define check constraints.
  • Since Liquibase 3.6, you can specify a name for a Not Null constraint with the notNullConstraintName attribute.
• The createSequence change has n cacheSize attribute that sets how many numbers of the sequence will be fetched into memory for each query that accesses the sequence.
• The documentation for version 3.1.1 of Liquibase mentions the new beforeColumn, afterColumn, and position attributes that you can put on a column statement to control where a new column is placed in an existing table. What the documentation leaves out is that these attributes don't work :-)
• Version 3.4.0 of Liquibase introduced two new attributes to the includeAll element of a databaseChangeLog, both of which are undocumented. The first one is the errorIfMissingOrEmpty attribute. It defaults to true, but if it is set to false, Liquibase will ignore errors caused by invalid or empty directories and move on. The second one is the resourceFilter attribute. A resourceFilter is the name of a class that implements liquibase.changelog.IncludeAllFilter interface, which allows developers to implement sophisticated logic to decide what files from a directory should be included (in addition to the .groovy extension filter that the Groovy DSL imposes).
• Liquibase 3.5.0 renamed the resourceFilter of includeAll to just filter. It also added resourceComparator, which lets you specify the name of a class that implements Comparator that will be used to determine how to sort files. The default is to just sort them by name.
• Liquibase 3.4.0 added the undocumented forIndexCatalogName, forIndexSchemaName, and forIndexName attributes to the addPrimaryKey and addUniqueConstraint changes. These attributes allow you to specify the index that will be used to implement the primary key and unique constraint, respectively.
• Liquibase 3.4.0 added the undocumented cacheSize and willCycle attributes to the alterSequence change. cacheSize sets how many numbers of the sequence will be fetched into memory for each query that accesses the sequence. willCycle determines if the sequence should start over when it reaches its maximum value.
• Liquibase 3.5.0 changed the willCycle attribute of alterSequence to be cycle. It does the same thing, but it remains undocumented.
• Liquibase added the context attribute to the include, includeAll, and changeLog elements. They work the same as the context attribute of a change set.
• Liquibase 3.5 added runOrder and created attributes to the changeSet element. runOrder lets you specify that a change set should always be first or last. I have no idea what created does.
• Liquibase 3.6 added the ignore attribute to a change set, which seems to be a way to ignore a change set.
• Liquibase 3.6 added validate and clustered to the addUniqueConstraint change. validate tells the database to validate the constraint when it is first created, and clustered tells the database to use a clustered index.
• Liquibase 3.6 added defaultValueConstraintName to the addDefaultValue change to give the constraint being created a name.
• Liquibase 3.5 added commentLineStartsWith to the loadData change. By default, Liquibase treats lines in the loaded file that start with a # to be comments. commentLineStartsWith lets you change that.
• Liquibase 3.6 added usePreparedStatements to the loadData change. When true, it tells Liquibase to use prepared statements in the inserts it generates.
• Liquibase 3.6 added validate to the addForeignKeyConstraint change to tell the database whether or not to validate a new constraint when it is made.
• Liquibase 3.6 added another way to create a view with the createView change. Previously, the SQL to create the view was in the closure of the createView element. Now you can specify a file with the SQL using the new path, encoding, and relativeToChangelogFile attributes. I have no idea what happens if you specify both a path to a file and a closure with SQL.
• Liquibase 3.6 added a timeout to the executeCommand change. It lets you specify a command timeout in seconds, minutes, or hours. For example, to set a 2 minute timeout, you'd use 2m as the value of the timeout attribute.

License

This code is released under the Apache Public License 2.0, just like Liquibase 2.0.

TODOs

• Support for the customChange. Using groovy code, liquibase changes and database SQL in a changeSet.
• Support for extensions. modifyColumn is probably a good place to start.