In our previous blog post, we discussed how we can apply different changelogs to different database environments. It is more than often, that when applying a changelog, changeset writer assumes database in a certain state. Like when you are adding a column to the database, you would assume that corresponding table is present. Or when you are dropping a table, it has no data in it. Or we assume that underlying database connection is of a particular nature. We can check for and decide what to do by using the concept of Preconditions in the Liquibase. Using preconditions allows to validate underlying assumption and decide the course of action. Preconditions can be attached to changelogs or changesets to control the execution of an update based on the state of the database.

Types of Preconditions and their meaning

Each of these preconditions can again have one or more attributes which are necessary to define the precondition.

Handling Failures and Errors

Naturally, you would want to take some actions when a precondition fails. There can be two scenarios:

Failure: When the precondition check fails
Error: When the check itself cannot be performed and an exception is thrown

Both of above situations can be handled by a number of attributes. Some of these include:

Again, now that the underlying Precondition failed or the check for precondition itself failed, you need to decide the course of action. It can be set using below attributes:

The possible actions for onUpdateSQL are a little different.

Fitting it all together

In our database, we have a table [dbo].[CustomerDetails] which has only two columns: CustomerTypeID and CustomerDesc. We want to add a new column to this table, say CustomerAddress but we want to make sure that the said table exists before we. For this, we can use below changelog:

Let’s go ahead and apply our change using liquibase update. Since there are no preconditions fails, we’ll not see any indication of same in the logs:

Now, let’s try to apply below changelog where we are trying to drop a table:

Here, in the precondition we have specified that table contains no records. However, our table contains 2 records. Let’s run liquibase and see what happens:

We can see that liquibase command has failed to proceed as instructed. At this point, if we check records in the DATABASECHANGELOG table, we would see no new entries as no change has been made.

Conditional logic in Preconditions

We can use as many preconditions as we need to make sure underlying assumptions are valid. By default all preconditions are associated by AND. The possible association can be defined using AND, OR and NOT.