Skip to content

Like DB Unit, but designed a lot better, easier to use, faster and its for Scala!

License

Notifications You must be signed in to change notification settings

barorion/scaladbtest

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

55 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

scaladbtest

A new, light-weight way to load test data into your database before individual tests run. It is intended to replace DBUnit. It is written in Scala, but you can use it in Java/Junit4 as well.

Getting Started is Simple

Scaladbtest uses a simple file format called DBT that is used to write and load test data. Unlike DBUnit, it does not use XML, depend on Schemas or other bloated text-based formats. I believe that XML schemas are maintenance heavy and violate the DRY principle. There must be a better way, and there is!

Introducing DBT

DBT is scaladbtest's way of specifying test data. When comparing DBT to an XML file that consisted of 500 lines and was used by DBUnit, there was nearly a 30% Reduction in the number of characters used to express the same test data!

There is only one way to load data in ScalaDBTest, and I did my best to make it a good one! ;) There's suddenly no confusion between which format to use among dozens... or putting up with the hassle of trying to figure out how to configure them! You also don't have to download a tool to generate a schema just to make the XML format usable.

DBT Usage

Make a file in your tests' resources directory. Usually it's located in "src/test/resources" as part of your project structure if you're using SBT or Maven.

You can name the DBT file whatever you want, but "data.dbt" is a pretty good name to get started until you start splitting up your files.

Inserting Records

Here's a simple example that inserts a single record for the table country:

country:
- country_id: 1, name: "Canada"

Table declarations start with an identifier like country, and are then followed by an immediate : character.

From there, you can list a record for this table with -, followed by a comma separated list of name/value pairs for the record's data as shown above.

Whitespace is not important, so feel free to shape the syntax however you like.

Inserting Multiple Records for the same Table

You can specify multiple records that belong to the same table by using - characters, like this:

province:
- province_id: 1, name: "Alberta", country_id: 1
- province_id: 2, name: "British Columbia", country_id: 1
- province_id: 3, name: "Manitoba", country_id: 1

See, you're already saving many characters compared to XML!

Numbers

Numbers can be represented with or without quotes. Most databases have no issues accepting the string '1' and converting it to the number 1, so you are free to choose whichever format pleases your eye.

For example, you could place the number 1 in quotes, like this:

country:
- country_id: "1", name: "Canada"

Strings

Strings in DBT must be surrounded in quotes. This is to allow for spaces in the actual text, such as the string "United States":

country:
- country_id: 2, name: "United States"

Booleans

Booleans are represented by the literals true and false, like this:

user_account:
- user_account_id: 1, name: "Ken Egervari", is_enabled: true
- user_account_id: 2, name: "George W. Bush", is_enabled: false

If you specify the value "true" instead of true, scaladbtest will think you mean the string "true" instead of the boolean value, which is probably not what you want.

Dates

Dates are represented just like strings, so there's no special mechanics for you to remember with DBT. However, scaladbtest goes one step further by providing you with an easy way to acess today's date and time with the expression $now.

So instead of specifying a lengthy record declaration like the following...

task:
- task_id: 1, name: "My Task", creation_date: "2010-05-15 01:00:00.00"

... Scaladbtest will let you specify it like this instead:

task:
- task_id: 1, name: "My Task", creation_date: $now

Null Values

One of the annoying things with DBUnit is that you had to instruct the framework to transform a special string, such as "[NULL]", to an actual null value in Java. This trick required a few lines of setup code... and to the uninitiated, you'd probably be scratching your head for a bit until you figured it out.

With DBT, you can plant null literals anywhere you like. Here's an example:

education_facility:
- id: 1, name: "Star Trek University", province_id: 5, director_id: null

Scaladbtest also doesn't force you to specify ALL the columns in the first record definition like DBUnit does... so if you don't specify a value - it is null (unlike DBUnit, which just pretends you didn't specify the column at all.)

education_facility:
- id: 1, name: "Scala University", province_id: 2
- id: 1, name: "Star Trek University", province_id: 5, director_id: 1

In this case, Scaladbtest will correctly assign null to director_id for the 1st record, and will assign the value 1 to the second record. Finally, we have a tool that does the right thing with no complex formats or extra work!

Optional Labels

You can use a new concept called a label, which can populate columns with data as well as give yourself an alternate way to search for the data without remembering the primary key in your java/scala code. You can define a label like this:

country:
-[Canada] country_id: 1, name: $label

$label is an expression value. When it comes time to put this value in the database, scaladbtest will replace it with the exact string you provided as a label.

In the future, they'll be more functionality to manipulate labels to make maintaining and creating test data easier.

Labels Generating Primary Keys (Not Yet Implemented)

All the records we've been defining so far have been anonymous records. We've been defining the individual ids on each record.

Imagine, however, that you've been away from the project for 3 months. You know that there's a country called "Canada" in the database, but you can't quite remember the ID. In fact, you can't remember what any of the ids relate to anymore. Trust me, I've been there!

Labels offer a more intuitive way to pull out records from test data:

country: [pk: country_id]
-[Canada] name: $label
-[United States] name: $label
-[Mexico] name: $label

By using the optional table parameter syntax, we can specify which column is supposed to be the primary key, and scaladbtest will generate a key using the hash code of the label for us!

Once the test data is loaded, whenever you want to lookup the record, you can simply refer to "Canada".hashCode in your code.

BENEFIT: This instantly improves the readability of your tests, and helps people get back in the groove if they've been out of a project for far too long.

Default Columns

Of course, specifying column values that are the same for all of the records is redundant, and it increases the cognitive load. A better way is to tell scaladbtest to use default values for all the values that are the same.

In this next example, Scaladbtest will make sure that all 3 of these records will have the value 1 for the column country_id:

province:
? country_id: 1
- province_id: 1, name: "Alberta"
- province_id: 2, name: "British Columbia"
- province_id: 3, name: "Manitoba"

To specify defaults, you begin the line with a ? instead of the regular -. There can only be one ? definition per table declaration (although you can have multiple declarations with the same name - more on this later). It must also be the first declaration under the table name before any records are defined.

Overriding Defaults

You can also override the default values if you wish, and scaladbtest will do the right thing that you would expect:

province:
? country_id: 1, nice_weather: true
- province_id: 1, name: "British Columbia"
- province_id: 2, name: "Manitoba", nice_weather: false
- province_id: 3, name: "New York", country_id: 2

All three of the province records will contain 4 columns. Manitoba will still have bad weather while New York will belong to the United States (2) instead of Canada (1).

This kind of expressive power can really save typing and space when your actual tests only care about a small subset of columns on a given table.

Record Insertion Order

DBUnit had a strange limitation where it inserted all the records grouped by the table they belonged to, regardless of the order you wrote them in. This caused a lot of headaches because foreign key constraints would get violated very easily, forcing you to turn them off altogether.

Scaladbtest does not have this problem - it remembers the exact order you wrote your records in and processes them exactly in that same order! Intuitive, eh? (Yes, I'm from Canada).

By combining default values and the ability to specify column definitions multiple times, it is very easy to maintain your test data:

category:
- category_id: 1, name: "Star Trek Characters"

category_answer:
? category_id: 1
- category_answer_id: 1, text: "Sisko"
- category_answer_id: 2, text: "Quark"
- category_answer_id: 3, text: "Odo"

category:
- category_id: 2, name: "Races"

category_answer:
? category_id: 2
- category_answer_id: 4, text: "Romulan",
- category_answer_id: 5, text: "Federation"
- category_answer_id: 6, text: "Klingon"

The data actually looks readable and you can infer the structure of it quite easily. This is a natural synergy that occured when you group default values with your own table orderings.

Comments

Naturally, you might find it useful to insert comments into your file. You can do so like this:

# countries
country:
# Why did I have to be born in a socialist country for? :(
- id: 1, name: "Canada"

Executing scaladbtest from Unit Tests

Setting up scaladbtest is easy - even easier than in DBUnit.

Here's a small snippet of what you need to do to get scaladbunit to work with Java, JUnit4 and the Spring Framework (taken from a real-world project). The steps would be similar for other testing frameworks.

@Autowired
protected DataSource dataSource;

protected ScalaDbTester scalaDbTester;

@PostConstruct
public void initialize() {
	scalaDbTester = new ScalaDbTester(dataSource);
}

@Before
public void loadTestData() {
	scalaDbTester.onBefore("src/test/resources/data.dbt");
}

@After
public void cleanTestData() {
	scalaDbTester.onAfter();
}

As long as you've specified your DBT file in the correct location, everything should just work out of the box. Before your tests are run, scaladbtest will parse your DBT files and load everything into the database. You can even specify multiple DBT files to scalaDbTester.onBefore() if you wish.

After your tests are run, scaladbtest will naturally clean out all the test data. It's really that simple! DBUnit requires twice as much code to setup, requires declaration of exceptions and has lots of other mandatory configuration doo-dads that should have been optional and set as defaults.

Scaladbtest is truly a better way to load test data!

Requirements

scaladbunit requires Scala 2.8.0, Spring Framework's 3.0's JDBC package and Scalaj-collection library if you intended to use it in Java.

Installation

From source

Clone the repository from Github:

git clone git://github.com/egervari/scaladbtest.git

Build the project and create the JAR (requires sbt version 0.7.4 or greater):

cd scaladbtest
sbt package

sbt (Not Yet In Central Repository)

If you're using simple-build-tool, simply add the following line to your project file:

val scaladbtest = "scaladbtest" % "scaladbtest_2.8.0" % "0.1"

Maven (Not Yet In Central Repository)

If you're using Maven, add the following to your pom.xml:

<dependency>
  <groupId>scaladbtest</groupId>
  <artifactId>scaladbtest_${scala.version}</artifactId>
  <version>0.1</version>
</dependency>

In Closing

I hope scaladbtest gives you enough reasons to switch from dbunit and provides you with the best database testing experience on the JVM!

About

Like DB Unit, but designed a lot better, easier to use, faster and its for Scala!

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published