Flowly

High performance async-like flow control library with data retention and short-circuiting

npm install flowly

This module is aimed at a simple and high performance variant of async and async.series. The primary benefits of Flow are the ability to flow.cbLast(), access data from previous steps in the current series via flow.data and flow.halt() to early exit with non-error return data. This allows developers to eliminate a lot of boilerplate error checking and variable stashing.

Another benefit is it runs roughly 4x faster than async.series.

Getting Started

The usage is very similar to async. In fact you can quickly replace almost any async.series with flowly.series and it will work with no regression and an instant speed boost.

var flowly = require("flowly");
flowly.series({
	step1 : (cb) => {
		// step1
	},
	step2 : (cb) => {
		// step2
	},
	step3 : (cb) => {
		// step3
	}
}, (err, results) => {
});

To see the benefits of flowly lets look at a simple async.series example.

var getPosts = function(cb) {
	var data;
	async.series({
		getData : (cb) => db.collection("foo").find({}, function(err, temp) {
			if (err) { return cb(err); }
			
			data = temp;
			
			return cb(null);
		}),
		transform : (cb) => {
			var temp = data.map(val => ({ url : val.title + "/" + val.recid + "/", title : val.title, recid : val.recid }));
			cb(null, temp);
		}
	}, (err, data) => {
		if (err) { return cb(err); }
		
		return cb(null, data.transform);
	});
}

If you call getPosts(function(err, data) {}) it will callback return the data from the transform step. Now lets do that same code in flowly.

var getPosts = function(cb) {
	async.series({
		getData : (cb) => db.collection("foo").find({}, cb),
		transform : (cb) => {
			var temp = flow.data.getData.map(val => ({ url : val.title + "/" + val.recid + "/", title : val.title, recid : val.recid }));
			cb(null, temp);
		}
	}, flow.cbLast(cb));
}

1. By utilizing flow.data we are able to get the returned data from the first getData step without setting aside the data variable. The need to set aside a variable also forced us to do if (err) { return cb(err); } boilerplate.
2. By using flow.cbLast(cb) we return the data from the transform step, again eliminating another set of error checking.
3. If we add more steps to our series in the future, it's easy and doesn't require a whole refactor of the function. In async if you add another step, often you end up moving the final callback into the primary series, and it gets more unwieldy.

Benchmark

npm run simplebench

As of 8/1/2017 on Node 7.10.0, higher ops/sec is better.

Group:  default
Winner - native callbacks

native callbacks - count: 1309506, ops/sec: 1309506
new flowly.Flow - count: 801229, ops/sec: 801229, diff: -38.81%
neoasync.series - count: 495107, ops/sec: 495107, diff: -62.19%
async.series - count: 183600, ops/sec: 183600, diff: -85.98%
native promises - count: 73923, ops/sec: 73923, diff: -94.35%

Documentation

flowly.series(calls, cb)

A shortcut execution of var flow = new flowly.Flow(); flow.series(calls, cb);. See the documentation for Flow.prototype.series for details.

flowly.batch(args, cb)

This function can help when you want to split an array up into multiple bulk calls and concat the result. A common example is if you need to make an api call to an external service, but that service only allows you to pass 50 ids at a time. This will allow you to split up any length of ids, burst them out 50 ids at a time and then work with the total result.

• args - object - required - Args
  • items - array - required - Array of items to batch across, can be an array of any kind of values.
  • batchSize - number - required - What size batches to break the groups up into.
  • concat - boolean - optional default false - If each run returns an array, concat will combine those into a single array. Cannot be used if each step does not return an array.
  • merge - boolean - optional default false - If each run returns an object, merge will Object.assign() them together into a single object. Cannot be used if each step does not return an object.
  • fn - function - required - The processor function, it will be called with like fn(itemsArr, cb) and the itemsArr will be an array of items assigned to that batch. This function should return an array. If you do not care about the return value from the batches, you will want to return [].

var batch = 0;
flowly.batch({
	items : [1,2,3,4,5,6,7,8,9,10],
	batchSize : 3,
	fn : function(items, cb) {
		var myBatch = batch++;
		return cb(null, items.map(val => `${myBatch}_${val}`));
	}
}, function(err, result) {
	if (err) { throw err; }
	
	// result
	// the fn was called 4 times, and the results of each batch were concatted together
	// ["1_1", "1_2", "1_3", "2_4", "2_5", "2_6", "3_7", "3_8", "3_9", "4_10"]
});

Flow

Constructor

var flow = new flowly.Flow(options);

• options - object - default undefined - Optional options argument
  • timers - boolean - default false - Whether to time each step in the flow. Should only be done when debugging.

Flow.prototype.series(calls, cb)

• calls - array or object - The calls object or array. For an object each key is the name of the step and it's value is a function. For array, each entry is a function
• cb - function - Overall callback function to receive the results object.

calls {} style

When you pass your calls in {} syntax, the flow.data is accessed by the keyname, and the results of the final callback is indexed by key

var flow = new flowly.Flow();
flow.series({
	call1 : function(cb) {
		return cb(null, "foo");
	},
	call2 : function(cb) {
		// flow.data.call1 === "foo"
		return cb(null, "bar");
	}
}, function(err, results) {
	// results === flow.data === { call1 : "foo", call2 : "bar" }
});

calls [] style

When you pass your calls in [] syntax, the flow.data is accessed by index, and the results of the final callback is indexed by index.

var flow = new flowly.Flow();
flow.series([
	function(cb) {
		return cb(null, "foo");
	},
	function(cb) {
		// flow.data[0] === "foo"
		return cb(null, "bar");
	}
], function(err, results) {
	// results === flow.data === ["foo", "bar"]
});

With all callbacks if the function returns 2 arguments (null, "foo") then the flow.data contains just the non-error portion. If the function returns more than 2 values, flow.data will contain an array of values.

step1 : (cb) => cb(null, "foo")
flow.data.step1 === "foo"

step1 : (cb) => cb(null, "foo", "bar", "baz")
flow.data.step1 === ["foo", "bar", "baz"]

As with async if the function ever returns a truthy value on the first return argument, it will short-circuit and skip all subsequent steps.

In the following example step2 is never executed, and the overall function will cb the error from step1.

flow.series({
	step1 : (cb) => cb(new Error("fail")),
	step2 : (cb) => will not get here!
}, flow.cbLast(cb))

Flow.prototype.cbLast(callback)

var getPosts = function(cb) {
	var flow = new flowly.Flow();
	flow.series({
		getData : (cb) => db.collection("foo").find({}, cb),
		transform : (cb) => {
			var temp = getData.map(val => ({ url : val.title + "/" + val.recid + "/", title : val.title, recid : val.recid }));
			cb(null, temp);
		}
	}, flow.cbLast(cb));
}

In this case if you execute getPosts(function(err, data) {}) the data returned will be the result of the final transform step, since it's the final executed step in the series.

flow.data

flow.data allows access to the data keys returned from all previous steps to the current step. See the syntaxes for the different calls formats to see what the data structure is like.

flow.last

flow.last returns the data from the last cb to execute. If a flow.halt was performed, it's the result of that. Usually the best practice is to use flow.cbLast(cb) directly.

Flow.prototype.halt(arg1, arg2, ...)

Halts the execution of the series at the current step and proceed to the series callback. flow.last will be set to the return of this function.

If you intend to cb an error, do not use flow.halt instead simply cb the error as it will already shortcircuit. flow.halt() is used when you want to skip further steps but still return valid.

In the following example you can see we get a user, if we don't find a user, it's not an error condition, but we don't want to run the permissions step, so we exit out.

var flow = new flowly.Flow();
flow.series({
	user : (cb) => getUser(username, password, cb),
	permissions : (cb) => {
		if (flow.data.user === undefined) { return flow.halt({ message : "User was not found." }); }
		
		getPermissions(flow.data.user.roleId);
	},
	result : (cb) => {
		flow.data.user.permissions = flow.data.permissions;
		return cb(null, { message : "Logged in", user : flow.data.user });
	}
}, flow.cbLast(cb));

Flow.prototype.if

Conditionally execute a callback only if the return condition is true.

In the following example, we only execute the getPermission call if the user is not an admin. If the condition is not true the step is equivalent to cb(null).

var flow = new flowly.Flow();
flow.series({
	user : (cb) => getUser(username, password, cb),
	permissions : flow.if(() => flow.data.user.isAdmin === false, getPermission(flow.data.user.roleId, cb)),
	result : (cb) => {
		flow.data.user.permissions = flow.data.permissions;
		return cb(null, { message : "Logged in", user : flow.data.user });
	}
})