I'm writing an app that talks to Apple to verifyReceipts. They have both a sandbox and production url that you can post to.
When communicating with Apple, if you receive a 21007 status, it means you were posting to the production url, when you should be posting to the sandbox one.
So I wrote some code to facilitate the retry logic. Here's a simplified version of my code:
var request = require('request')
, Q = require('q')
;
var postToService = function(data, url) {
var deferred = Q.defer();
var options = {
data: data,
url: url
};
request.post(options, function(err, response, body) {
if (err) {
deferred.reject(err);
} else if (hasErrors(response)) {
deferred.reject(response);
} else {
deferred.resolve(body);
}
});
return deferred.promise;
};
exports.verify = function(data) {
var deferred = Q.defer();
postToService(data, "https://production-url.com")
.then(function(body) {
deferred.resolve(body);
})
.fail(function(err) {
if (err.code === 21007) {
postToService(data, "https://sandbox-url.com")
.then(function(body){
deferred.resolve(body);
})
.fail(function(err) {
deferred.reject(err);
});
} else {
deferred.reject(err);
}
});
return deferred.promise;
};
The retry portion in the verify function is pretty ugly and difficult to read with the nested promises. Is there a better way of doing this?
You can re-throw an error in the rejection handler to continue rejecting the promise, or you can return a new promise to replace the rejection.
exports.verify = function(data) {
return postToService(data, "https://production-url.com")
.fail(function(err) {
if (err.code === 21007) {
return postToService(data, "https://sandbox-url.com")
} else {
throw err
}
});
};
Here are a couple of possibilities. Because this question has an element of personal taste to it, you may or may not like what you see!
(Admission - I have not tested this code)
Option 1 - Use a wrapper for resolve
and reject
. This adds 'noise' in the form of the helper functions, but tidies up the rest.
var resolve = function (deferred, ob) {
return function () {
deferred.resolve(ob);
};
};
var reject = function (deferred, ob) {
return function () {
deferred.reject(ob);
};
};
exports.verify = function(data) {
var deferred = Q.defer();
postToService(data, "https://production-url.com")
.then(resolve(deferred, body))
.fail(function(err) {
if (err.code === 21007) {
postToService(data, "https://sandbox-url.com")
.then(resolve(deferred, body))
.fail(reject(deferred, err));
} else {
deferred.reject(err);
}
});
return deferred.promise;
};
Option 2 - Use bind. This has the advantage of using existing JS functionality, but you have duplicate references to deferred
when creating the callbacks.
exports.verify = function(data) {
var deferred = Q.defer();
postToService(data, "https://production-url.com")
.then(deferred.resolve.bind(deferred, body))
.fail(function(err) {
if (err.code === 21007) {
postToService(data, "https://sandbox-url.com")
.then(deferred.resolve.bind(deferred, body))
.fail(deferred.reject.bind(deferred, err));
} else {
deferred.reject(err);
}
});
return deferred.promise;
};
Option 3 - Use bind and 'method handles' (minor variation on #2).
exports.verify = function(data) {
var deferred = Q.defer();
var resolve = deferred.resolve;
var reject = deferred.reject;
postToService(data, "https://production-url.com")
.then(resolve.bind(deferred, body))
.fail(function(err) {
if (err.code === 21007) {
postToService(data, "https://sandbox-url.com")
.then(resolve.bind(deferred, body))
.fail(reject.bind(deferred, err));
} else {
deferred.reject(err);
}
});
return deferred.promise;
};
Option 4 - Monkey patch deferred.
function patch(deferred) {
deferred.resolveFn = function (ob) {
return function () {
deferred.resolve(ob);
};
};
deferred.rejectFn = function (ob) {
return function () {
deferred.reject(ob);
};
};
return deferred;
}
exports.verify = function(data) {
var deferred = patch(Q.defer());
postToService(data, "https://production-url.com")
.then(deferred.resolveFn(body))
.fail(function(err) {
if (err.code === 21007) {
postToService(data, "https://sandbox-url.com")
.then(deferred.resolveFn(body))
.fail(deferred.rejectFn(err));
} else {
deferred.reject(err);
}
});
return deferred.promise;
};
You might consider something like the following. I think judicious use of whitespace can help readability. You'll probably want to find a reasonable style standard that your team feels good about and stick with it!
exports.verify = function(data) {
var deferred = Q.defer();
postToService(data, "https://production-url.com")
.then(deferred.resolve, function(err) {
if (err.code === 21007) {
postToService(data, "https://sandbox-url.com")
.then(deferred.resolve, deferred.reject);
} else { deferred.reject(err); }
});
return deferred.promise;
};
Stuart's answer is right, the point was to chain promises. and I would like to clarify that it is not needed to use Q.defer just for wrapping. it's even considered an anti-pattern. See reasons here The Deferred anti-pattern
var request = require('request')
, Q = require('q');
var PRODUCTION_URL = "https://production-url.com",
var SANDBOX_URL = "https://sandbox-url.com",
export.verify = function() {
return postToProduction(data)
.fail( function(error) {
if (error.code === 21007 ) return postToSanbox(data);
throw error;
});
}
function postToProduction(data) {
return postToService(data, PRODUCTION_URL);
}
function postToSandbox(data) {
return postToService(data, SANDBOX_URL);
}
function postToService(data, url) {
var deferred = Q.defer();
var options = {
data: data,
url: url
};
request.post(options, function(err, response, body) {
if (err) return deferred.reject(err);
if (hasErrors(response)) return deferred.reject(response);
deferred.resolve(body);
});
return deferred.promise;
}