I ran bal test and got one passing test. Then I changed the test, ran it again, and got exactly the same result. If both versions passed, what had I improved?
I found the answer by changing one line in the function and watching what happened.
What the first test checked
I was working through Ballerina’s HTTP client mocking example. It includes a function called getRandomJoke(name) that fetches a joke and replaces “Chuck Norris” with the name you give it. If I call getRandomJoke("Alex"), the joke I get back should be about Alex.
The example uses a mock HTTP client. That lets the test supply the same joke every time without calling a live API. The original test called the client directly:
http:Response result = check clientEndpoint->/random;
It then checked that the response contained the joke configured in the mock. That explained the first pass. It also showed me the gap: getRandomJoke never ran. Whatever happened inside that function, the test would keep asking the mock for the same joke and getting the same answer.
What I changed
I kept “Chuck Norris” in the mock response and changed the test to call the function with “Alex”:
string result = check getRandomJoke("Alex");
test:assertEquals(result, "When Alex wants an egg, he cracks open a chicken.");
Keeping the two names different was important. “Chuck Norris” was what the mock supplied. “Alex” was what I expected the function to return. When I ran bal test, it passed again.
The one-line experiment
To see what the updated test could catch, I opened main.bal and temporarily changed return replacedText; to return joke;. The function still calculated the sentence about Alex, but it returned the original joke.
This time, bal test failed. These two lines from the output showed why:
expected: 'When Alex wants an egg, he cracks open a chicken.'
actual : 'When Chuck Norris wants an egg, he cracks open a chicken.'
Ballerina also warned that replacedText was unused. That warning gave me a clue in the code, and the failed assertion showed the effect a caller would see.
I restored return replacedText; and ran bal test again:
1 passing
0 failing
0 skipped
What I learned
Both versions passed when the function worked. The difference was what would happen when it broke. The original test could pass even if every caller got a joke about Chuck Norris. The updated test caught that mistake and showed me the wrong name right in the output.
The next time I write a test, I’ll ask myself: what small change to the function should make this test fail?
Ballerina tutorial — Test services and clients · My documentation pull request
I ran the Ballerina tests and collected the results myself. I used AI to draft this article.
Top comments (0)