QUnit with Ajax, QUnit passes the failing tests

Question

I am looking into QUnit for JavaScript unit testing. I am in a strange situation where I am checking against the value returned from the Ajax call.

For the following test I am purposely trying to fail it.

// test to check if the persons are returned! 
test("getPersons", function() {
  getPersons(function(response) {
    // persons = $.evalJSON(response.d);
    equals("boo", "Foo", "The name is valid");
  });
});

But it ends up passing all the time. Here is the getPersons method that make the Ajax call.

function getPersons(callback) {
  var persons = null;

  $.ajax({
    type: "POST",
    dataType: "json",
    data: {},
    contentType: "application/json",
    url: "AjaxService.asmx/GetPersons",
    success: function(response) {
      callback(response);
    }
  });
}

Solution 1

Starting and stopping using the QUnit library seems to be working!

// test to check if the persons are returned!
test("getPersons", function() {
  stop();
  getPersons(function(response) {
    persons = $.evalJSON(response.d);
    equals(persons[0].FirstName, "Mohammad");
    start();
  });
});

Solution 2

The real problem here isn't needing to call the start() and stop() methods - in fact you could get into trouble using that approach if you are not careful in calling stop() again at the end of your callback if you have additional .ajax() methods. Which then means you find yourself in some snarled mess of having to keep track if all the callbacks have been fired to know if you still need to call stop() again.

The root of the problem involves the default behavior of asynchronous requests - and the simple solution is to make the .ajax() request happen synchronously by setting the async property to false:

test("synchronous test", function() {
  $.ajax({
    url:'Sample.asmx/Service',
    async:false,
    success: function(d,s,x) { ok(true, "Called synchronously"); }
  });
});

Even still, the best approach is to allow the asynchronous behavior and use the right test method invocation: asyncTest(). According to the docs "Asynchronous tests are queued and run one after the other. Equivalent to calling a normal test() and immediately calling stop()."

asyncTest("a test", function() {
  $.ajax({
    url: 'Sample.asmx/Service',
    success: function(d,s,x) {
      ok(true, "asynchronous PASS!");
      start();
    }
  });
});

Solution 3

There a lot of qunit test in my project. like:

    module("comment");
    asyncTest("comment1", function() {
      expect(6);
      $.ajax({
        url: 'http://xxx.com/comment',
        dataType: "json",
        type: "GET",
        timeout: 1000
      }).done(function(data) {
        ok(true, "loaded");
        ok(data.data.length>1, "array size");
        ok(data.total, "attr total");
        var c = data.data[0];
        ok(c.id, "attr c.id");
        ok(c.user_id, "attr c.user_id");
        ok(c.type == 4, "attr c.type equal 4");
      }).fail(function(x, text, thrown) {
        ok(false, "ajax failed: " + text);
      }).always(function(){
        start();
      });
    });

Solution 4

ive done some qunit testing with ajax. its not pretty. the best thing i could come with is stopping the test when ajax is fired, and starting it again in the success callback. (using start() and stop()) methods. This meant one ajax request at a time, but i could live with that. Good luck

View more solutions

Comments

[azamsharp]

Starting and stopping seems to be working! The solution is posted in the post below. Thanks!

[azamsharp]

But I wonder why my above approach was not working. I could access the response and everything. It seemed that the equals was failing!

[mkoryak]

i think that because of async nature of ajax, the test was finishing before the response came back

[dbslone]

asyncTest(name, ...) is a convenience method for test(name, function() { stop(); ... }). Nesting = bad idea.

[Goyuix]

@Jörn - thanks for pointing out the nesting. Reading back over this answer I am not sure why I wrapped it in the first place other than to perhaps illustrate you could. I have unwrapped the asyncTest method to make it more clear how it is intended to be used.

[Giovanni Cappellotto]

As @Goyuix said you can also use asyncTest() instead of test() to avoid calling the stop() function and to make it explicit that this is an async test.