promise using Q about error propagation

q_error_propagation.js

    var d = Q.defer(),
            ret = [],
            p = d.promise;

    var p2 = p.then(
            function (v) {
                ret.push("e1 :" + v);
                throw "e1";
            },
            function (r) {
                ret.push("e2 :" + r);
                return "e2";
            });

    var p3 = p2.then(
            function (v) {
                ret.push("e3 :" + v);
                throw "e3";
            },
            function (r) {
                ret.push("e4 :" + r);
                return "e4";
            });

    var p4 = p3.then(function (v) {
        ret.push("e5 :" + v);
        throw "e5";
    }, function (r) {
        ret.push("e6 :" + r);
        return "e6";
    });

    setTimeout(function () {
        d.resolve(1);
    }, 100);

    setTimeout(function () {
        alert(ret.join("\n"));
    }, 200);

Why would you expect e6: e4?

Consider:

var p4 = p3.then(function (v) {
    ret.push("e5 :" + v);
    throw "e5";
}, function (r) {
    ret.push("e6 :" + r);
    return "e6";
});

p3 is a fulfilled promise, since the error handler does return "e4" instead of re-throwing. This signals successful recovery, just like not rethrowing in a catch block.

Thus, the fulfillment handler for p3 will be run with the value p3 is fulfilled with, viz. "e4". So "e5 :e4" is as expected.

To get your expected behavior, just re-throw instead of recovering: change line #22 to throw "e4" instead of return "e4".

To be further illustrative, your code is the equivalent of:

function p() {
  return 1;
}

function p2() {
  var v;
  try {
    v = p();
  } catch (r) {
    ret.push("e2 :" + r);
    return "e2";
  }

  ret.push("e1 :" + v);
  throw "e1";
}

function p3() {
  var v;
  try {
    v = p2();
  } catch (r) {
    ret.push("e4: " + r);
    return "e4";
  }

  ret.push("e3 :" + v);
  throw "e3";
}

function p4() {
  var v;
  try {
    v = p3();
  } catch (r) {
    ret.push("e6 :" + r);
    return "e6"; 
  }

  ret.push("e5 :" + v);
  throw "e5";
}

p4();

As you can see, the code path that gets followed is:

• p() executes without errors; `"e1 :1"
• p2() then throws "e1"
• So p3's catch clause gets entered: "e4: e1"
• That catch clause returns "e4", i.e. p3() returns "e4"
• So p4's catch clause is never entered: "e5 :e4".

@DomenicDenicola :

Thank you , i misunderstood the error callback of promise at first, i thought it is just used for error propagation and notification.

Now i see it can also be used for recovery :)

@yiminghe happy to help! :)