SQLTest

SQLTest is a test tool that makes it easy to write tests for Prana.

You specify your test a script to be run, together with a text file containing expected output, and another text file containing test data

A test with name <test_name> is made up of three files:

The script file

This is named <test_name>_test_script.txt

Important note! Every statement or comment in a script file MUST end in a semicolon, or it will not be parsed correctly

It contains a sequence of SQL statements, just like the ones you would type into a Prana CLI.

Statements can be DDL statements such as create source..., create materialized view..., drop source..., drop materialized view..., or queries such as select * from some_table.

SQLTest runs your script, collects the output and compares it to the expected output

Test scripts can also contain SQL comments which begin with --.

Special directives

We define some special comments which do special things when they are encountered. This allows us to do lots of powerful things in tests. Here are the currently supported special comment directives:

• --load_data dataset_name [no wait]; This loads the dataset with the specified name. If no wait is specified then it does not wait for the dataset to be processed before continuing. If not specified then it does wait for processing to complete.
• --repeat n; This repeats the test n times in a loop. Very useful for reproducing non-deterministic test failures. If there's a repeat directive, it must be the first line in a script.
• --close session; This closes the current SQL session at that point. A new session will then be created for the next statement. Used for testing scope of prepared statements.
• --create topic topic_name; Creates a topic on the Fake Kafka used by SQLTest with the specified name
• --delete topic topic_name; Deletes a topic on the Fake Kafka used by SQLTest with the specified name
• --reset offsets topic_name; Resets the committed offsets for the specified topic
• --restart cluster; Restarts the Prana cluster
• --kafka fail fail_time; Tells the Fake Kafka to return errors when attempts to get messages or commit offsets are made. Errors are returned for fail_time milliseconds, then normal operation is resumed.
• --wait for rows table_name num_rows; Waits for num_rows rows to be present in the specified table before proceeding. table_name is the name of the source or materialized view.
• wait for committed source_name num_messages; Waits for the source to commit num_messages messages from Kafka. Does not include duplicates.
• wait for duplicates source_name num_duplicates; Waits for the source to receive num_duplicates duplicate messages from Kafka.
• enable commit offsets source_name; Enables committing of offsets for the specified source. Default is enabled.
• disable commit offsets source_name; Disables committing of offsets for the specified source.

Note on prepared statements

We also support some special statements that can be used for testing prepared statements.

To create a prepared statement for a query:

prepare select * from foo order by bar;

This will return the prepared statement id.

Then to execute a prepared statement:

execute 3 hello 23.2

The first argument is the id of the prepared statement. The remaining arguments are the arguments for the prepared statement. They are space delimited.

The expected output file

This is named <test_name>_test_out.txt.

It contains the expected output of the test. SQLTest compares the actual output of running the script with the expected output and if they are the same, the test passes, otherwise it fails.

If the test fails, the actual output will be written to stdout, so you can see what happened.

The test data file

This is named <test_name>_test_data.txt.

A test data file can contain multiple data-sets. Each dataset is a set of rows, in CSV format, with one row per line. Each data-set starts with a descriptor line, like the following:

dataset:my_data_set1 my_source_1

This says, here is a dataset called my_data_set1 and it gets inserted into a source with name my_source_1

Datasets are inserted into sources at points in the script where a special load-data directive in a SQL comment is encountered, e.g:

--load data dataset_1;

When the script encounters this line, it loads the data from the dataset with that name into the source described in the dataset descriptor. It then waits for the data to be fully propoagated and any cascading MVs across the cluster to be updated before proceeding.

Running tests

By default tests run both against a fake single node cluster and a real three node cluster.

You can run these individually by running the test cases TestSqlFakeCluster and TestSqlFakeCluster in SQLTest.

Tests that clean-up

SQLTest will check that the cluster is in a clean state at the end of every test run - that means there are no sources or materialized views, no data in user tables in storage, and no sessions in memory.

That means you must drop all sources and materialized views that you create at the end of each test script.

Running an individual test

You can run an individual test or subset of tests which start with the same prefix by setting the variable TestPrefix in SQLTest