Combining multiple external datasets with cucumber/selenium test environments - selenium-webdriver

Question

When developing automated tests using cucumber and selenium webdriver in java, I use excel spreadsheets as datasets for the gherkin scenarios instead of the traditional Examples tables, using a simple table with only the row numbers in my feature files. This works very well when doing tests that only make use of data from one spreadsheet at a time, but when implementing tests that make use of multiple spreadsheets, how does one ensure the it iterates over every combination.
For example, when testing multiple configurations and their impact on the main interface, I provide the configuration data, let's say 3 combinations of different configurations, using the first spreadsheet, and in my gherkin feature I only enter the row numbers and use the code to handle the actual reading of data.
When the user uses configuration from row <ExcelRow>
...
Examples:
| ExcelRow |
| 1 |
| 2 |
| 3 |
The problem arises when I want to test such configurations with different combinations of inputs in the main interface, also provided via a separate excel spreadsheet. I want configuration from row 1 to be run with all rows from the second spreadsheet, before moving on to row 2's configuration and so on.
Manually using the examples table to do the combinations does the job when working with smaller data sets
Examples:
| ConfigRow | InputRow |
| 1 | 1 |
| 1 | 2 |
| 2 | 1 |
| 2 | 2 |
| 3 | 1 |
| 3 | 2 |
The problem arises when using very large datasets, where the examples table starts to clutter the feature file despite only containing the row numbers.
I tried implementing the actual input testing as a single step that loops over the entire excel spreadsheet for each configuration, but that forced me to do my assertions in the same loop and not in the Then step.

Answer 1

If you want to mention only config row in feature file and you want that some other rows to be executed for each config row then you may want to utilize cucumber-guice and make it #ScenarioScoped . Cucumber-guice can initialize same classes for each scenario independently. You would need these dependencies in your pom
<dependency>
<groupId>io.cucumber</groupId>
<artifactId>cucumber-guice</artifactId>
<version>${cucumber.version}</version>
<scope>test</scope>
</dependency>
<dependency>
<groupId>com.google.inject</groupId>
<artifactId>guice</artifactId>
<version>4.2.3</version>
<scope>test</scope>
</dependency>
Then , in a global class you can do
import io.cucumber.guice.ScenarioScoped;
#ScenarioScoped
public class Global {
public Helpers help;
//constructor
public Global() {
//helper class would contain code that does all the second excel sheet work
help = new Helpers();
}
}
In step def you can do
//import Global and guice dependencies
import yourPackage.Global;
import com.google.inject.Inject;
...
...
public class stepDef {
#Inject
Global global;
#When ("the user uses configuration from row {int}")
public void useConfigs(){
global.help.doSomeExcelWork();
}
#Then ("I assert from excel sheet")
public void doAssertions(){
//do assertions here. These
global.help.doAssertion();
}
}
Your helper class could be something like this
public class Helper {
public void doSomeExcelWork(){
//do excel work
}
public void doAssertion(){
//return values for your assertions
}
}
Your feature file would look like
When the user uses configuration from row <ExcelRow>
Examples:
| ExcelRow |
| 1 |
| 2 |
Then I assert from excel sheet
Now , for all your examples (scenarios) global would be injected independently and the then statement also would be called for each example row

Answer 2

I am not sure if that is possible to do via Cucumber. You may try searching for Cucumber Java dynamic examples generation like here.
I just would like to question if Cucumber/Gherkin are the right tools for what you wanted to achieve. The primary goal of Gherkin / Cucumber / Specflow scenarios is demonstrate the behavior of the system to anyone reading the feature file. So hiding the data in "linked" files can be accepted if they hold a complex piece of data which is as a single unit, provided the file name demonstrates what is special about the data inside.
What you might be looking for are Parameterized Tests and Data Providers that are available in JUnit 5 and TestNG 2, 3. If you write automation framework in the way that Cucumber, or other test framework becomes only a thin wrapper around it which "assembles" the test, you can generate tests on the fly.
For example your step: "When the user uses configuration from row", becomes
public void whenUserUsesConfiguration(SutConfiguration configuration) {
// your configuration setup goes here
// but you do not read configuration from file in this method
}
Method above can be used in both Cucumber steps and JUnit/TestNG test without loosing any readability, or understandability.
By splitting your tests into two parts, one for understanding general system behavior and accessible to all stakeholders, and the one that check lots of small nuances, but both using the same framework will allow you to have greater flexibility, and more pleasant development experience.