Nodejs的进阶指南


The Node Craftsman Book An advanced Node.js tutorial Manuel Kiessling This book is for sale at http://leanpub.com/nodecraftsman This version was published on 2015-04-02 This is a Leanpub book. Leanpub empowers authors and publishers with the Lean Publishing process. Lean Publishing is the act of publishing an in-progress ebook using lightweight tools and many iterations to get reader feedback, pivot until you have the right book and build traction once you do. ©2013 - 2015 Manuel Kiessling Tweet This Book! Please help Manuel Kiessling by spreading the word about this book on Twitter! The suggested hashtag for this book is #nodecraftsman. Find out what other people are saying about the book by clicking on this link to search for this hashtag on Twitter: https://twitter.com/search?q=#nodecraftsman Also By Manuel Kiessling The Node Beginner Book Node入门 El Libro Principiante de Node Livro do Iniciante em Node For Nora, Aaron and Melinda. Contents Preface ............................................... i About .............................................. i Status .............................................. i Notes on code formatting .................................... i Intended audience ........................................ i Part 1: Node.js basics in detail ............................... 1 Introduction to Part 1 ....................................... 2 Working with NPM and Packages ................................ 3 Test-Driven Node.js Development ................................ 5 Object-oriented JavaScript .................................... 11 Blueprints versus finger-pointing ................................ 11 A classless society ........................................ 12 Creating objects ......................................... 12 Object-orientation, prototyping, and inheritance ....................... 20 A classless society, revisited .................................. 21 Summary ............................................ 23 Synchronous and Asynchronous operations explained .................... 25 Visualizing the Node.js execution model ............................ 25 Blocking and non-blocking operations ............................. 29 Using and creating Event Emitters ............................... 33 Introduction ........................................... 33 Creating your own Event Emitter object ............................ 37 Summary ............................................ 47 Optimizing code performance and control flow management using the async library .. 48 Executing expensive asynchronous background tasks in parallel . . . . . . . . . . . . . . . 49 Optimizing code structure with async ............................. 55 CONTENTS Node.js and MySQL ........................................ 61 Using the node-mysql library .................................. 61 A first database application ................................... 61 Using node-mysql’s Streaming API .............................. 65 Making SQL queries secure against attacks .......................... 70 Summary ............................................ 78 Node.js and MongoDB ...................................... 79 Some MongoDB basics ..................................... 79 Applying CRUD operations with the low-level mongodb driver . . . . . . . . . . . . . . . 80 Retrieving specific documents using filters .......................... 86 More complex update operations ................................ 90 Working with indexes ...................................... 92 Querying collections efficiently ................................ 95 Summary ............................................ 99 Part 2: Building a complete web application with Node.js and AngularJS . 100 Introduction ............................................ 101 The requirements from a user’s perspective . . . . . . . . . . . . . . . . . . . . . . . . . . 101 High level architecture overview ................................102 Setting up the development environment . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 Milestone 1: A first passing test against the server ...................... 106 Milestone 2: The API responds with actual database content ................ 110 Abstracting database access ..................................110 Ensuring a clean slate for test runs ...............................112 Completing the first spec ....................................113 Milestone 3: Setting the stage for a continuous delivery workflow ............. 117 Introducing automatic database migrations . . . . . . . . . . . . . . . . . . . . . . . . . . 120 Milestone 4: Giving users a frontend .............................. 125 Setting up frontend dependencies through bower . . . . . . . . . . . . . . . . . . . . . . . 125 Serving the frontend through the backend server . . . . . . . . . . . . . . . . . . . . . . . 125 Adding the frontend code ....................................126 Adding AngularJS view templates ...............................133 Milestone 5: More work on the backend ............................ 136 Adding a route for retrieving categories ............................136 Making the backend recognize different environments . . . . . . . . . . . . . . . . . . . . 142 Milestone 6: Completing the backend and finalizing the application ............ 147 CONTENTS Creating the route for adding new keywords . . . . . . . . . . . . . . . . . . . . . . . . . 147 Creating the route for updating keywords . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 Creating the route for deleting keywords . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 Summary .............................................. 160 Epilogue .............................................. 161 Preface About The aim of this book is to help beginning JavaScript programmers who already know how to write basic Node.js applications in mastering JavaScript and Node.js thoroughly. Status This book is finished and will only receive updates regarding errors. It was last updated on April 2, 2015. All code examples have been tested to work with Node.js v0.12. Notes on code formatting Please note that long lines in code examples may receive a line-break, denoted by the backslash sign \. Take care to put those lines into your editor as one line. Take, for example, this code block: 1 dbSession.fetchAll('SELECTid,value,categoryIDFROMkeywordORDERBYi\ 2 d', function(err,rows){ The PDF version of this book shows a line break right after the i of the ORDER BY id part, denoted by a \ after the i. However, in your editor, you need to make sure that the whole code resides on one line. Also, note that from what I can see, code blocks can not be copy-pasted from the PDF version of this book. While this might seem like a major annoyance, experience shows that learning to program works way better if one is forced to transcribe code from a book into the editor. Intended audience This book will fit best for readers that are familiar with the basic concepts of JavaScript and have already written some basic Node.js applications. As this book is a sequel to The Node Beginner Book, I recommend reading it before starting with this book. Part 1: Node.js basics in detail Introduction to Part 1 The goal of this book is to enable you to write Node.js programs of any level of complexity, from simple utilities to complex applications that use several external modules and consist of several layers of code, talking to external systems and serving thousands of users. In order to do so, you need to learn about all the different aspects of Node.js - the tools, the methodologies, the libraries, the APIs, the best practices - and you need to learn how to put all of that together to create a working whole. Therefore, I have split this book into two parts: A collection of different basics in Part 1, and a thorough tutorial on how to put these basics together to build a complex application in Part 2. In Part 1, every chapter stands on its own and isn’t directly related to the other chapters. Part 2 is more like a continuous story that starts from scratch and gives you a finished and working application at the end. Thus, let’s start with Part 1 and look at all the different facets of Node.js software development. Working with NPM and Packages We already used NPM, the Node Package Manager, in order to install a single package and its dependencies for the example project in The Node Beginner Book. However, there is much more to NPM, and a more thorough introduction is in order. The most useful thing that NPM does isn’t installing packages. This could be done manually with slightly more hassle. What’s really useful about NPM is that it handles package dependencies. A lot of packages that we need for our own project need other packages themselves. Have a look at https://npmjs.org/package/request, for example. It’s the overview page for the NPM package request. According to its description, it provides a “simplified HTTP request client”. But in order to do so, request not only uses its own code. It also needs other packages for doing its job. These are listed under “Dependencies”: qs, json-stringify-safe, and others. Whenever we use the NPM command line tool, npm, in order to install a package, NPM not only pulls the package itself, but also its dependencies, and installs those as well. Using npm install request is simply a manual way to implicitly say “my project depends on request, please install it for me”. However, there is an explicit way of defining dependencies for our own projects, which also allows to have all dependencies resolved and installed automatically. In order to use this mechanism, we need to create a control file within our project that defines our dependencies. This control file is then used by NPM to resolve and install those dependencies. This control file must be located at the top level folder of our project, and must be named package.json. This is what a package.json file looks like for a project that depends on request: { "dependencies":{ "request":"" } } Having this file as part of our code base makes NPM aware of the dependencies of our project without the need to explicitly tell NPM what to install by hand. We can now use NPM to automatically pull in all dependencies of our project, simply by executing npm install within the top level folder of our code base. In this example, this doesn’t look like much of a convenience win compared to manually installing request, but once we have more than a handful of dependencies in our projects, it really makes a difference. Working with NPM and Packages 4 The package.json file also allows us to “pin” dependencies to certain versions, i.e., we can define a version number for every dependency, which ensures that NPM won’t pull the latest version automatically, but exactly the version of a package we need: { "dependencies":{ "request":"2.27.0" } } In this case, NPM will always pull request version 2.27.0 (and the dependencies of this version), even if newer versions are available. Patterns are possible, too: { "dependencies":{ "request":"2.27.x" } } The x is a placeholder for any number. This way, NPM would pull in request version 2.27.0 and 2.27.5, but not 2.28.0. The official documentation at https://npmjs.org/doc/json.html#dependencies has more examples of possible dependency definitions. Please note that the package.json file does much more than just defining dependencies. We will dig deeper in the course of this book. For now, we are prepared to use NPM for resolving the dependencies that arise in our first project - our first test-driven Node.js application. Test-Driven Node.js Development The code examples in The Node Beginner Book only described a toy project, and we came away with not writing any tests for it. If writing tests is new for you, and you have not yet worked on software in a test-driven manner, then I invite you to follow along and give it a try. We need to decide on a test framework that we will use to implement our tests. A lack of choice is not an issue in the JavaScript and Node.js world, as there are dozens of frameworks available. Personally, I prefer Jasmine, and will therefore use it for my examples. Jasmine is a framework that follows the philosophy of behaviour-driven development, which is kind of a “subculture” within the community of test-driven developers. This topic alone could easily fill its own book, thus I’ll give only a brief introduction. The idea is to begin development of a new software unit with its specification, followed by its implementation (which, by definition, must satisfy the specification). Let’s make up a real world example: we order a table from a carpenter. We do so by specifying the end result: “I need a table with a top that is 6 x 3 ft. The height of the top must be adjustable between 2.5 and 4.0 ft. I want to be able to adjust the top’s height without standing up from my chair. I want the table to be black, and cleaning it with a wet cloth should be possible without damaging the material. My budget is $500.” Such a specification allows to share a goal between us and the carpenter. We don’t have to care for how exactly the carpenter will achieve this goal. As long as the delivered product fits our specification, both of us can agree that the goal has been reached. With a test-driven or behaviour-driven approach, this idea is applied to building software. You wouldn’t build a piece of software and then define what it’s supposed to do. You need to know in advance what you expect a unit of software to do. Instead of doing this vaguely and implicitly, a test-driven approach asks you to do the specification exactly and explicitly. Because we work on software, our specification can be software, too: we only need to write functions that check if our unit does what it is expected to do. These check functions are unit tests. Let’s create a software unit which is covered by tests that describe its expected behaviour. In order to actually drive the creation of the code with the tests, let’s write the tests first. We then have a clearly defined goal: making the tests pass by implementing code that fulfills the expected behaviour, and nothing else. In order to do so, we create a new Node.js project with two folders in it: Test-Driven Node.js Development 6 src/ spec/ spec is where our test cases go - in Jasmine lingo, these are called “specifications”, hence “spec”. The spec folder mirrors the file structure under src, i.e., a source file at src/foo.js is mirrored by a specification at spec/fooSpec.js. Following the tradition, we will test and implement a “Hello World” code unit. Its expected behaviour is to return a string “Hello Joe!” if called with “Joe” as its first and only parameter. This behaviour can be specified by writing a unit test. To do so, we create a file spec/greetSpec.js, with the following content: 1 'use strict'; 2 3 var greet = require('../src/greet'); 4 5 describe('greet', function() { 6 7 it('should greet the given name', function() { 8 expect(greet('Joe')).toEqual('Hello Joe!'); 9 }); 10 11 it('should greet no-one special if no name is given', function() { 12 expect(greet()).toEqual('Hello world!'); 13 }); 14 15 }); This is a simple, yet complete specification. It is a programmatical description of the behaviour we expect from a yet-to-be-written function named greet. The specification says that if the function greet is called with Joe as its first and only parameter, the return value of the function call should be the string “Hello Joe!”. If we don’t supply a name, the greeting should be generic. As you can see, Jasmine specifications have a two-level structure. The top level of this structure is a describe block, which consists of one or more it blocks. An it block describes a single expected behaviour of a single unit under test, and a describe block summarizes one or more blocks of expected behaviours, therefore completely specifying all expected behaviours of a unit. Let’s illustrate this with a real-world “unit” described by a Jasmine specification: Test-Driven Node.js Development 7 describe('A candle', function() { it('should burn when lighted', function() { //... }); it('should grow smaller while burning', function() { //... }); it('should no longer burn when all wax has been burned', function() { //... }); it('should go out when no oxygen is available to it', function() { //... }); }); As you can see, a Jasmine specification gives us a structure to fully describe how a given unit should behave. Not only can we describe expected behaviour, we can also verify it. This can be done by running the test cases in the specification against the actual implementation. After all, our Jasmine specification is just another piece of JavaScript code which can be executed. The NPM package jasmine-node ships with a test case runner which allows us to execute the test case, with the added benefit of a nice progress and result output. Let’s create a package.json file that defines jasmine-node as a dependency of our application - then we can start running the test cases of our specification. As described earlier, we need to place the package.json file at the topmost folder of our project. Its content should be as follows: { "devDependencies":{ "jasmine-node":"" } } We talked about the dependencies section of package.json before - but here we declare jasmine-node in a devDependencies block. The result is basically the same: NPM knows about this dependency Test-Driven Node.js Development 8 and installs the package and its dependencies for us. However, dev dependencies are not needed to run our application - as the name suggests, they are only needed during development. NPM allows to skip dev dependencies when deploying applications to a production system - we will get to this later. In order to have NPM install jasmine-node, please run npm install in the top folder of your project. We are now ready to test our application against its specification. Of course, our greet function cannot fulfill its specification yet, simply because we have not yet implemented it. Let’s see how this looks by running the test cases. From the root folder of our new project, execute the following: ./node_modules/jasmine-node/bin/jasmine-node spec/greetSpec.js As you can see, Jasmine isn’t too happy with the results yet. We refer to a Node module in src/greet.js, a file that doesn’t even exist, which is why Jasmine bails out before even starting the tests: Exception loading: spec/greetSpec.js { [Error: Cannot find module '../src/greet'] code: 'MODULE_NOT_FOUND' } Well, let’s create the module, in file src/greet.js: 1 'use strict'; 2 3 var greet = function() {}; 4 5 module.exports = greet; Now we have a general infrastructure, but of course we do not yet behave as the specification wishes. Let’s run the test cases again: Test-Driven Node.js Development 9 FF Failures: 1) greet should greet the given name Message: TypeError: object is not a function Stacktrace: TypeError: object is not a function at null. (./spec/greetSpec.js:8:12) 2) greet should greet no-one special if no name is given Message: TypeError: object is not a function Stacktrace: TypeError: object is not a function at null. (./spec/greetSpec.js:12:12) Finished in 0.015 seconds 2 tests, 2 assertions, 2 failures, 0 skipped Jasmine tells us that it executed two test cases that contained a total of two assertions (or expectations), and because these expectations could not be satisfied, the test run ended with two failures. It’s time to satisfy the first expectation of our specification, in file src/greet.js: 1 'use strict'; 2 3 var greet = function(name) { 4 return 'Hello ' + name + '!'; 5 }; 6 7 module.exports = greet; Another test case run reveals that we are getting closer: Test-Driven Node.js Development 10 .F Failures: 1) greet should greet no-one special if no name is given Message: Expected 'Hello undefined!' to equal 'Hello world!'. Stacktrace: Error: Expected 'Hello undefined!' to equal 'Hello world!'. at null. (spec/greetSpec.js:12:21) Finished in 0.015 seconds 2 tests, 2 assertions, 1 failure, 0 skipped Our first test case passes - greet can now correctly greet people by name. We still need to handle the case where no name was given: 1 'use strict'; 2 3 var greet = function(name) { 4 if (name === undefined){ 5 name = 'world'; 6 } 7 return 'Hello ' + name + '!'; 8 }; 9 10 module.exports = greet; And that does the job: .. Finished in 0.007 seconds 2 tests, 2 assertions, 0 failures, 0 skipped We have now created a piece of software that behaves according to its specification. You’ll probably agree that our approach to create this unbelievably complex unit of software - the greet function - in a test-driven way doesn’t prove the greatness of test-driven development in any way. That’s not the goal of this chapter. It merely sets the stage for what’s to come. We are going to create real, comprehensive software through the course of this book, and this is where the advantages of a test-driven approach can be experienced. Object-oriented JavaScript Let’s talk about object-orientation and inheritance in JavaScript. The good news is that it’s actually quite simple, but the bad news is that it works completely different than object-orientation in languages like C++, Java, Ruby, Python or PHP, making it not-quite-so simple to understand. But fear not, we are going to take it step by step. Blueprints versus finger-pointing Let’s start by looking at how “typical” object-oriented languages actually create objects. We are going to talk about an object called myCar. myCar is our bits-and-bytes representation of an incredibly simplified real world car. It could have attributes like color and weight, and methods like drive and honk. In a real application, myCar could be used to represent the car in a racing game - but we are going to completely ignore the context of this object, because we will talk about the nature and usage of this object in a more abstract way. If you would want to use this myCar object in, say, Java, you need to define the blueprint of this specific object first - this is what Java and most other object-oriented languages call a class. If you want to create the object myCar, you tell Java to “build a new object after the specification that is laid out in the class Car”. The newly built object shares certain aspects with its blueprint. If you call the method honk on your object, like so: myCar.honk(); then the Java VM will go to the class of myCar and look up which code it actually needs to execute, which is defined in the honk method of class Car. Ok, nothing shockingly new here. Enter JavaScript. Object-oriented JavaScript 12 A classless society JavaScript does not have classes. But as in other languages, we would like to tell the interpreter that it should build our myCar object following a certain pattern or schema or blueprint - it would be quite tedious to create every car object from scratch, “manually” giving it the attributes and methods it needs every time we build it. If we were to create 30 car objects based on the Car class in Java, this object-class relationship provides us with 30 cars that are able to drive and honk without us having to write 30 drive and honk methods. How is this achieved in JavaScript? Instead of an object-class relationship, there is an object-object relationship. Where in Java our myCar, asked to honk, says “go look at this class over there, which is my blueprint, to find the code you need”, JavaScript says “go look at that other object over there, which is my prototype, it has the code you are looking for”. Building objects via an object-object relationship is called Prototype-based programming, versus Class-based programming used in more traditional languages like Java. Both are perfectly valid implementations of the object-oriented programming paradigm - it’s just two different approaches. Creating objects Let’s dive into code a bit, shall we? How could we set up our code in order to allow us to create our myCar object, ending up with an object that is a Car and can therefore honk and drive? Well, in the most simple sense, we can create our object completely from scratch, or ex nihilo if you prefer the boaster expression. It works like this: 1 var myCar = {}; 2 3 myCar.honk = function() { 4 console.log('honk honk'); 5 }; 6 7 myCar.drive = function() { 8 console.log('vrooom...'); 9 }; This gives us an object called myCar that is able to honk and drive: Object-oriented JavaScript 13 myCar.honk(); // outputs "honk honk" myCar.drive(); // outputs "vrooom..." However, if we were to create 30 cars this way, we would end up defining the honk and drive behaviour of every single one, something we said we want to avoid. In real life, if we made a living out of creating, say, pencils, and we don’t want to create every pencil individually by hand, then we would consider building a pencil-making machine, and have this machine create the pencils for us. After all, that’s what we implicitly do in a class-based language like Java - by defining a class Car, we get the car-maker for free: Car myCar = new Car(); will build the myCar object for us based on the Car blueprint. Using the new keyword does all the magic for us. JavaScript, however, leaves the responsibility of building an object creator to us. Furthermore, it gives us a lot of freedom regarding the way we actually build our objects. In the most simple case, we can write a function which creates “plain” objects that are exactly like our “ex nihilo” object, and that don’t really share any behaviour - they just happen to roll out of the factory with the same behaviour copied onto every single one, if you want so. Or, we can write a special kind of function that not only creates our objects, but also does some behind-the-scenes magic which links the created objects with their creator. This allows for a true sharing of behaviour: functions that are available on all created objects point to a single implementation. If this function implementation changes after objects have been created, which is possible in JavaScript, the behaviour of all objects sharing the function will change accordingly. Let’s examine all possible ways of creating objects in detail. Using a simple function to create plain objects In our first example, we created a plain myCar object out of thin air - we can simply wrap the creation code into a function, which gives us a very basic object creator: 1 var makeCar = function() { 2 var newCar = {}; 3 newCar.honk = function() { 4 console.log('honk honk'); 5 }; 6 }; For the sake of brevity, the drive function has been omitted. We can then use this function to mass-produce cars: Object-oriented JavaScript 14 1 var makeCar = function() { 2 var newCar = {} 3 newCar.honk = function() { 4 console.log('honk honk'); 5 }; 6 return newCar; 7 }; 8 9 myCar1 = makeCar(); 10 myCar2 = makeCar(); 11 myCar3 = makeCar(); One downside of this approach is efficiency: for every myCar object that is created, a new honk function is created and attached - creating 1,000 objects means that the JavaScript interpreter has to allocate memory for 1,000 functions, although they all implement the same behaviour. This results in an unnecessarily high memory footprint of the application. Secondly, this approach deprives us of some interesting opportunities. These myCar objects don’t share anything - they were built by the same creator function, but are completely independent from each other. It’s really like with real cars from a real car factory: They all look the same, but once they leave the assembly line, they are totally independent. If the manufacturer should decide that pushing the horn on already produced cars should result in a different type of honk, all cars would have to be returned to the factory and modified. In the virtual universe of JavaScript, we are not bound to such limits. By creating objects in a more sophisticated way, we are able to magically change the behaviour of all created objects at once. Using a constructor function to create objects In JavaScript, the entities that create objects with shared behaviour are functions which are called in a special way. These special functions are called constructors. Let’s create a constructor for cars. We are going to call this function Car, with a capital C, which is common practice to indicate that this function is a constructor. In a way, this makes the constructor function a class, because it does some of the things a class (with a constructor method) does in a traditional OOP language. However, the approach is not identical, which is why constructor functions are often called pseudo- classes in JavaScript. I will simply call them classes or constructor functions. Because we are going to encounter two new concepts that are both necessary for shared object behaviour to work, we are going to approach the final solution in two steps. Object-oriented JavaScript 15 Step one is to recreate the previous solution (where a common function spilled out independent car objects), but this time using a constructor: 1 var Car = function() { 2 this.honk = function() { 3 console.log('honk honk'); 4 }; 5 }; When this function is called using the new keyword, like so: var myCar = new Car(); it implicitly returns a newly created object with the honk function attached. Using this and new makes the explicit creation and return of the new object unnecessary - it is created and returned “behind the scenes” (i.e., the new keyword is what creates the new, “invisible” object, and secretly passes it to the Car function as its this variable). You can think of the mechanism at work a bit like in this pseudo-code: 1 // Pseudo-code, for illustration only! 2 3 var Car = function(this){ 4 this.honk = function() { 5 console.log('honk honk'); 6 }; 7 return this; 8 }; 9 10 var newObject = {}; 11 var myCar = Car(newObject); As said, this is more or less like our previous solution - we don’t have to create every car object manually, but we still cannot modify the honk behaviour only once and have this change reflected in all created cars. But we laid the first cornerstone for it. By using a constructor, all objects received a special property that links them to their constructor: Object-oriented JavaScript 16 1 var Car = function() { 2 this.honk = function() { 3 console.log('honk honk'); 4 }; 5 }; 6 7 var myCar1 = new Car(); 8 var myCar2 = new Car(); 9 10 console.log(myCar1.constructor); // outputs [Function: Car] 11 console.log(myCar2.constructor); // outputs [Function: Car] All created myCars are linked to the Car constructor. This is what actually makes them a class of related objects, and not just a bunch of objects that happen to have similar names and identical functions. Now we have finally reached the moment to get back to the mysterious prototype we talked about in the introduction. Using prototyping to efficiently share behaviour between objects As stated there, while in class-based programming the class is the place to put functions that all objects will share, in prototype-based programming, the place to put these functions is the object which acts as the prototype for our objects at hand. But where is the object that is the prototype of our myCar objects - we didn’t create one! It has been implicitly created for us, and is assigned to the Car.prototype property (in case you wondered, JavaScript functions are objects too, and they therefore have properties). Here is the key to sharing functions between objects: Whenever we call a function on an object, the JavaScript interpreter tries to find that function within the queried object. But if it doesn’t find the function within the object itself, it asks the object for the pointer to its prototype, then goes to the prototype, and asks for the function there. If it is found, it is then executed. This means that we can create myCar objects without any functions, create the honk function in their prototype, and end up having myCar objects that know how to honk - because everytime the interpreter tries to execute the honk function on one of the myCar objects, it will be redirected to the prototype, and execute the honk function which is defined there. Here is how this setup can be achieved: Object-oriented JavaScript 17 1 var Car = function() {}; 2 3 Car.prototype.honk = function() { 4 console.log('honk honk'); 5 }; 6 7 var myCar1 = new Car(); 8 var myCar2 = new Car(); 9 10 myCar1.honk(); // executes Car.prototype.honk() and outputs "honk honk" 11 myCar2.honk(); // executes Car.prototype.honk() and outputs "honk honk" Our constructor is now empty, because for our very simple cars, no additional setup is necessary. Because both myCars are created through this constructor, their prototype points to Car.prototype - executing myCar1.honk() and myCar2.honk() always results in Car.prototype.honk() being executed. Let’s see what this enables us to do. In JavaScript, objects can be changed at runtime. This holds true for prototypes, too. Which is why we can change the honk behaviour of all our cars even after they have been created: 1 var Car = function() {}; 2 3 Car.prototype.honk = function() { 4 console.log('honk honk'); 5 }; 6 7 var myCar1 = new Car(); 8 var myCar2 = new Car(); 9 10 myCar1.honk(); // executes Car.prototype.honk() and outputs "honk honk" 11 myCar2.honk(); // executes Car.prototype.honk() and outputs "honk honk" 12 13 Car.prototype.honk = function() { 14 console.log('meep meep'); 15 }; 16 17 myCar1.honk(); // executes Car.prototype.honk() and outputs "meep meep" 18 myCar2.honk(); // executes Car.prototype.honk() and outputs "meep meep" Of course, we can also add additional functions at runtime: Object-oriented JavaScript 18 1 var Car = function() {}; 2 3 Car.prototype.honk = function() { 4 console.log('honk honk'); 5 }; 6 7 var myCar1 = new Car(); 8 var myCar2 = new Car(); 9 10 Car.prototype.drive = function() { 11 console.log('vrooom...'); 12 }; 13 14 myCar1.drive(); // executes Car.prototype.drive() and outputs "vrooom..." 15 myCar2.drive(); // executes Car.prototype.drive() and outputs "vrooom..." But we could even decide to treat only one of our cars differently: 1 var Car = function() {}; 2 3 Car.prototype.honk = function() { 4 console.log('honk honk'); 5 }; 6 7 var myCar1 = new Car(); 8 var myCar2 = new Car(); 9 10 myCar1.honk(); // executes Car.prototype.honk() and outputs "honk honk" 11 myCar2.honk(); // executes Car.prototype.honk() and outputs "honk honk" 12 13 myCar2.honk = function() { 14 console.log('meep meep'); 15 }; 16 17 myCar1.honk(); // executes Car.prototype.honk() and outputs "honk honk" 18 myCar2.honk(); // executes myCar2.honk() and outputs "meep meep" It’s important to understand what happens behind the scenes in this example. As we have seen, when calling a function on an object, the interpreter follows a certain path to find the actual location of that function. While for myCar1, there still is no honk function within that object itself, that no longer holds true for myCar2. When the interpreter calls myCar2.honk(), there now is a function within myCar2 itself. Object-oriented JavaScript 19 Therefore, the interpreter no longer follows the path to the prototype of myCar2, and executes the function within myCar2 instead. That’s one of the major differences to class-based programming: while objects are relatively “rigid” e.g. in Java, where the structure of an object cannot be changed at runtime, in JavaScript, the prototype-based approach links objects of a certain class more loosely together, which allows to change the structure of objects at any time. Also, note how sharing functions through the constructor’s prototype is way more efficient than creating objects that all carry their own functions, even if they are identical. As previously stated, the engine doesn’t know that these functions are meant to be identical, and it has to allocate memory for every function in every object. This is no longer true when sharing functions through a common prototype - the function in question is placed in memory exactly once, and no matter how many myCar objects we create, they don’t carry the function themselves, they only refer to their constructor, in whose prototype the function is found. To give you an idea of what this difference can mean, here is a very simple comparison. The first example creates 1,000,000 objects that all have the function directly attached to them: 1 var C = function() { 2 this.f = function(foo) { 3 console.log(foo); 4 }; 5 }; 6 7 var a = []; 8 for (var i = 0; i < 1000000; i++){ 9 a.push(new C()); 10 } In Google Chrome, this results in a heap snapshot size of 328 MB. Here is the same example, but now the function is shared through the constructor’s prototype: 1 var C = function() {}; 2 3 C.prototype.f = function(foo) { 4 console.log(foo); 5 }; 6 7 var a = []; 8 for (var i = 0; i < 1000000; i++){ 9 a.push(new C()); 10 } This time, the size of the heap snapshot is only 17 MB, i.e., only about 5% of the non-efficient solution. Object-oriented JavaScript 20 Object-orientation, prototyping, and inheritance So far, we haven’t talked about inheritance in JavaScript, so let’s do this now. It’s useful to share behaviour within a certain class of objects, but there are cases where we would like to share behaviour between different, but similar classes of objects. Imagine our virtual world not only had cars, but also bikes. Both drive, but where a car has a horn, a bike has a bell. Being able to drive makes both objects vehicles, but not sharing the honk and ring behaviour distinguishes them. We could illustrate their shared and local behaviour as well as their relationship to each other as follows: Vehicle > drive | /---------/ \--------\ | | Car Bike > honk > ring Designing this relationship in a class-based language like Java is straightforward: We would define a class Vehicle with a method drive, and two classes Car and Bike which both extend the Vehicle class, and implement a honk and a ring method, respectively. This would make the car as well as bike objects inherit the drive behaviour through the inheritance of their classes. How does this work in JavaScript, where we don’t have classes, but prototypes? Let’s look at an example first, and then dissect it. To keep the code short for now, let’s only start with a car that inherits from a vehicle: 1 var Vehicle = function() {}; 2 3 Vehicle.prototype.drive = function() { 4 console.log('vrooom...'); 5 }; 6 7 8 var Car = function() {}; 9 10 Car.prototype = new Vehicle(); Object-oriented JavaScript 21 11 12 Car.prototype.honk = function() { 13 console.log('honk honk'); 14 }; 15 16 17 var myCar = new Car(); 18 19 myCar.honk(); // outputs "honk honk" 20 myCar.drive(); // outputs "vrooom..." In JavaScript, inheritance runs through a chain of prototypes. The prototype of the Car constructor is set to a newly created vehicle object, which establishes the link structure that allows the interpreter to look for methods in “parent” objects. The prototype of the Vehicle constructor has a function drive. Here is what happens when the myCar object is asked to drive(): • The interpreter looks for a drive method within the myCar object, which does not exist • The interpreter then asks the myCar object for its prototype, which is the prototype of its constructor Car • When looking at Car.prototype, the interpreter sees a vehicle object which has a function honk attached, but no drive function • Thus, the interpreter now asks this vehicle object for its prototype, which is the prototype of its constructor Vehicle • When looking at Vehicle.prototype, the interpreter sees an object which has a drive function attached - the interpreter now knows which code implements the myCar.drive() behaviour, and executes it A classless society, revisited We just learned how to emulate the traditional OOP inheritance mechanism. But it’s important to note that in JavaScript, that is only one valid approach to create objects that are related to each other. It was Douglas Crockford who came up with another clever solution, which allows objects to inherit from each other directly. It’s a native part of JavaScript by now - it’s the Object.create() function, and it works like this: Object-oriented JavaScript 22 1 Object.create = function(o) { 2 var F = function() {}; 3 F.prototype = o; 4 return new F(); 5 }; We learned enough now to understand what’s going on. Let’s analyze an example: 1 var vehicle = {}; 2 vehicle.drive = function () { 3 console.log('vrooom...'); 4 }; 5 6 var car = Object.create(vehicle); 7 car.honk = function() { 8 console.log('honk honk'); 9 }; 10 11 var myCar = Object.create(car); 12 13 myCar.honk(); // outputs "honk honk" 14 myCar.drive(); // outputs "vrooom..." While being more concise and expressive, this code achieves exactly the same behaviour, without the need to write dedicated constructors and attaching functions to their prototype. As you can see, Object.create() handles both behind the scenes, on the fly. A temporary constructor is created, its prototype is set to the object that serves as the role model for our new object, and a new object is created from this setup. Conceptually, this is really the same as in the previous example where we defined that Car.prototype shall be a new Vehicle();. But wait! We created the functions drive and honk within our objects, not on their prototypes - that’s memory-inefficient! Well, in this case, it’s actually not. Let’s see why: Object-oriented JavaScript 23 1 var vehicle = {}; 2 vehicle.drive = function () { 3 console.log('vrooom...'); 4 }; 5 6 var car = Object.create(vehicle); 7 car.honk = function() { 8 console.log('honk honk'); 9 }; 10 11 var myVehicle = Object.create(vehicle); 12 var myCar1 = Object.create(car); 13 var myCar2 = Object.create(car); 14 15 myCar1.honk(); // outputs "honk honk" 16 myCar2.honk(); // outputs "honk honk" 17 18 myVehicle.drive(); // outputs "vrooom..." 19 myCar1.drive(); // outputs "vrooom..." 20 myCar2.drive(); // outputs "vrooom..." We have now created a total of 5 objects, but how often do the honk and drive methods exist in memory? Well, how often have they been defined? Just once - and therefore, this solution is basically as efficient as the one where we built the inheritance manually. Let’s look at the numbers: 1 var c = {}; 2 c.f = function(foo) { 3 console.log(foo); 4 }; 5 6 var a = []; 7 for (var i = 0; i < 1000000; i++){ 8 a.push(Object.create(c)); 9 } Turns out, it’s not exactly identical - we end up with a heap snapshot size of 40 MB, thus there seems to be some overhead involved. However, in exchange for much better code, this is probably more than worth it. Summary By now, it’s probably clear what the main difference between classical OOP languages and JavaScript is, conceptually: While classical languages like Java provide one way to manage object creation and Object-oriented JavaScript 24 behaviour sharing (through classes and inheritance), and this way is enforced by the language and “baked in”, JavaScript starts at a slightly lower level and provides building blocks that allow us to create several different mechanisms for this. Whether you decide to use these building blocks to recreate the traditional class-based pattern, let your objects inherit from each other directly, with the concept of classes getting in the way, or if you don’t use the object-oriented paradigm at all and just solve the problem at hand with pure functional code: JavaScript gives you the freedom to choose the best methodology for any situation. Synchronous and Asynchronous operations explained Visualizing the Node.js execution model For the chapters that follow it’s important to fully understand what it means, conceptually, that a Node.js application has synchronous and asynchronous operations, and how both operations interact with each other. Let’s try to build this understanding step by step. The first concept that we need to understand is that of the Node.js event loop. The event loop is the execution model of a running Node.js application. We can visualize this model as a row of loops: +----> -----+ +----> -----+ +----> -----+ +----> -----+ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | +-----------+ +-----------+ +-----------+ +-----------+ I’ve drawn boxes because circles look really clumsy in ascii art. So, these here look like rectangles, but please imagine them as circles - circles with an arrow, which means that one circle represents one iteration through the event loop. Another visualization could be the following pseudo-code: while (I still have stuff to do){ do stuff; } Conceptually, at the very core of it, it’s really that simple: Node.js starts, loads our application, and then loops until there is nothing left to do - at which point our application terminates. What kind of stuff is happening inside one loop iteration? Let’s first look at a very simple example, a Hello World application like this: Synchronous and Asynchronous operations explained 26 console.log('Hello'); console.log('World'); This is the visualization of the execution of this application in our ascii art: +----> ---------+ | | | Write | | 'Hello' | | to the screen | | | | Write | | 'World' | | to the screen | | | +---------------+ Yep, that’s it: Just one iteration, and then, exit the application. The things we asked our application to do - writing text to the screen, and then writing another text to the screen, using console.log - are synchronous operations. They both happen within the same (and in this case, only) iteration through the event loop. Let’s look at the model when we bring asynchronous operations into the game, like this: console.log('Hello'); setTimeout(function() { console.log('World'); }, 1000); This still prints Hello and then World to the screen, but the second text is printed with a delay of 1000 ms. setTimeout, you may have guessed it, is an asynchronous operation. We pass the code to be executed in the body of an anonymous function - the so-called callback function. Why do we do so? The visualization helps to understand why: Synchronous and Asynchronous operations explained 27 +----> ------------+ +----> ----+ +----> -----+ +----> ------+ | | | | | | | | | Write | | | | | | Write | | 'Hello' to | | | | | | 'World' to | | the screen | | | | | | the screen | | | | | | | | | | Start setTimeout | | | | | | | | operation | | | | | | | | | | | | | | | +-----+------------+ +----------+ +-----------+ ^------------+ | | | setTimeout(callback, 1000) | callback() | | +--------------------------------------------+ As time goes by... This, again at the very core of it, is what calling an asynchronous function does: It starts an operation outside the event loop. Conceptually, Node.js starts the asynchronous operation and makes a mental note that when this operations triggers an event, the anonymous function that was passed to the operation needs to be called. Hence, the event loop: as long as asynchronous operations are ongoing, the Node.js process loops, waiting for events from those operations. As soon as no more asynchronous operations are ongoing, the looping stops and our application terminates. Note that the visualization isn’t detailed enough to show that Node.js checks for outside events between loop iterations. Just to be clear: callback functions don’t need to be anonymous inline functions: var write_world = function() { console.log('World'); }; console.log('Hello'); setTimeout(write_world, 1000); results in: Synchronous and Asynchronous operations explained 28 +----> ------------+ +----> ----+ +----> ----+ +----> ------+ | | | | | | | | | Write | | | | | | Write | | 'Hello' to | | | | | | 'World' to | | the screen | | | | | | the screen | | | | | | | | | | Start setTimeout | | | | | | | | operation | | | | | | | | | | | | | | | +-----+------------+ +----------+ +----------+ ^------------+ | | | setTimeout(write_world, 1000) | write_world() | | +-------------------------------------------+ As time goes by... It’s just that more often than not, declaring a named function and passing it as the callback isn’t worth the hassle, because very often we need the functionality described in the function only once. Let’s look at a slightly more interesting asynchronous operation, the fs.stat call - it starts an IO operation that looks at the file system information for a given path, that is, stuff like the inode number and the size etc.: var fs = require('fs'); fs.stat('/etc/passwd', function(err, stats) { console.dir(stats); }); The visualization: +----> -----+ +----> -----+ +----> -----+ +----> -----+ | | | | | | | | | Start | | | | | | Write | | fs.stat | | | | | | stats | | operation | | | | | | object to | | | | | | | | screen | +-----+-----+ +-----------+ +-----------+ +-----^-----+ | | | fs.stat('/etc/passwd', callback) | callback() | | +--------------------------------------------+ Ongoing asynchronous file system I/O operation Synchronous and Asynchronous operations explained 29 Really not that different - instead of Node.js merely counting milliseconds in the background, it starts an IO operation; IO operations are expensive, and several loop iterations go by where nothing happens. Then, Node.js has finished the background IO operation, and triggers the callback we passed in order to jump back - right into the current loop iteration. We then print - very synchronously - the stats object to the screen. Another classical example: When we start an HTTP server, we create a background operation which continuously waits for HTTP requests from clients: var http = require('http'); http.createServer(function(request, response) { response.writeHead(200,{'Content-Type':'text/html'}); response.write('Hello World'); response.end(); }).listen(8080); Whenever this event occurs, the passed callback is triggered: +----> -----+ +----> -----+ +----> -----+ +----> -----+ | | | | | | | | | Start | | | | Send | | Send | | HTTP | | | | response | | response | | server | | | | to client | | to client | | | | | | | | | +-----+-----+ +-----------+ ^-----------+ ^-----------+ | | | | http.createServer() | callback() | callback() | | | +-------------------^---+----------^---+ | | | | | | a client a client requests requests the server the server Blocking and non-blocking operations From the understanding of this conceptual model, we can get to understanding the difference between blocking and non-blocking operations. The key to understanding these is the realization that every synchronous operation results in a blocking operation. That’s right, even an innocent Synchronous and Asynchronous operations explained 30 console.log('Hello World'); results in a blocking operation - while the Node.js process writes the text to the screen, this is the only thing that it does. There is only one single piece of JavaScript code that can be executed within the event loop at any given time. The reason this doesn’t result in a problem in the case of a console.log() is that it is an extremely cheap operation. In comparison, even the most simple IO operations are way more expensive. Whenever the network or harddrives (yes, including SSDs) are involved, expect things to be incredibly much slower compared to operations where only the CPU and RAM are involved, like var a = 2 * 21. Just how much slower? The following table gives you a good idea - it shows actual times for different kind of computer operations, and shows how they relate to each other compared to how time spans that human beings experience relate to each other: 1 CPU cycle 0.3 ns 1 s Level 1 cache access 0.9 ns 3 s Level 2 cache access 2.8 ns 9 s Level 3 cache access 12.9 ns 43 s Main memory access 120 ns 6 min Solid-state disk I/O 50-150 μs 2-6 days Rotational disk I/O 1-10 ms 1-12 months Internet: SF to NYC 40 ms 4 years Internet: SF to UK 81 ms 8 years Internet: SF to Australia 183 ms 19 years OS virtualization reboot 4 s 423 years SCSI command time-out 30 s 3000 years Hardware virtualization reboot 40 s 4000 years Physical system reboot 5 m 32 millenia So, the difference between setting a variable in your code and reading even a tiny file from disk is like the difference between preparing a sandwich and going on vacation for a week. You can prepare a lot of sandwichs during one week of vacation. And that’s the sole reason why all those Node.js functions that result in IO operations also happen to work asynchronously: it’s because the event loop needs being kept free of any long- running operations during which the Node.js process would practically stall, which would result in unresponsive applications. Just to be clear: we can easily stall the event loop even if no IO operations are involved and we only use cheap synchronous functions - if only there are enough of them within one loop iteration. Take the following code for example: Synchronous and Asynchronous operations explained 31 var http = require('http'); http.createServer(function(request, response) { console.log('Handling HTTP request'); response.writeHead(200,{'Content-Type':'text/html'}); response.write('Hello World'); response.end(); }).listen(8080); var a; for (var i=0; i < 10000000000; i += 1){ a = i; } console.log('For loop has finished'); It’s our minimalistic web server again, plus a for loop with 10,000,000,000 iterations. Our event loop visualization basically looks the same: +----> -----+ +----> -----+ +----> -----+ +----> -----+ | | | | | | | | | Start | | | | Send | | Send | | HTTP | | | | response | | response | | server | | | | to client | | to client | | | | | | | | | +-----+-----+ +-----------+ ^-----------+ ^-----------+ | | | | http.createServer() | callback() | callback() | | | +-------------------^---+----------^---+ | | | | | | a client a client requests requests the server the server But here is the problem: The HTTP server will start listening in the background as an asynchronous operation, but we will stay within the first loop operation as long as the for loop is running. And although we only have a very cheap operation happening within the for loop, it happens 10,000,000,000 times, and on my machine, this takes around 20 seconds. Synchronous and Asynchronous operations explained 32 When you start the server application and then open your browser at http://localhost:8080/, you won’t get an answer right away. External events, like our HTTP request, are only handled between one loop iteration and the next; however, the first loop iteration takes 20 seconds because of our for loop, and only then Node.js switches to the next iteration and has a chance to handle our request by calling the HTTP server callback. As you will see, the application will print For loop has finished, and right after that, it will answer the HTTP request and print Handling HTTP request. This demonstrates how external events from asynchronous operations are handled at the beginning of a new event loop iteration. You’ve probably heard time and again how Node.js isn’t suited for writing applications with CPU intensive tasks - as we can see, the reason for this is the event loop model. From this, we can distill the two most important rules for writing reponsive Node.js applications: • Handle IO-intensive operations through asynchronous operations • Keep your own code (that is, everything that happens synchronously within event loop iterations) as lean as possible This leaves the question: what are sensible solutions if you have to do expensive CPU- bound operations within your JavaScript code? As we will learn in later chapters, we can mitigate the problem that Node.js itself simply isn’t particularly suited for these kinds of operations. Using and creating Event Emitters Introduction By now, you are probably more than familiar with this: someFunction(function(err) { if (!err) { console.log('Hey, looks like someFunction has finished and called me.'); } else { console.log('Oh no, something terrible happened!'); } }); We call a function, someFunction in this case, which does something asynchronously in the background, and calls the anonymous function we passed in (the callback) once it has finished, passing an Error object if something went wrong, or null if all went fine. That’s the standard callback pattern, and you will encounter it regularly when writing Node.js code. For simple use cases, where we call a function that finishes its task some time in the future, either successfully or with an error, this pattern is just fine. For more complex use cases, this pattern does not scale well. There are cases where we call a function which results in multiple events happening over time, and also, different types of events might happen. One example is reading the contents of a file through a ReadStream. The details of handling files are discussed in a later chapter, but we will use this example to illustrate event emitters. This snippet demonstrates how to read data from a large file using a ReadStream: 'use strict'; var fs = require('fs'); fs.createReadStream('/path/to/large/file'); When reading data from a file, two different things can happen: We either receive content, or we reach the end of the file. Both cases could be handled using the callback pattern - for example, by using one callback with two parameters, one with the data and another one that is false as long as Using and creating Event Emitters 34 the end of the file has not been reached, and true once it has been reached. Or we could provide two separate callback functions, one that is called when content is retrieved and one that is called when the end of the file has been reached. But there is a more elegant way. Instead of working with classical callbacks, createReadStream allows us to use an Event Emitter. That’s a special object which can be used to attach callback functions to different events. Using it looks like this: 1 'use strict'; 2 3 var fs = require('fs'); 4 5 var stream = fs.createReadStream('/path/to/large/file'); 6 7 stream.on('data', function(data) { 8 console.log('Received data: ' + data); 9 }); 10 11 stream.on('end', function() { 12 console.log('End of file has been reached'); 13 }); Here is what happens in detail: • On line 5, we create a read stream that will start to retrieve the contents of file /path/to/large/- file. The call to fs.createReadStream does not take a function argument to use it as a callback. Instead, it returns an object, which we assign as the value of the variable stream. • On line 7, we attach a callback to one type of events the ReadStream emits: data events • On line 11, we attach another callback to another type of event the ReadStream emits: the end event The object that is returned by fs.createReadStream is an Event Emitter. These objects allow us to attach different callbacks to different events while keeping our code readable and sane. A ReadStream retrieves the contents of a file in chunks, which is more efficient than to load the whole data of potentially huge files into memory at once in one long, blocking operation. Because of this, the data event will be emitted multiple times, depending on the size of the file. The callback that is attached to this event will therefore be called multiple times. When all content has been retrieved, the end event is fired once, and no other events will be fired from then on. The end event callback is therefore the right place to do whatever we want to do after we retrieved the complete file content. In practice, this would look like this: Using and creating Event Emitters 35 1 'use strict'; 2 3 var fs = require('fs'); 4 5 var stream = fs.createReadStream('/path/to/large/file'); 6 7 var content = ''; 8 9 stream.on('data', function(data) { 10 content = content + data; 11 }); 12 13 stream.on('end', function() { 14 console.log('File content has been retrieved: ' + content); 15 }); It doesn’t make too much sense to efficiently read a large file’s content in chunks, only to assign the whole data to a variable and therefore using the memory anyways. In a real application, we would read a file in chunks to, for example, send every chunk to a web client that is downloading the file via HTTP. We will talk about this in more detail in a later chapter. The Event Emitter pattern itself is simple and clearly defined: a function returns an event emitter object, and using this object’s on method, callbacks can be attached to events. However, there is no strict rule regarding the events themselves: an event type, like data or end, is just a string, and the author of an event emitter can define any name she wants. Also, it’s not defined what arguments are passed to the callbacks that are triggered through an event - the author of the event emitter should define this through some kind of documentation. There is one recurring pattern, at least for the internal Node.js modules, and that is the error event: Most event emitters emit an event called “error” whenever an error occurs, and if we don’t listen to this event, the event emitter will raise an exception. You can easily test this by running the above code: as long as you don’t happen to have a file at /path/to/large/file, Node.js will bail out with this message: events.js:72 throw er; // Unhandled 'error' event ^ Error: ENOENT, open '/path/to/large/file' But if you attach a callback to the error event, you can handle the error yourself: Using and creating Event Emitters 36 1 'use strict'; 2 3 var fs = require('fs'); 4 5 var stream = fs.createReadStream('/path/to/large/file'); 6 7 var content = ''; 8 9 stream.on('error', function(err) { 10 console.log('Sad panda: ' + err); 11 }); 12 13 stream.on('data', function(data) { 14 content = content + data; 15 }); 16 17 stream.on('end', function() { 18 console.log('File content has been retrieved: ' + content); 19 }); Instead of using on, we can also attach a callback to an event using once. Callbacks that are attached this way will be called the first time that the event occurs, but will then be removed from the list of event listeners and not be called again: stream.once('data', function(data) { console.log('I have received the first chunk of data'); }); Also, it’s possible to detach an attached callback manually. This only works with named callback functions: var callback = function(data) { console.log('I have received a chunk of data: ' + data); } stream.on('data', callback); stream.removeListener('data', callback); And last but not least, you can remove all attached listeners for a certain event: Using and creating Event Emitters 37 stream.removeAllListeners('data'); Creating your own Event Emitter object We can create event emitters ourselves. This is even supported by Node.js by inheriting from the built-in events.EventEmitter class. But let’s first implement a simple event emitter from scratch, because this explains the pattern in all its details. For this, we are going to create a module whose purpose is to regularly watch for changes in the size of a file. Once implemented, it can be used like this: 'use strict'; watcher = new FilesizeWatcher('/path/to/file'); watcher.on('error', function(err) { console.log('Error watching file:', err); }); watcher.on('grew', function(gain) { console.log('File grew by', gain, 'bytes'); }); watcher.on('shrank', function(loss) { console.log('File shrank by', loss, 'bytes'); }); watcher.stop(); As you can see, the module consists of a class FilesizeWatcher which can be instantiated with a file path and returns an event emitter. We can listen to three different events from this emitter: error, grew, and shrank. Let’s start by writing a spec that describes how we expect to use our event emitter. To do so, create a new project directory, and add the following package.json: { "devDependencies":{ "jasmine-node":"" } } Afterwards, run npm install. Now create a file FilesizeWatcherSpec.js, with the following content: Using and creating Event Emitters 38 1 'use strict'; 2 3 var FilesizeWatcher = require('./FilesizeWatcher'); 4 var exec = require('child_process').exec; 5 6 describe('FilesizeWatcher', function() { 7 8 var watcher; 9 10 afterEach(function() { 11 watcher.stop(); 12 }); 13 14 it('should fire a "grew" event when the file grew in size', function(done) { 15 16 var path = '/var/tmp/filesizewatcher.test'; 17 exec('rm -f ' + path + ' ; touch ' + path, function() { 18 watcher = new FilesizeWatcher(path); 19 20 watcher.on('grew', function(gain) { 21 expect(gain).toBe(5); 22 done(); 23 }); 24 25 exec('echo "test" > ' + path, function() {}); 26 27 }); 28 29 }); 30 31 it('should fire a "shrank" event when the file grew in size', function(done) { 32 33 var path = '/var/tmp/filesizewatcher.test'; 34 exec('rm -f ' + path + ' ; echo "test" > ' + path, function() { 35 watcher = new FilesizeWatcher(path); 36 37 watcher.on('shrank', function(loss) { 38 expect(loss).toBe(3); 39 done(); 40 }); 41 42 exec('echo "a" > ' + path, function() {}); Using and creating Event Emitters 39 43 44 }); 45 46 }); 47 48 it('should fire "error" if path does not start with a slash', function(done) { 49 50 var path = 'var/tmp/filesizewatcher.test'; 51 watcher = new FilesizeWatcher(path); 52 53 watcher.on('error', function(err) { 54 expect(err).toBe('Path does not start with a slash'); 55 done(); 56 }); 57 58 }); 59 60 }); Because this is just an example application, we will not create a spec and a src directory, but instead just put both the specification file and the implementation file in the top folder of our project. Before we look at the specification itself in detail, let’s discuss the done() call we see in each of the it blocks. The done function is a callback that is passed to the function parameter of an it block by Jasmine. This pattern is used when testing asynchronous operations. Our emitter emits events asyn- chronously, and Jasmine cannot know by itself when events will fire. It needs our help by being told “now the asynchronous operation I expected to occur did actually occur” - and this is done by triggering the callback. Now to the specification itself. The first expectation is that when we write “test” into our testfile, the grew event is fired, telling us that the file gained 5 bytes in size. Note how we use the exec function from the child_process module to manipulate our test file through shell commands within the specification. Next, we specify the behaviour that is expected if the monitored test file shrinks in size: the shrank event must fire and report how many bytes the file lost. At last, we specify that if we ask the watcher to monitor a file path that doesn’t start with a slash, an error event must be emitted. Using and creating Event Emitters 40 I’m creating a very simplified version of a file size watcher here for the sake of brevity - for a realworld implementation, more sophisticated checks would make sense. We will create two different implementations which both fulfill this specification. First, we will create a version of the file size watcher where we manage the event listener and event emitting logic completely ourselves. This way, we experience first hand how the event emitter pattern works. Afterwards, we will implement a second version where we make use of existing Node.js functionality in order to implement the event emitter pattern without the need to reinvent the wheel. The following shows a possible implementation of the first version, where we take care of the event listener callbacks ourselves: 1 'use strict'; 2 3 var fs = require('fs'); 4 5 var FilesizeWatcher = function(path) { 6 var self = this; 7 8 self.callbacks = {}; 9 10 if (/^\//.test(path) === false){ 11 self.callbacks['error']('Path does not start with a slash'); 12 return; 13 } 14 15 fs.stat(path, function(err, stats) { 16 self.lastfilesize = stats.size; 17 }); 18 19 self.interval = setInterval( 20 function() { 21 fs.stat(path, function(err, stats) { 22 if (stats.size > self.lastfilesize) { 23 self.callbacks['grew'](stats.size - self.lastfilesize); 24 self.lastfilesize = stats.size; 25 } 26 if (stats.size < self.lastfilesize) { 27 self.callbacks['shrank'](self.lastfilesize - stats.size); 28 self.lastfilesize = stats.size; 29 } Using and creating Event Emitters 41 30 }, 1000); 31 }); 32 }; 33 34 FilesizeWatcher.prototype.on = function(eventType, callback) { 35 this.callbacks[eventType] = callback; 36 }; 37 38 FilesizeWatcher.prototype.stop = function() { 39 clearInterval(this.interval); 40 }; 41 42 module.exports = FilesizeWatcher; Let’s discuss this code. On line 3, we load the fs module - we need its stat function to asynchronously retrieve file information. On line 5 we start to build a constructor function for FilesizeWatcher objects. They are created by passing a path to watch as a parameter. On line 6, we assign the object instance variable to a local self variable - this way we can access our instantiated object within callback functions, where this would point to another object. We then create the self.callbacks object - we are going to use this as an associative array where we will store the callback to each event. Next, on line 10, we check if the given path starts with a slash using a regular expression - if it doesn’t, we trigger the callback associated with the error event. If the check succeeds, we start an initial stat operation in order to store the file size of the given path - we need this base value in order to recognize future changes in file size. The actual watch logic starts on line 19. We set up a 1-second interval where we call stat on every interval iteration and compare the current file size with the last known file size. Line 22 handles the case where the file grew in size, calling the event handler callback associated with the grew event; line 26 handles the opposite case. In both cases, the new file size is saved. Event handlers can be registered using the FilesizeWatcher.on method which is defined on line 34. In our implementation, all it does is to store the callback under the event name in our callbacks object. Finally, line 38 defines the stop method which cancels the interval we set up in the constructor function. Let’s see if this implementation works by running ./node_modules/jasmine-node/bin/jasmine-node ./FilesizeWatcherSpec.js: Using and creating Event Emitters 42 ..F Failures: 1) FilesizeWatcher should fire "error" if the path does not start with a slash Message: TypeError: Object # has no method 'error' Stacktrace: TypeError: Object # has no method 'error' at new FilesizeWatcher (FilesizeWatcher.js:11:28) at null. (FilesizeWatcherSpec.js:51:15) at null. (...mine-node/lib/jasmine-node/async-callback.js:45:37) Finished in 0.087 seconds 3 tests, 3 assertions, 1 failure, 0 skipped Well… nearly. The core functionality works: the grew and shrank events are fired as expected. But the file path check makes problems. According to the message, the problem arises on line 11: self.callbacks['error']('Path does not start with a slash'); The error message says that the self.callbacks object doesn’t have a method called error. But in our specification, we defined a callback for this event, just as we did for the other events, on line 53 of FilesizeWatcherSpec.js: watcher.on('error', function(err) { The problem here is quite subtle and can only be understood if one remembers how the execution model of Node.js works. Let’s recall the chapter on synchronous and asynchronous operations. Our specification code looks like this: watcher = new FilesizeWatcher(path); watcher.on('error', function(err) { expect(err).toBe('Path does not start with a slash'); done(); }); According to the visualization technique we used to explain the inner workings of the event loop model, this is what actually happens: Using and creating Event Emitters 43 +----> ------+ +----> -----+ +----> -----+ +----> -----+ | | | | | | | | | Create | | | | | | | | watcher | | | | | | | | object | | | | | | | | | | | | | | | | Check path | | | | | | | | | | | | | | | | Trigger | | | | | | | | error | | | | | | | | callback | | | | | | | | | | | | | | | | Attach | | | | | | | | callback | | | | | | | | to error | | | | | | | | event | | | | | | | | | | | | | | | +------------+ +-----------+ +-----------+ +-----------+ Now it becomes clear that the order in which things are happening is simply wrong: We first call the error callback, and then we attach it. This can’t work. When creating the watcher object, we call the Filesizewatcher constructor function. Up to the point where it checks the path and calls the error event handler callback, everything happens synchronously: var FilesizeWatcher = function(path) { var self = this; self.callbacks = {}; if (/^\//.test(path) === false){ self.callbacks['error']('Path does not start with a slash'); return; } The constructor function returns and Node.js execution continues with the main code, all the while we are still in the first event loop iteration. Only now - afterwards - we attach the event handler callback for the error event; but the constructor already tried to call it! Why doesn’t this problem arise with the grew and shrank events? Here is the visualization: Using and creating Event Emitters 44 +----> ------+ +----> -----+ +----> -----+ +----> -----+ | | | | | | | | | Create | | | | | | | | watcher | | | | | | | | object | | | | | | | | | | | | | | | | Start | | | | | | | | check | | | | | | | | interval | | | | | | | | | | | | | | | | Attach | | | | | | | | callback | | | | | | | | to grew | | | | | | | | event | | | | | | | | | | | | | | | +-----+------+ +-----------+ +-----^-----+ +-----------+ | | | setInterval() | callback() | | +--------------+ | | | | fs.stat() | | | +---------------+ This shows how our main code in the first event loop starts the asynchronous setInterval operation which in turn starts an asynchronous stat operation. Only several event loop iterations later, when the callback for the grew event has long been attached, is this callback triggered. No problem arises. How can we solve the problem for our error event? The solution is to put the triggering of the not- yet-attached callback into the future, into the next event loop iteration; this way, the callback will be triggered in the second event loop iteration after it has been attached in the first event loop iteration. Moving code execution into the next event loop iteration is simple thanks to the process.nextTick method. Here is how the relevant part in file FilesizeWatcher.js needs to be rewritten: if (/^\//.test(path) === false){ process.nextTick(function() { self.callbacks['error']('Path does not start with a slash'); }); return; } With this, we put the callback triggering for the error event listener into a callback function that will be executed by process.nextTick on the next iteration of the event loop: Using and creating Event Emitters 45 +----> ------+ +----> -----+ +----> -----+ +----> -----+ | | | | | | | | | Create | | | | | | | | watcher | | | | | | | | object | | | | | | | | | | | | | | | | Check path | | | | | | | | | | | | | | | | Put | | | | | | | | callback | | | | | | | | into next | | | | | | | | iteration | | | | | | | | | | | | | | | | Attach | | | | | | | | callback | | | | | | | | to error | | | | | | | | event | | | | | | | | | | | | | | | +-----+------+ +-----^-----+ +-----------+ +-----------+ | | | process.nextTick() | callback() | | +--------------------+ Thus we ensure that it is called after it has been defined. As mentioned, Node.js supports us in creating event emitters, sparing us from writing the callback handling code for our own emitters. We can use this support by inheriting from an existing class EventEmitter in module events. Because the expected behaviour of our MyEmitter doesn’t change from an outside perspective, we will leave the spec unchanged, and only change our implementation from using a self-made solution to utilizing the existing Node.js class: 1 'use strict'; 2 3 var fs = require('fs'); 4 var util = require('util'); 5 var EventEmitter = require('events').EventEmitter; 6 7 var FilesizeWatcher = function(path) { 8 var self = this; 9 Using and creating Event Emitters 46 10 if (/^\//.test(path) === false){ 11 process.nextTick(function() { 12 self.emit('error','Path does not start with a slash'); 13 }); 14 return; 15 } 16 17 fs.stat(path, function(err, stats) { 18 self.lastfilesize = stats.size; 19 }); 20 21 self.interval = setInterval( 22 function() { 23 fs.stat(path, function(err, stats) { 24 if (stats.size > self.lastfilesize) { 25 self.emit('grew', stats.size - self.lastfilesize); 26 self.lastfilesize = stats.size; 27 } 28 if (stats.size < self.lastfilesize) { 29 self.emit('shrank', self.lastfilesize - stats.size); 30 self.lastfilesize = stats.size; 31 } 32 }, 1000); 33 }); 34 }; 35 36 util.inherits(FilesizeWatcher, EventEmitter); 37 38 FilesizeWatcher.prototype.stop = function() { 39 clearInterval(this.interval); 40 }; 41 42 module.exports = FilesizeWatcher; The interesting part here is line 36. We use a helper function that ships with Node.js, util.inherits. We use it to make our FilesizeWatcher class a descendant of events.EventEmitter, which also ships with Node.js. This class implements all the logic that is needed to act as an event emitter, and our FilesizeWatcher class inherits this logic. This way, our own code now concentrates on implementing its specific logic, without having to care about the details of registering callbacks and checking if these can be called etc. According to our tests, FilesizeWatcher still behaves as it did before: Using and creating Event Emitters 47 ... Finished in 0.052 seconds 3 tests, 3 assertions, 0 failures, 0 skipped Instead of using util.inherits, we can also create the inheritance by placing a new events.EventEmitter object on the prototype of FilesizeWatcher: FilesizeWatcher.prototype = new EventEmitter(); Although util.inherits adds some extra properties to our constructor, the result is the same, behaviour-wise. Give it a try if you like and see if the tests still pass. Summary The asynchronous development style using callbacks that is a recurring pattern in JavaScript and Node.js applications can at times result in a code structure that is hard to understand. Applying the event emitter pattern comes in handy in situations where more than one callback event must be handled, resulting in a cleaner code structure that is easier to follow and reason about. The pattern is so useful that we should use it for our own objects and functions where it makes sense, something that’s made easy by utilizing the existing util.inherits class. Optimizing code performance and control flow management using the async library Writing or using functions or methods in your code that are executed one after the other gets you a long way in your applications. Sometimes, those functions are simple synchronous steps: console.log('Starting calculation...'); var result = 5 + 4; console.log('The result is', result); Often, callbacks are used when operations are executed in the background and jump back into our code’s control flow asynchronously at a later point in time: console.log('Starting calculation...'); startExpensiveCalculation(5, 4, function(err, result) { if (!err) { console.log('The result is', result); } }); If those asynchronous background operations bring forth a more complex callback behaviour, they might be implemented as an event emitter: console.log('Starting calculation...'); calculation = Calculator.start(5, 4); calculation.on('error', function(err) { console.log('An error occured:', err); }); calculation.on('interim result', function(result) { console.log('Received interim result:', result); Optimizing code performance and control flow management using the async library 49 }); calculation.on('done', function(result) { console.log('Received final result:', result); }); Handling expensive operations asynchronously in the background, especially if they are IO-bound, is an important key to making Node.js applications perform efficiently - reading a large file or writing a lot of records to a database will always be a costly procedure, but handling it asynchronously at least ensures that the other parts of our application won’t be blocked while that procedure is going on. Nevertheless, there is often potential for optimization within our own code and its control flow. Executing expensive asynchronous background tasks in parallel Let’s consider an example where our application queries several different remote web services, presenting the retrieved data on the console. We are not going to query real remote web services, instead we will write a very simple Node.js HTTP server that will serve as a dummy web service. Our web server doesn’t really do anything significant, and therefore we will make it react to requests a bit slower than neccessary, in order to simulate a real web service that has a certain workload - as you will see, this makes it easier for us to show the performance gain in our own code optimizations. Please create a new project folder and create a file server.js with the following content: 1 'use strict'; 2 3 var http = require('http'); 4 var url = require('url'); 5 var querystring = require('querystring'); 6 7 http.createServer(function(request, response) { 8 9 var pathname = url.parse(request.url).pathname; 10 var query = url.parse(request.url).query; 11 var id = querystring.parse(query)['id']; 12 13 var result = { 14 'pathname': pathname, Optimizing code performance and control flow management using the async library 50 15 'id': id, 16 'value': Math.floor(Math.random() * 100) 17 }; 18 19 setTimeout(function() { 20 response.writeHead(200,{"Content-Type":"application/json"}); 21 response.end(JSON.stringify(result)); 22 }, 2000 + Math.floor(Math.random() * 1000)); 23 24 }).listen( 25 8080, 26 function() { 27 console.log('Echo Server listening on port 8080'); 28 } 29 ); This gives us a very simple “echo” server - if we request the URL http://localhost:8080/getUser?id=4, we receive {"pathname":"/getUser","id":"4","value":67} as the response. This is good enough to give us the simulation of a remote webservice API to play around with. But alas, it’s a slow webservice! Someone didn’t do his optimization homework, and we now have to deal with an API where every response takes between 2 to 3 seconds (this is simulated with the setTimeout construct on lines 19-22). This allows to show how different request patterns will result in different runtime characteristics. We will now write a Node.js client for this webserver. This client will make two consecutive requests to the server, and print the output for both requests on the command line: 1 'use strict'; 2 3 var request = require('request'); 4 5 request.get( 6 'http://localhost:8080/getUserName?id=1234', 7 function(err, res, body) { 8 var result = JSON.parse(body); 9 var name = result.value; 10 11 request.get( 12 'http://localhost:8080/getUserStatus?id=1234', 13 function(err, res, body) { 14 var result = JSON.parse(body); 15 var status = result.value; Optimizing code performance and control flow management using the async library 51 16 17 console.log('The status of user', name, 'is', status); 18 }); 19 20 }); Save this in a file called client.js. Also, you need to install the request module via npm install request. This is probably the most straight-forward solution. We start the first request, wait for it to finish, print the result, then start the second request, wait for it to finish, and print that result, too. How long does that take? Let’s see: $ time node client.js The status of user 62 is 68 real 0m5.810s user 0m0.172s sys 0m0.033s Don’t forget to start the server via node server.js beforehand! Not surprisingly, it takes around 5-6 seconds, because we only start the second request after the first request has been completed, and each request takes around 2-3 seconds. We can’t do anything about the terribly slow remote webservice, but our own code isn’t exactly optimal either. Our two requests don’t inherently depend on each other, and yet, we are executing them serially. Of course starting these requests in parallel is simple, because both are asynchronous operations: Optimizing code performance and control flow management using the async library 52 1 'use strict'; 2 3 var request = require('request'); 4 5 var name, status; 6 7 request.get( 8 'http://localhost:8080/getUserName?id=1234', 9 function(err, res, body) { 10 var result = JSON.parse(body); 11 name = result.value; 12 }); 13 14 request.get( 15 'http://localhost:8080/getUserStatus?id=1234', 16 function(err, res, body) { 17 var result = JSON.parse(body); 18 status = result.value; 19 }); 20 21 console.log('The status of user', name, 'is', status); No, wait, sorry! That’s not going to work - console.log will execute within the first event loop iteration while the request callbacks are triggered in later iterations. Mh, how about… 1 'use strict'; 2 3 var request = require('request'); 4 5 var name, status; 6 7 request.get( 8 'http://localhost:8080/getUserName?id=1234', 9 function(err, res, body) { 10 var result = JSON.parse(body); 11 name = result.value; 12 }); 13 14 request.get( 15 'http://localhost:8080/getUserStatus?id=1234', 16 function(err, res, body) { 17 var result = JSON.parse(body); Optimizing code performance and control flow management using the async library 53 18 status = result.value; 19 20 console.log('The status of user', name, 'is', status); 21 }); No, that’s not good either: We start both request in parallel, but we have no guarantee that they will finish at the same time. We risk printing 1 The status of user undefined is 75 if the second request finishes earlier than the first. Well, looks like we need some additional code to synchronize our finished calls. How about this: 1 'use strict'; 2 3 var request = require('request'); 4 5 var name, status; 6 var firstHasFinished, secondHasFinished = false; 7 8 request.get( 9 'http://localhost:8080/getUserName?id=1234', 10 function(err, res, body) { 11 var result = JSON.parse(body); 12 name = result.value; 13 markFinished('first'); 14 }); 15 16 request.get( 17 'http://localhost:8080/getUserStatus?id=1234', 18 function(err, res, body) { 19 var result = JSON.parse(body); 20 status = result.value; 21 markFinished('second'); 22 }); 23 24 function markFinished(step) { 25 if (step == 'first'){ 26 firstHasFinished = true; 27 } 28 29 if (step == 'second'){ Optimizing code performance and control flow management using the async library 54 30 secondHasFinished = true; 31 } 32 33 if (firstHasFinished && secondHasFinished) { 34 console.log('The status of user', name, 'is', status); 35 } 36 } Seriously now - that’s not even funny! What if we need to synchronize dozens or hundreds of operations? We could use an array where we store the state of each operation… no, this whole thing doesn’t feel right. async to the rescue, I say! async is a clever little module that makes managing complex control flows in our code a breeze. After installing the module via npm install async, we can write our client like this: 1 'use strict'; 2 3 var request = require('request'); 4 var async = require('async'); 5 6 var name, status; 7 8 var getUsername = function(callback) { 9 request.get( 10 'http://localhost:8080/getUserName?id=1234', 11 function(err, res, body) { 12 var result = JSON.parse(body); 13 callback(err, result.value); 14 }); 15 }; 16 17 var getUserStatus = function(callback) { 18 request.get( 19 'http://localhost:8080/getUserStatus?id=1234', 20 function (err, res, body) { 21 var result = JSON.parse(body); 22 callback(err, result.value); 23 }); 24 }; 25 26 async.parallel([getUsername, getUserStatus], function(err, results) { Optimizing code performance and control flow management using the async library 55 27 console.log('The status of user', results[0], 'is', results[1]); 28 }); Let’s analyze what we are doing here. On line 4, we load the async library. We then wrap our requests into named functions. These functions will be called with a callback parameter. Inside our functions, we trigger this callback when our operation has finished - in this case, when the requests have been answered. We call the callbacks with two parameters: an error object (which is null if no errors occured), and the result value. The orchestration happens on lines 26-28. We use the parallel method of the async object and pass an array of all the functions we want to run in parallel. Additionally, we pass a callback function which expect two parameters, err and results. async.parallel will trigger this callback as soon as the slowest of the parallel operations has finished (and called its callback), or as soon as one of the operations triggers its callback with an error. Let’s see what this does to the total runtime of our script: $ time node client.js The status of user 95 is 54 real 0m3.176s user 0m0.240s sys 0m0.044s As one would expect, the total runtime of our own code matches the runtime of one request because both requests are started in parallel and will finish roughly at the same time. Optimizing code structure with async async offers several other mechanisms for managing the control flow of our code. These are interesting even if our concern isn’t performance optimization. Let’s investigate them. For these cases, let’s remove the artificial slowness from the API server by removing the setTimeout operation on line 19 and 22, making the server respond immediately: Optimizing code performance and control flow management using the async library 56 1 'use strict'; 2 3 var http = require('http'); 4 var url = require('url'); 5 var querystring = require('querystring'); 6 7 http.createServer(function(request, response) { 8 9 var pathname = url.parse(request.url).pathname; 10 var query = url.parse(request.url).query; 11 var id = querystring.parse(query)['id']; 12 13 var result = { 14 'pathname': pathname, 15 'id': id, 16 'value': Math.floor(Math.random() * 100) 17 }; 18 19 response.writeHead(200,{"Content-Type":"application/json"}); 20 response.end(JSON.stringify(result)); 21 22 }).listen( 23 8080, 24 function() { 25 console.log('Echo Server listening on port 8080'); 26 } 27 ); Sometimes we want to run operations in series. This is of course possible by putting method calls into the callback functions of previous method calls, but the code quickly becomes ugly if you do this with a lot of methods: 'use strict'; var request = require('request'); var url = 'http://localhost:8080/'; request.get(url + 'getUserName?id=1234', function(err, res, body) { console.log('Name:', JSON.parse(body).value); request.get(url + 'getUserStatus?id=1234', function(err, res, body) { Optimizing code performance and control flow management using the async library 57 console.log('Status:', JSON.parse(body).value); request.get(url + 'getUserCountry?id=1234', function(err, res, body) { console.log('Country:', JSON.parse(body).value); request.get(url + 'getUserAge?id=1234', function(err, res, body) { console.log('Age:', JSON.parse(body).value); }); }); }); }); This is already starting to look messy, and we haven’t even added any notable “business logic” to our code. Note how our code is intended another level with every method call, creating the so-called “boomerang pattern” that is typical for multi-level nested callback control flows. We can use async.series to achieve the same control flow with much cleaner code: 1 'use strict'; 2 3 var request = require('request'); 4 var async = require('async'); 5 6 var url = 'http://localhost:8080/'; 7 8 async.series([ 9 10 function(callback) { 11 request.get(url + 'getUserName?id=1234', function(err, res, body) { 12 console.log('Name:', JSON.parse(body).value); 13 callback(null); 14 }); 15 }, 16 17 function(callback) { 18 request.get(url + 'getUserStatus?id=1234', function(err, res, body) { 19 console.log('Status:', JSON.parse(body).value); 20 callback(null); Optimizing code performance and control flow management using the async library 58 21 }); 22 }, 23 24 function(callback) { 25 request.get(url + 'getUserCountry?id=1234', function(err, res, body) { 26 console.log('Country:', JSON.parse(body).value); 27 callback(null); 28 }); 29 }, 30 31 function(callback) { 32 request.get(url + 'getUserAge?id=1234', function(err, res, body) { 33 console.log('Age:', JSON.parse(body).value); 34 callback(null); 35 }); 36 } 37 38 ]); Just as with async.parallel, we can use async.series to collect the results of each step and do something with them once all steps have finished. This is again achieved by passing the result of each step to the callback each step triggers, and by providing a callback function to the async.series call which will receive an array of all results: 1 'use strict'; 2 3 var request = require('request'); 4 var async = require('async'); 5 6 var url = 'http://localhost:8080/'; 7 8 async.series([ 9 10 function(callback) { 11 request.get(url + 'getUserName?id=1234', function(err, res, body) { 12 callback(null,'Name: ' + JSON.parse(body).value); 13 }); 14 }, 15 16 function(callback) { 17 request.get(url + 'getUserStatus?id=1234', function(err, res, body) { 18 callback(null,'Status: ' + JSON.parse(body).value); Optimizing code performance and control flow management using the async library 59 19 }); 20 }, 21 22 function(callback) { 23 request.get(url + 'getUserCountry?id=1234', function(err, res, body) { 24 callback(null,'Country: ' + JSON.parse(body).value); 25 }); 26 }, 27 28 function(callback) { 29 request.get(url + 'getUserAge?id=1234', function(err, res, body) { 30 callback(null,'Age: ' + JSON.parse(body).value); 31 }); 32 } 33 34 ], 35 36 function(err, results) { 37 for (var i=0; i < results.length; i++){ 38 console.log(results[i]); 39 } 40 } 41 42 ); In case that one of the series steps passes a non-null value to its callback as the first parameter, the series is immediately stopped, and the final callback is triggered with the results that have been collected to far, and the err parameter set to the error value passed by the failing step. async.waterfall is similar to async.series, as it executes all steps in series, but it also enables us to access the results of a previous step in the step that follows: 1 'use strict'; 2 3 var request = require('request'); 4 var async = require('async'); 5 6 var url = 'http://localhost:8080/'; 7 8 async.waterfall([ 9 10 function(callback) { 11 request.get(url + 'getSessionId', function(err, res, body) { Optimizing code performance and control flow management using the async library 60 12 callback(null, JSON.parse(body).value); 13 }); 14 }, 15 16 function(sId, callback) { 17 request.get(url + 'getUserId?sessionId=' + sId, function(err, res, body) { 18 callback(null, sId, JSON.parse(body).value); 19 }); 20 }, 21 22 function(sId, uId, callback) { 23 request.get(url + 'getUserName?userId=' + uId, function(err, res, body) { 24 callback(null, JSON.parse(body).value, sId); 25 }); 26 } 27 28 ], 29 30 function(err, name, sId) { 31 console.log('Name:', name); 32 console.log('SessionID:', sId); 33 } 34 35 ); Note how for each step function, callback is passed as the last argument. It follows a list of arguments for each parameter that is passed by the previous function, minus the error argument which each step function always passes as the first parameter to the callback function. Also note the difference in the final callback: instead of results, it too expects a list of result values, passed by the last waterfall step. async provides several other interesting methods which help us to bring order in our control flow and allows us to orchestrate tasks in an efficient manner. Check out the async README¹ for more details. ¹https://github.com/caolan/async/blob/master/README.md Node.js and MySQL Using the node-mysql library Node.js is able to connect to MySQL database servers, but support for this is not built in. We need to make use of a Node.js package that handles the low level details of establishing a connection and talking to the database. Additionally, such a package should enable us to issue statements to the database written in SQL. The most widely accepted package for this is node-mysql by Node.js core contributor Felix Geisendörfer. In order to use it, let’s start by declaring mysql as a dependency for our project in our package.json file: { "dependencies":{ "mysql":"~2.0.0" } } As always, npm install pulls in everything we need to start working with a MySQL database. For what follows, I assume that you have a working MySQL database up and running on your local machine. You should be able to connect to it locally. You can verify that this is the case by running mysql -h127.0.0.1 -uroot -p on your command line. Provide the database root password if needed - you will also need this password in your Node.js code in order to establish a connection. A first database application Once you are up and running, let’s start with a first, simple database application. Create a file index.js in the directory where you installed the node-mysql module, and fill it with the following code: Node.js and MySQL 62 1 'use strict'; 2 3 var mysql = require('mysql'); 4 5 var connection = mysql.createConnection({ 6 host:'localhost', 7 user:'root', 8 password:'root' 9 }); 10 11 connection.query( 12 'SELECT "foo" AS first_field, "bar" AS second_field', 13 function (err, results, fields) { 14 console.log(results); 15 connection.end(); 16 } 17 ); Let’s see what exactly happens here. First, we include Felix’ external mysql library (line 3). We then create a new object dubbed connection on line 5. This object is a representation of our connection with the database. No connection without connecting, which is why we call the createConnection function with the parameters necessary to connect and authenticate against the database server: the host name, the username, and the password. Note that there is a parameter port which defaults to 3306 if omitted. You only need to set this parameter if your database server does not run on this default port. For a list of all possible parameters (a total of 17 are available), see the node-mysql documentation at https://github.com/felixge/node-mysql#connection-options. Through the newly created connection object, we can now talk to the database server, which is what we do on line 11. To do so, the connection object provides a method named query. We provide two parameters to the method call. The first is an SQL statement, the second is an anonymous function. This function gets called when Node.js receives the answer to our query from the server - a callback. This is once again the classical asynchronous pattern of JavaScript applications that makes sense whenever our code triggers an operation whose execution could potentially block the program flow if executed synchronously: Imagine the database needing 5 seconds before it has compiled the answer to our query - that’s not unlikely for more complex queries. If we wouldn’t wait for the database’s answer asynchronously, then our whole application would have to wait for it. No other code could be executed for these 5 seconds, rendering our whole application unusable during this time. Node.js and MySQL 63 Line 12 shows which parameters the query method passes to our callback function: err, results, and fields. For now, let’s concentrate on results, because that’s where the fun is. What kind of thing is results? It’s simply an array of objects, with each object representing one row of the result set and each attribute of such an object representing one field of the according row. Let’s run our application and examine its output to see how this looks. Please run node index.js. The result should look like this: [ { first_field:'foo', second_field:'bar' }] Let’s add a legend: |-------------- An array ---------------------| | | | |------------ with an object -------------| | | | | | | | |------- that has an attribute | | | | | | | | | | with a value | | | | | | | | | | ----------- --- | | [ { first_field: 'foo', second_field: 'bar' } ] | | ----------- --- | | | | | | | | | | | with a value | | | | | | | | | |------- that has a field | | | | | | | |------------ with a row -----------------| | | | |-------------- A result set -----------------| (Note that we didn’t query for any real data from any real database table, but relational databases are really stubborn and treat everything as a set of rows of fields, even if we only query for some text strings mapped to made up fields like we just did.) According to this structure we can of course access the value of any field in any row directly, by accessing the right attribute of the right object: Node.js and MySQL 64 console.log(results[0].first_field); // Outputs 'foo'; Let’s examine the other parameters that are passed to our anonymous callback function. err is either an error object, if a problem occured, or null if everything went well. Let’s modify our code a bit to see how this works: 1 'use strict'; 2 3 var mysql = require('mysql'); 4 5 var connection = mysql.createConnection({ 6 host:'localhost', 7 user:'root', 8 password:'root' 9 }); 10 11 connection.query( 12 'SLECT "foo" AS first_field, "bar" AS second_field', 13 function (err, results, fields) { 14 console.log(err); 15 console.log(results); 16 connection.end(); 17 } 18 ); Note how I changed the SQL statement in line 12 and added logging of the err parameter on line 14. The altered SQL statement is invalid and results in the database answering the query with an error. node-mysql translates this error into an error object which is then passed as the first parameter to our callback function. If you run the new code, a string representation of the error object is written to the console: { [Error: ER_PARSE_ERROR: You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'SLECT "foo" AS first_field, "bar" AS second_field' at line 1] code: 'ER_PARSE_ERROR', errno: 1064, sqlState: '42000', index: 0 } The fact that the err parameter is null if no error occured makes error handling in our code very simple: Node.js and MySQL 65 1 'use strict'; 2 3 var mysql = require('mysql'); 4 5 var connection = mysql.createConnection({ 6 host:'localhost', 7 user:'root', 8 password:'root' 9 }); 10 11 connection.query( 12 'SLECT "foo" AS first_field, "bar" AS second_field', 13 function (err, results, fields) { 14 if (err) { 15 console.log('A database error occured!'); 16 } else { 17 console.log(results); 18 } 19 connection.end(); 20 } 21 ); That’s a very typical pattern in a lot of Node.js applications: Call a library function while providing a callback, and check if the first parameter passed to this callback is null in order to decide if an error occured, continuing otherwise. fields returns an array with one object for every field that was requested in the query - each object contains meta information about the according field. It isn’t of any significant use for us here. Feel free to console.log() the fields parameter if you like. Using node-mysql’s Streaming API Next, we will query some real data from existing tables. For this, we first need to create a database, add a table to it, and insert some rows. Let’s do this through a Node.js application. Please create a file create.js with the following content: Node.js and MySQL 66 1 'use strict'; 2 3 var mysql = require('mysql'); 4 5 var connection = mysql.createConnection({ 6 host:'localhost', 7 user:'root', 8 password:'root' 9 }); 10 11 connection.query('CREATE DATABASE node', function(err) { 12 if (err) { 13 console.log('Could not create database "node".'); 14 } 15 }); 16 17 connection.query('USE node', function(err) { 18 if (err) { 19 console.log('Could not switch to database "node".'); 20 } 21 }); 22 23 connection.query('CREATE TABLE test ' + 24 '(id INT(11) AUTO_INCREMENT, ' + 25 ' content VARCHAR(255), ' + 26 ' PRIMARY KEY(id))', 27 function(err) { 28 if (err) { 29 console.log('Could not create table "test".'); 30 } 31 } 32 ); 33 34 connection.query('INSERT INTO test (content) VALUES ("Hello")'); 35 connection.query('INSERT INTO test (content) VALUES ("World")'); 36 37 connection.end(); This program is meant to be executed just once in order to set up a new database with some example data. Please run it now. Let’s rewrite our index.js file and try to query for this newly added data: Node.js and MySQL 67 1 'use strict'; 2 3 var mysql = require('mysql'); 4 5 var connection = mysql.createConnection({ 6 host:'localhost', 7 user:'root', 8 password:'root', 9 database:'node' 10 }); 11 12 connection.query( 13 'SELECT id, content FROM test', 14 function (err, results, fields) { 15 if (err) { 16 console.log('A database error occured!'); 17 } else { 18 console.log(results); 19 } 20 connection.end(); 21 } 22 ); Please note how on line 9 I have added an additional connection parameter named database. This way, a default database is already selected for subsequent queries on this connection, and doesn’t have to be selected through a USE database query. If we execute the script, the result should look like this: [ { id: 1, content:'Hello' }, { id: 2, content:'World' }] As one would expect, we now receive an array with two objects, because the resultset consists of two rows. This gives us the opportunity to discuss an even better way to handle resultsets from queries. I have explained how the classical callback pattern should be used on the query call because otherwise expensive database operations could block the whole Node.js application. However, there is another source of potentially blocking behaviour: large resultsets. Imagine that our test database consists of 1,000,000 rows instead of just two, and we would like to output the content of the whole resultset to the console like this: Node.js and MySQL 68 connection.query( 'SELECT id, content FROM test', function (err, results, fields) { for (var i = 0; i < results.length; i++){ console.log('Content of id ' + results[i].id + ' is ' + results[i].content); } } ); Using the callback pattern ensures that our Node.js application won’t block while the database retrieves these 1,000,000 rows. However, once we receive the resultset with these rows in our application - that is, within the callback - and we start operating on them, executing these operations also will take a long time. During this time, our application will block, too. Remember that in Node.js, only one piece of our JavaScript code is executed at any given time. If we iterate over these 1,000,000 array entries and do something with each entry, and this takes, say, 10 seconds, then our application cannot do something else during these 10 seconds. The solution: instead of using the all-or-nothing callback pattern, we can utilize node-mysql’s streaming API. Instead of providing a callback function to the query call, we only provide the SQL statement, but we use the return value of the query function: a Query object. This object emits several events that happen during the lifetime of the query: error, field, row and end. Let’s concentrate on row and see how a more efficient version of our 1,000,000 row output application would look like: var query = connection.query('SELECT id, content FROM test'); query.on('row', function(row) { console.log('Content of id ' + row.id + ' is ' + row.content); }); Now we have a piece of our code running 1,000,000 times, once for each row. While this doesn’t make our application faster in the narrower sense (it will still take the assumed 10 seconds until all rows have been handled), it does make it way more efficient. Our application now only blocks during the handling of one row, immediately returning control to the event loop of our application, allowing to execute other code in between. Of course, this isn’t critical for an application that is used by only one person on one shell console. But it is very critical if such a database operation takes place within a multi-user client-server application: Node.js and MySQL 69 Imagine you use a website to retrieve a small amount of data from a database - you wouldn’t want to wait 10 seconds for your page to load only because another user requested a page that retrieved 1,000,000 rows for him from the database. Let’s rewrite our index.js script once again to demonstrate usage of the complete streaming API of node-mysql: 1 'use strict'; 2 3 var mysql = require('mysql'); 4 5 var connection = mysql.createConnection({ 6 host:'localhost', 7 user:'root', 8 password:'root', 9 database:'node' 10 }); 11 12 var query = connection.query('SELECT id, content FROM test'); 13 14 query.on('error', function(err) { 15 console.log('A database error occured:'); 16 console.log(err); 17 }); 18 19 query.on('fields', function(fields) { 20 console.log('Received fields information.'); 21 }); 22 23 query.on('result', function(result) { 24 console.log('Received result:'); 25 console.log(result); 26 }); 27 28 query.on('end', function() { 29 console.log('Query execution has finished.'); 30 connection.end(); 31 }); It’s pretty much self explanatory. Note how the order of the on blocks doesn’t say anything about when (and if) they are executed - it’s up to node-mysql to decide when each block (or, to be more precise: the anonymous callback function associated with each block) will be called. When executing this script, this is what the output should look like: Node.js and MySQL 70 Received fields information. Received result: { id: 1, content: 'Hello' } Received result: { id: 2, content: 'World' } Query execution has finished. Making SQL queries secure against attacks Let’s now go full circle and create a simple web application that allows to insert data into our table and also reads and displays the data that was entered. We need to start a web server with two routes (one for displaying data, one for taking user input), and we need to pass user input to the database and database results to the webpage. Here is the application in one go: 1 'use strict'; 2 3 var mysql = require('mysql'), 4 http = require('http'), 5 url = require('url'), 6 querystring = require('querystring'); 7 8 9 // Start a web server on port 8888. Requests go to function handleRequest 10 11 http.createServer(handleRequest).listen(8888); 12 13 14 // Function that handles http requests 15 16 function handleRequest(request, response) { 17 18 // Page HTML as one big string, with placeholder "DBCONTENT" for data from 19 // the database 20 var pageContent = '' + 21 '' + 22 '' + 24 '' + 25 '' + 26 '
' + Node.js and MySQL 71 27 '' + 28 '' + 29 '
' + 30 '
' + 31 'Content in database:' + 32 '
' +  33 'DBCONTENT' +  34 '
' + 35 '
' + 36 '
' + 37 '' + 38 '' + 39 '
' + 40 '' + 41 ''; 42 43 // Parsing the requested URL path in order to distinguish between 44 // the / page and the /add route 45 var pathname = url.parse(request.url).pathname; 46 47 // User wants to add content to the database (POST request to /add) 48 if (pathname == '/add'){ 49 var requestBody = ''; 50 var postParameters; 51 request.on('data', function (data) { 52 requestBody += data; 53 }); 54 request.on('end', function() { 55 postParameters = querystring.parse(requestBody); 56 // The content to be added is in POST parameter "content" 57 addContentToDatabase(postParameters.content, function() { 58 // Redirect back to homepage when the database has finished adding 59 // the new content to the database 60 response.writeHead(302,{'Location':'/'}); 61 response.end(); 62 }); 63 }); 64 65 // User wants to read data from the database (GET request to /) 66 } else { 67 // The text to use for filtering is in GET parameter "q" 68 var filter = querystring.parse(url.parse(request.url).query).q; Node.js and MySQL 72 69 getContentsFromDatabase(filter, function(contents) { 70 response.writeHead(200,{'Content-Type':'text/html'}); 71 // Poor man's templating system: Replace "DBCONTENT" in page HTML with 72 // the actual content we received from the database 73 response.write(pageContent.replace('DBCONTENT', contents)); 74 response.end(); 75 }); 76 } 77 } 78 79 80 // Function that is called by the code that handles the / route and 81 // retrieves contents from the database, applying a LIKE filter if one 82 // was supplied 83 84 function getContentsFromDatabase(filter, callback) { 85 var connection = mysql.createConnection({ 86 host:'localhost', 87 user:'root', 88 password:'root', 89 database:'node' 90 }); 91 var query; 92 var resultsAsString = ''; 93 94 if (filter) { 95 query = connection.query('SELECT id, content FROM test ' + 96 'WHERE content LIKE "' + filter + '%"'); 97 } else { 98 query = connection.query('SELECT id, content FROM test'); 99 } 100 101 query.on('error', function(err) { 102 console.log('A database error occured:'); 103 console.log(err); 104 }); 105 106 // With every result, build the string that is later replaced into 107 // the HTML of the homepage 108 query.on('result', function(result) { 109 resultsAsString += 'id: ' + result.id; 110 resultsAsString += ', content: ' + result.content; Node.js and MySQL 73 111 resultsAsString += '\n'; 112 }); 113 114 // When we have worked through all results, we call the callback 115 // with our completed string 116 query.on('end', function(result) { 117 connection.end(); 118 callback(resultsAsString); 119 }); 120 } 121 122 123 // Function that is called by the code that handles the /add route 124 // and inserts the supplied string as a new content entry 125 126 function addContentToDatabase(content, callback) { 127 var connection = mysql.createConnection({ 128 host:'localhost', 129 user:'root', 130 password:'root', 131 database:'node', 132 }); 133 134 connection.query('INSERT INTO test (content) ' + 135 'VALUES ("' + content + '")', 136 function(err) { 137 if (err) { 138 console.log('Could not insert content "' + content + 139 '" into database.'); 140 } 141 callback(); 142 }); 143 } Simply start the application by running node index.js, open up its homepage in your browser by visiting http://localhost:8888/, and play around with it. So, that’s quite a bit of code here. Let’s walk through the important parts. If the user requests the homepage, here is what happens: • The HTTP server that is started on line 11 receives the requests and calls the handleRequest function that is declared on line 16. Node.js and MySQL 74 • On line 45 the variable pathname is declared and initialized with the requested path, which is / in this case, by parsing the request URL. • The code continues on line 66, where the request for / is handled. • The handling starts by parsing the URL for the value of the GET parameter q which is later used to filter the database content. • On line 69 things start to get interesting. We call getContentsFromDatabase with two parameters: The filter string we just parsed from the URL, and an anonymous callback function. • Execution continues on line 84, starting with the creation of a new database connection. • Then, on lines 94 to 99, the query to retrieve the contents from the database is started. • Each retrieved row triggers the execution of the result event handler on line 108, where we extend the string resultsAsString with the contents of each row. • Execution jumps to line 116 once all rows have been retrieved. Here, in the end event handler, we close the connection to the database and then call the callback we passed into the getContentsFromDatabase function. This returns control back to the request handler, in line 70. • There, we start responding to the request, first by sending a status 200 header, followed by the page content - with the database content replaced into the HTML at the proper place on line 74. With this, the request has been handled and responded to. When the user uses the Filter content form, the same procedure is executed, with the only difference that now the GET parameter q has a value which is then used to filter the database results using the SQL LIKE clause. When using the Add content button, a different execution path is taken: • The HTTP server that is started on line 11 receives the requests and calls the handleRequest function that is declared on line 16. • On line 45 the variable pathname is declared and initialized with the requested path, which is /add in this case, by parsing the request URL. • The code continues on line 48, where the request for /add is handled. • Then, on line 51, the body of the request (the POST data) is retrieved step-by-step, in the same way that large resultsets from a database can be retrieved when using the streaming API. With every step, a variable is extended which finally holds the complete request body. • Once the request has been fully received, execution continues on line 54 when the end event is triggered for the request. • We then parse the request body in order to translate the POST data into a JavaScript object which allows us to easily access the value of POST parameters, as is done one line 57 where we call the addContentToDatabase function with the string that is to be added to the database as its first parameter, followed by a callback function that we expect to be triggered once the data has been inserted. Node.js and MySQL 75 • With this, execution continues on line 126 where we start by connecting to the database, followed by an INSERT query on line 134. The query function triggers our anonymous callback defined on line 135 when the query execution has finished, which in turn triggers the callback that was passed in from the request handler function. • This leads to execution continuing on line 60, where we redirect the user’s browser back to the homepage. So, nice application, does what it’s supposed to do. And has a huge problem. Why? Because the way we treat the incoming data that is provided by the user puts our database at risk. Our application is vulnerable to so-called SQL injection attacks. Let’s demonstrate how. To do so, let’s create another table within our database node, one that’s supposed to carry sensitive information that’s not meant for any user’s eyes. We will just pretend that our nice little content- viewer app is part of a bigger application that somewhere also stores user passwords. We will call this table passwords, and you need to create this within the node database on your local server, either using the MySQL command line, a MySQL GUI client you happen to use, or, if you like, you can simply rewrite the create.js script we used earlier, like this: 1 'use strict'; 2 3 var mysql = require('mysql'); 4 5 var connection = mysql.createConnection({ 6 host:'localhost', 7 user:'root', 8 password:'root' 9 }); 10 11 connection.query('USE node', function(err) { 12 if (err) { 13 console.log('Could not switch to database "node".'); 14 } 15 }); 16 17 connection.query('CREATE TABLE passwords ' + 18 '(id INT(11) AUTO_INCREMENT, ' + 19 ' password VARCHAR(255), ' + 20 ' PRIMARY KEY(id))', 21 function(err) { 22 if (err) { 23 console.log('Could not create table "passwords".'); 24 } Node.js and MySQL 76 25 } 26 ); 27 28 connection.query('INSERT INTO passwords (password) VALUES ("secret")'); 29 connection.query('INSERT INTO passwords (password) VALUES ("dont_tell")'); 30 31 connection.end(); Execute create.js through Node.js in order to create the new table. Now, here is the scenario: We have a table test that contains public information, and a table passwords that contains sensitive information. We have written an application that operates only on the test table, thus allowing public access to the public information in that table, but not allowing access to any sensitive information like the passwords in the passwords table. And yet, the passwords can be retrieved through the application. Here’s how to do that: • Start the application with node index.js • Visit http://localhost:8888 in your browser • Enter the following text into the second text input field and hit Filter content %" UNION SELECT id, password AS content FROM passwords WHERE password LIKE " Note that you need to enter the text exactly as shown above, starting with a percentage sign and ending with a quotation mark. You will receive a list of data from the test table (if any content has been added before), but at the end of this list, you will also see the two passwords we have added to the passwords table, like this: id: 1, content: foo bar id: 1, content: secret id: 2, content: dont_tell Whoops. How is that possible? It’s because a set of SQL commands has been injected into our application’s SQL, although only data - the string to filter for - was meant to be injected. Hence an SQL injection attack. If you look at the critical piece of our code, what happens becomes clear. On lines 95 and 96, our SQL query is constructed: Node.js and MySQL 77 query = connection.query('SELECT id, content FROM test ' + 'WHERE content LIKE "' + filter + '%"'); If the content of the filter variable is “foo”, this is what the final SQL looks like: SELECT id, content FROM test WHERE content LIKE "foo%" But if the malicious filter text is entered into the web page, we end up with the following SQL query: SELECT id, content FROM test WHERE content LIKE "%" UNION SELECT id, password AS content FROM passwords WHERE password LIKE "%" (I’ve added a line break in order to stay within the boundaries of the text page.) The malicious filter text simply closes the LIKE query of the original SQL and continues with an additional SELECT on the passwords table, combining the results of both SELECTs into a single (and identically structured) result set using the UNION clause. Another LIKE clause is added at the end of the malicious code, which makes sure that the whole SQL is syntactically correct. This is a made up example for demo purposes, but that’s how actual attacks look like. A real attack might need a lot more effort - e.g., the attacker needs to find out the structure of the passwords table, but supporting information can be gathered by querying meta data from the database; if an attacker finds a part of the application that allows to extend harmless SQL queries into more potent ones, finding sensitive information is only a question of time and patience. What can we, as application developers, do against such attacks? There is no 100% secure system, as the saying goes, however, there are a lot of ways to raise the bar, making attacks more difficult. In our scenario, a very effective counter-measure is to use placeholders with value assignment when building SQL statements, instead of hand-crafting the final query by concatenating strings. This way, we can guarantee that any part of your query that is supposed to be treated as a value is indeed interpreted by node-mysql as a value, escaping it in a way that doesn’t allow the value to end up at the database server as a command. To achieve this, we need a way to create a query that can unambiguously differentiate between the command parts of a query and the value parts. This is done by creating a query where the value parts are not directly included within the query string - instead, their position is marked with a special character, the quotation mark (?), as a placeholder. This is how our query looks when using this syntax: Node.js and MySQL 78 SELECT id, content FROM test WHERE content LIKE ? Now node-mysql knows that whatever is going to be used at the position marked by the ? placeholder is a value, treating it accordingly. Of course, when issueing the query, we also need a way to tell what value to use. This is done by mapping the value onto the query-with-placeholder, like so: filter = filter + '%'; query = connection.query('SELECT id, content FROM test WHERE content LIKE ?', filter); In our specific case, we need to add the search wildcard, %, to the filter parameter. We then construct the query with the placeholder, passing the value for this placeholder as the second parameter. The rest of the code remains unchanged. We should change the INSERT query that is used for adding content, too, by rewriting line 135 to connection.query('INSERT INTO test (content) VALUES (?)', content, function(err) { ... This changes the SQL query to placeholder syntax, and adds the content variable as the second parameter. The callback function is now the third parameter. Mapping multiple values onto multiple placeholders is done by providing an array of values: query('SELECT foo FROM bar WHERE id = ? AND name = ?',[42,'John Doe']); Here, “42” maps to the id placeholder, and “John Doe” to the name placeholder. If you restart the application ad try to attack it again, you will see that the attack no longer works. Mission accomplished. Summary In this chapter we learned about the node-mysql module, how it enables us to connect to relational databases, how using asynchronous callbacks and the Streaming API enable us to write efficient applications even when large resultsets need to be handled, and analyzed how database applications can be attacked and, more importantly, how we can protect our applications against these kinds of attacks. Node.js and MongoDB Some MongoDB basics MongoDB is a document-oriented NoSQL database that stores objects in the following hierarchy: Server \_Database \_Collection \_Document \_Document \_Document \_ ... \_Collection \_Document \_Document \_Document \_ ... If you are coming from a relational SQL database background (I do), then it might help to think of collections as tables, and of documents as rows in that table. It’s not a perfect analogy, however; the key idea of a document-oriented database like MongoDB is that documents within the same collection do not need to have the same structure, while each row in a table of a relational database does have the same structure. Here is how actual data in this hierarchy might actually look like: localhost \_accounting \_customers \_ { _id: 53c6c2, name: 'Jane Doe' } \_ { _id: 9dc42e, name: 'John Doe', age: 24 } \_ { _id: 63a76d, name: 'Jim Doe', options: { billtime: 3 } } \_invoices \_ { _id: 98ef5a, value: 192.87 } \_ { _id: f4eb21, value: 192.87, recurring: true } Node.js and MongoDB 80 This example represents the case where one has a MongoDB server running on localhost, with two databases customers and invoices. Each database contains some documents that are similar, but not identical, in structure. As you can see, I have represented the document objects in JavaScript object notation - this is how we are going to encounter them when working with a MongoDB server from within Node.js. Every document has an _id attribute, which is a random 24-digit hex value - I have shortened them for the illustration. We will now examine how to create and work with such a structure. Applying CRUD operations with the low-level mongodb driver Again, I won’t cover installation of the server software. Please refer to the official installation instructions² in order to get a MongoDB server up and running on your machine. Once this is done, create a new project folder called mongodb-test, and in there create an initial package.json as follows: { "name":"mongodb-test", "version":"0.0.1", "description":"", "dependencies":{ "mongodb":"^1.4.7" }, "devDependencies":{} } As always, we pull in the dependencies with npm install Now we can create a first program in file index.js: ²http://docs.mongodb.org/manual/installation/ Node.js and MongoDB 81 1 'use strict'; 2 3 var MongoClient = require('mongodb').MongoClient; 4 5 MongoClient.connect( 6 'mongodb://127.0.0.1:27017/accounting', 7 function(err, connection) { 8 var collection = connection.collection('customers'); 9 10 collection.insert({'name':'Jane Doe'}, function(err, count) { 11 12 collection.find().toArray(function(err, documents) { 13 console.dir(documents); 14 connection.close(); 15 }); 16 17 }); 18 19 }); Let’s see what this code does. On line 3, we require the MongoClient class from the mongodb library. We use the connect method of this class on line 5 to establish a connection to our MongoDB server, which is listening on localhost, port 27017. Upon connecting, we declare that we want to use the accounting database. We then create an object collection on line 8 which represents the customers collection within the accounting database. Using the insert method of the collection object, we add a new document to the customers collection on line 10. When the insert operation has finished, we query the collection for all existing documents using the find method on line 12. Our callback receives an array with the documents, prints them to the screen, and then closes the database connection. Running this code by executing node index.js will result in output similar to [ { _id: 53c8ba1b517c86d3c9df71e6, name: 'Jane Doe' } ] and running it again will result in output similar to Node.js and MongoDB 82 [ { _id: 53c8ba1b517c86d3c9df71e6, name: 'Jane Doe' }, { _id: 53c8ba21d13eb8dbc96e2c19, name: 'Jane Doe' } ] As you can see, each run creates a new document in the database. The name attribute is set by our code, the _id is generated by MongoDB. Let’s rewrite our code a bit in order to update existing documents: 1 'use strict'; 2 3 var MongoClient = require('mongodb').MongoClient; 4 5 MongoClient.connect( 6 'mongodb://127.0.0.1:27017/accounting', 7 function(err, connection) { 8 var collection = connection.collection('customers'); 9 10 collection.update({}, {'$set':{'age': 24}}, function(err, count) { 11 12 console.log('Updated', count, 'documents'); 13 14 collection.find().toArray(function(err, documents) { 15 console.dir(documents); 16 connection.close(); 17 }); 18 19 }); 20 21 }); The change is on line 10. Instead of inserting a new document, we want to update our documents. The first parameter is the filter that defines which documents to match when applying the update. In this case, we pass an empty object, that is, we don’t want to match any specific document and instead we want to apply the update to all documents in the collection. The second parameter defines what to update within each matched document. In this case, we want to set the age attribute of the documents to 24. The documents we already stored do not have an age attribute yet, but that’s not a problem: because MongoDB works schema-less, it will happily add attributes that did not yet exist on a given document. This is how the content of our collection looks like after running the new code: Node.js and MongoDB 83 Updated 1 documents [ { _id: 53c8ba1b517c86d3c9df71e6, name: 'Jane Doe' }, { _id: 53c8ba21d13eb8dbc96e2c19, name: 'Jane Doe', age: 24 } ] Oh, well. Not quite what we expected - we didn’t define a filter to the update statement because we wanted to match all existing documents, and yet only one of our two documents has been updated with the age attribute. The reason is that this is the default behaviour for MongoDB updates: only the first match for a given filter - even if it matches every document - is updated. By changing line 10 from collection.update({}, {'$set':{'age': 24}}, function(err, count) { to collection.update( {}, {'$set':{'age': 24}}, {'multi': true}, function(err, count) { (that is, by adding a parameter {'multi': true}), we override this behaviour, which then results in MongoDB updating all matching documents. Let’s have a look at how actual filters that match only some of our documents look like. We are going to rewrite our script because we need some more documents to play with: 1 'use strict'; 2 3 var MongoClient = require('mongodb').MongoClient; 4 5 MongoClient.connect( 6 'mongodb://127.0.0.1:27017/accounting', 7 function (err, connection) { 8 var collection = connection.collection('customers'); 9 10 var doFind = function (callback) { 11 collection.find().toArray(function (err, documents) { 12 console.dir(documents); 13 callback(); 14 }); Node.js and MongoDB 84 15 }; 16 17 var doInsert = function (i) { 18 if (i < 20){ 19 var value = Math.floor(Math.random() * 10); 20 collection.insert( 21 {'n':'#' + i, 'v': value}, 22 function (err, count) { 23 doInsert(i + 1); 24 }); 25 } else { 26 console.log(); 27 console.log('Inserted', i, 'documents:'); 28 doFind(function () { 29 doUpdate(); 30 }); 31 } 32 }; 33 34 var doUpdate = function () { 35 collection.update( 36 {'v':{'$gt': 5}}, 37 {'$set':{'valuable': true}}, 38 {'multi': true}, 39 function (err, count) { 40 console.log(); 41 console.log('Updated', count, 'documents:'); 42 doFind(function () { 43 collection.remove({}, function () { 44 connection.close(); 45 }); 46 }); 47 }); 48 }; 49 50 doInsert(0); 51 52 }); As you can see, we created three functions, doFind, doInsert and doUpdate. The script starts by calling doInsert, which is a recursive function that inserts 20 documents into the collection, and does so in a serial fashion by recursively calling itself (on line 23) every time the previous insert operation Node.js and MongoDB 85 has finished. Each documents has an n attribute with the serial number of its insert operation, and a v value, which is a random number between 0 and 9. As soon as the limit of 20 inserts has been reached, the function calls doFind (on line 28) which prints all existing documents to the screen. We pass doFind a callback function which calls doUpdate. This function adds the field valuable to some, but not all, documents. Whether a document is updated or not depends on it’s v value: using the $gt operator (on line 36), we filter the documents that are affected by the update to those whose value is greater than 5. After the update operation has finished, doFind is called once again, printing the results of the update to the screen. We then delete all documents using remove (on line 43), again using a filter that matches any document - however, remove doesn’t need a multi parameter in order to work on all matched documents; it does so by default. Here is a typical result for a run of this script: Inserted 20 documents: [ { _id: 53d60cc676371927029f95bd, n: '#0', v: 5 }, { _id: 53d60cc676371927029f95be, n: '#1', v: 8 }, { _id: 53d60cc676371927029f95bf, n: '#2', v: 0 }, { _id: 53d60cc676371927029f95c0, n: '#3', v: 4 }, { _id: 53d60cc676371927029f95c1, n: '#4', v: 6 }, { _id: 53d60cc676371927029f95c2, n: '#5', v: 2 }, { _id: 53d60cc676371927029f95c3, n: '#6', v: 4 }, { _id: 53d60cc676371927029f95c4, n: '#7', v: 1 }, { _id: 53d60cc676371927029f95c5, n: '#8', v: 0 }, { _id: 53d60cc676371927029f95c6, n: '#9', v: 7 }, { _id: 53d60cc676371927029f95c7, n: '#10', v: 5 }, { _id: 53d60cc676371927029f95c8, n: '#11', v: 8 }, { _id: 53d60cc676371927029f95c9, n: '#12', v: 2 }, { _id: 53d60cc676371927029f95ca, n: '#13', v: 8 }, { _id: 53d60cc676371927029f95cb, n: '#14', v: 1 }, { _id: 53d60cc676371927029f95cc, n: '#15', v: 0 }, { _id: 53d60cc676371927029f95cd, n: '#16', v: 1 }, { _id: 53d60cc676371927029f95ce, n: '#17', v: 4 }, { _id: 53d60cc676371927029f95cf, n: '#18', v: 4 }, { _id: 53d60cc676371927029f95d0, n: '#19', v: 6 } ] Updated 6 documents: [ { _id: 53d60cc676371927029f95bd, n: '#0', v: 5 }, { _id: 53d60cc676371927029f95bf, n: '#2', v: 0 }, { _id: 53d60cc676371927029f95c0, n: '#3', v: 4 }, { _id: 53d60cc676371927029f95c2, n: '#5', v: 2 }, { _id: 53d60cc676371927029f95c3, n: '#6', v: 4 }, Node.js and MongoDB 86 { _id: 53d60cc676371927029f95c4, n: '#7', v: 1 }, { _id: 53d60cc676371927029f95c5, n: '#8', v: 0 }, { _id: 53d60cc676371927029f95c7, n: '#10', v: 5 }, { _id: 53d60cc676371927029f95c9, n: '#12', v: 2 }, { _id: 53d60cc676371927029f95cb, n: '#14', v: 1 }, { _id: 53d60cc676371927029f95cc, n: '#15', v: 0 }, { _id: 53d60cc676371927029f95cd, n: '#16', v: 1 }, { _id: 53d60cc676371927029f95ce, n: '#17', v: 4 }, { _id: 53d60cc676371927029f95cf, n: '#18', v: 4 }, { _id: 53d60cc676371927029f95be, n: '#1', v: 8, valuable: true }, { _id: 53d60cc676371927029f95c1, n: '#4', v: 6, valuable: true }, { _id: 53d60cc676371927029f95c6, n: '#9', v: 7, valuable: true }, { _id: 53d60cc676371927029f95c8, n: '#11', v: 8, valuable: true }, { _id: 53d60cc676371927029f95ca, n: '#13', v: 8, valuable: true }, { _id: 53d60cc676371927029f95d0, n: '#19', v: 6, valuable: true } ] As you can see, 20 documents with a random v value are inserted. Afterwards, all documents whose value is greater than 5 have been updated to include the attribute valuable set to true - a total of 6 documents in this case. Retrieving specific documents using filters We have not yet performed any truly interesting queries against our collection. Up until now, all we did was to retrieve all documents contained in our collection by performing collection.find() without any filter. More complex filters are available of course. Let’s rewrite the existing code in order to retrieve only specific documents in our doFind function. First, a very simple example - we are going to retrieve all documents where the v attribute has a value of 5. To do so, we need to change line 11 from collection.find().toArray(function (err, documents) { to collection.find({'v': 5}).toArray(function (err, documents) { which results in only the matching documents being printed: Node.js and MongoDB 87 Inserted 20 documents: [ { _id: 53d7698d99c6107303ad204c, n: '#8', v: 5 }, { _id: 53d7698d99c6107303ad2050, n: '#12', v: 5 } ] Updated 10 documents: [ { _id: 53d7698d99c6107303ad204c, n: '#8', v: 5 }, { _id: 53d7698d99c6107303ad2050, n: '#12', v: 5 } ] We can write filters that match more than just one attribute, like this: collection.find( { 'v': 6, 'valuable': true } ).toArray(function (err, documents) { Both attributes of each document in the collection must match the given values for a document to be included. In other words, this is an AND query. For the given collection, the above query doesn’t make too much sense, because after the update, every document with a v value of 6 has the valuable attribute set to true anyways. But we might want to filter for all valuable documents whose v value is less than 8, like so: collection.find( { 'v':{'$lt': 8 }, 'valuable': true } ).toArray(function (err, documents) { We already encountered the greater than operator $gt, and $lt is the opposite operator, less than. Other available operators are: • $lte: less than or equal • $gte: greater than or equal • $ne: not equal Filters can also be combined into an OR query, if, for example, we want to query for all documents whose v value is either 5 or 8: Node.js and MongoDB 88 collection.find( { '$or':[ {'v': 5}, {'v': 8} ] } ).toArray(function (err, documents) { And of course it is also possible to combine an AND query and an OR query: collection.find( { 'v':{'$gt': 3}, '$or':[ {'n':'#5'}, {'n':'#10'} ] } ).toArray(function (err, documents) { This will retrieve all documents that have a v value greater than 3 AND whose n value is #5 OR #10. Speaking of the n attribute: it holds a string value, which cannot be queried with operators like $gt or $lt, of course. We can, however, apply regular expressions: collection.find( { 'n':/^#1/ } ).toArray(function (err, documents) { This matches all documents whose n value starts with #1. All the filter variations discussed so far can also be applied to the update and remove methods of the collection object. This way, you can update or remove only specific documents in a given collection. The find method also takes an options parameter. Possible options are limit, skip, and sort: Node.js and MongoDB 89 collection.find( { 'n':/^#1/ }, { 'limit': 5, 'skip' : 2, 'sort' :'v' } ).toArray(function (err, documents) { This one retrieves five documents whose n value starts with #1, skips the first two matches, and sorts the result set by the v value: Inserted 20 documents: [ { _id: 53d7739bbe04f81c05c13b6f, n: '#19', v: 5 }, { _id: 53d7739bbe04f81c05c13b66, n: '#10', v: 5 }, { _id: 53d7739bbe04f81c05c13b6b, n: '#15', v: 6 }, { _id: 53d7739bbe04f81c05c13b67, n: '#11', v: 7 }, { _id: 53d7739bbe04f81c05c13b68, n: '#12', v: 7 } ] Updated 9 documents: [ { _id: 53d7739bbe04f81c05c13b6f, n: '#19', v: 5 }, { _id: 53d7739bbe04f81c05c13b66, n: '#10', v: 5 }, { _id: 53d7739bbe04f81c05c13b6b, n: '#15', v: 6, valuable: true }, { _id: 53d7739bbe04f81c05c13b68, n: '#12', v: 7, valuable: true }, { _id: 53d7739bbe04f81c05c13b67, n: '#11', v: 7, valuable: true } ] We can also sort over multiple fields and in different directions: collection.find( { 'n':/^#1/ }, { 'limit': 5, 'skip' : 2, 'sort' :[['v','asc'], ['n','desc']] }).toArray(function (err, documents) { This sorts the matching documents first by their v value in ascending order, and documents with an identical v value are then ordered by their n value, in descending order, as can be seen here with the documents whose v value is 1: Node.js and MongoDB 90 Inserted 20 documents: [ { _id: 53d77475c0a6e83b0556d817, n: '#14', v: 1 }, { _id: 53d77475c0a6e83b0556d815, n: '#12', v: 1 }, { _id: 53d77475c0a6e83b0556d818, n: '#15', v: 3 }, { _id: 53d77475c0a6e83b0556d81a, n: '#17', v: 4 }, { _id: 53d77475c0a6e83b0556d814, n: '#11', v: 6 } ] Updated 7 documents: [ { _id: 53d77475c0a6e83b0556d817, n: '#14', v: 1 }, { _id: 53d77475c0a6e83b0556d815, n: '#12', v: 1 }, { _id: 53d77475c0a6e83b0556d818, n: '#15', v: 3 }, { _id: 53d77475c0a6e83b0556d81a, n: '#17', v: 4 }, { _id: 53d77475c0a6e83b0556d814, n: '#11', v: 6, valuable: true } ] More complex update operations We already learned how to add attributes or change the value of attributes in existing documents: collection.update( {'n':/^#1/}, {'$set':{'v': 5} }, {'multi': true}, function (err, count) { // ... } This would set the value of the v attribute to 5 for all documents whose n value starts with #1. This is probably the most regularly used update operation. Some other ways to update documents are available, however. Numeric attributes can be increased, for example: collection.update( {'n':/^#1/}, {'$inc':{'v': +1} }, {'multi': true}, function (err, count) { // ... } There is no $dec operation, but of course, we can increase by -1: Node.js and MongoDB 91 collection.update( {'n':/^#1/}, {'$inc':{'v':-1} }, {'multi': true}, function (err, count) { // ... } Multiplication is possible, too: collection.update( {'n':/^#1/}, {'$mul':{'v': 2} }, {'multi': true}, function (err, count) { // ... } There are several other operators: • $rename Renames a field. • $unset Removes the specified field from a document. • $min Only updates the field if the specified value is less than the existing field value. • $max Only updates the field if the specified value is greater than the existing field value. • $currentDate Sets the value of a field to current date, either as a Date or a Timestamp. Please refer to the MongoDB documentation³ for a detailed discussion of each operation. The update method can be used to perform an operation where documents are updated if the filter being used is able to match existing documents, but a new document is inserts instead if no existing document could be matched. This is called an upsert: ³http://docs.mongodb.org/manual/reference/operator/update-field/ Node.js and MongoDB 92 collection.update( {'n':'#20'}, {'$set':{'n':'20','v': 1} }, {'multi': true,'upsert': true}, function (err, count) { // ... } In our initial collection, only entries with n from #0 to #19 exist. Therefore, an update call with a filter set to {'n': '#20'} cannot match any document in the collection. Because the upsert option is set to true, a new document is inserted: Inserted 20 documents: [ { _id: 53d789d4fff187d607f03b3a, n: '#0', v: 7 }, ... { _id: 53d789d4fff187d607f03b4c, n: '#18', v: 8 }, { _id: 53d789d4fff187d607f03b4d, n: '#19', v: 5 } ] Updated 1 documents: [ { _id: 53d789d4fff187d607f03b3a, n: '#0', v: 7 }, ... { _id: 53d789d4fff187d607f03b4c, n: '#18', v: 8 }, { _id: 53d789d4fff187d607f03b4d, n: '#19', v: 5 }, { _id: 53d789d44266a0f5acef6f05, n: '#20', v: 1 } ] Working with indexes Like other databases, MongoDB needs to index data sets if we want queries on large sets to perform well. In order to experience this firsthand, let’s write a script which inserts 200,000 documents instead of just 20: Node.js and MongoDB 93 1 'use strict'; 2 3 var MongoClient = require('mongodb').MongoClient; 4 5 MongoClient.connect( 6 'mongodb://127.0.0.1:27017/accounting', 7 function (err, connection) { 8 var collection = connection.collection('customers'); 9 10 var doInsert = function (i) { 11 if (i < 200000){ 12 var value = Math.floor(Math.random() * 10); 13 collection.insert( 14 {'n':'#' + i, 'v': value}, 15 function (err, count) { 16 doInsert(i + 1); 17 }); 18 } else { 19 connection.close(); 20 } 21 }; 22 23 doInsert(0); 24 25 }); Call this script insert.js and execute it running node insert.js. Now, we time how long it takes MongoDB to return 10 specific rows from this collection, using the following script: 1 'use strict'; 2 3 var MongoClient = require('mongodb').MongoClient; 4 5 MongoClient.connect( 6 'mongodb://127.0.0.1:27017/accounting', 7 function (err, connection) { 8 var collection = connection.collection('customers'); 9 10 collection.find( 11 {'v':{'$gt': 5}}, 12 { Node.js and MongoDB 94 13 'skip:': 100000, 14 'limit': 10, 15 'sort':'v' 16 } 17 ).toArray(function (err, documents) { 18 console.dir(documents); 19 connection.close(); 20 }); 21 22 }); This script queries the database for 10 documents whose v attribute is greater than 5, starting at document 100,000, sorting results by their v value. Name this script query.js and run it like this: time node query.js This way, you’ll get a poor man’s performance benchmark. This is the result on my machine: ~$ time node query.js [ { _id: 53d7955c3ac66b9e0af26745, n: '#36271', v: 6 }, { _id: 53d7955c3ac66b9e0af26757, n: '#36289', v: 6 }, { _id: 53d7955c3ac66b9e0af26761, n: '#36299', v: 6 }, { _id: 53d7955c3ac66b9e0af2677a, n: '#36324', v: 6 }, { _id: 53d7955c3ac66b9e0af2677b, n: '#36325', v: 6 }, { _id: 53d7955c3ac66b9e0af26792, n: '#36348', v: 6 }, { _id: 53d7955c3ac66b9e0af26794, n: '#36350', v: 6 }, { _id: 53d7955c3ac66b9e0af267b2, n: '#36380', v: 6 }, { _id: 53d7955c3ac66b9e0af267b3, n: '#36381', v: 6 }, { _id: 53d7955c3ac66b9e0af267b6, n: '#36384', v: 6 } ] real 0m1.078s user 0m0.238s sys 0m0.040s The script runtime is 1 second on average. Now, write the script that adds an index on the v attribute: Node.js and MongoDB 95 1 'use strict'; 2 3 var MongoClient = require('mongodb').MongoClient; 4 5 MongoClient.connect( 6 'mongodb://127.0.0.1:27017/accounting', 7 function (err, connection) { 8 var collection = connection.collection('customers'); 9 10 collection.ensureIndex('v', function(err, indexName) { 11 connection.close(); 12 }); 13 14 } 15 ); Call this one addIndex.js and execute it. Then, run the benchmark again. This is my result: ~$ time node query.js [ { _id: 53d7955c3ac66b9e0af26745, n: '#36271', v: 6 }, { _id: 53d7955c3ac66b9e0af26757, n: '#36289', v: 6 }, { _id: 53d7955c3ac66b9e0af26761, n: '#36299', v: 6 }, { _id: 53d7955c3ac66b9e0af2677a, n: '#36324', v: 6 }, { _id: 53d7955c3ac66b9e0af2677b, n: '#36325', v: 6 }, { _id: 53d7955c3ac66b9e0af26792, n: '#36348', v: 6 }, { _id: 53d7955c3ac66b9e0af26794, n: '#36350', v: 6 }, { _id: 53d7955c3ac66b9e0af267b2, n: '#36380', v: 6 }, { _id: 53d7955c3ac66b9e0af267b3, n: '#36381', v: 6 }, { _id: 53d7955c3ac66b9e0af267b6, n: '#36384', v: 6 } ] real 0m0.269s user 0m0.236s sys 0m0.033s From 1 second to 0.2 seconds - quite an improvement. This effect becomes more significant the larger the collection is. With a 2,000,000 documents collection, the query takes 5.6 seconds without an index and still only 0.2 seconds with an index on my machine. Querying collections efficiently When retrieving large result sets from a MongoDB collection, the same rule that applies to MySQL database result sets also applies here: reading the complete result set into our Node.js process at once Node.js and MongoDB 96 isn’t going to be efficient resource-wise. Handling a result array with 2,000,000 entries will eat a lot of memory no matter what. This is what the toArray method, which we used until now, does: collection.find().toArray(function (err, documents) { // We now have one large documents array }); What collection.find() returns is a so-called cursor object. It can be transformed into an array, which is very convenient, but when handling lots of documents, it’s better to handle each one separately using the cursor’s each method: collection.find().each(function (err, document){ // We retrieve one document with each callback invocation }); This way, the cursor object will invoke the callback function for each document in the query result set, which means that Node.js only needs to claim memory for one document at a time. Thus, instead of collection.find().toArray(function (err, documents) { console.dir(documents); }); we can do collection.find().each(function (err, document){ console.dir(document); }); if we want to print a whole result set to the screen. The callback will be invoked with the document parameter set to null if the end of the result set has been reached - therefore, we can rewrite our doFind-doInsert-doUpdate script like this: Node.js and MongoDB 97 1 'use strict'; 2 3 var MongoClient = require('mongodb').MongoClient; 4 5 MongoClient.connect( 6 'mongodb://127.0.0.1:27017/accounting', 7 function (err, connection) { 8 var collection = connection.collection('customers'); 9 10 var doFind = function (callback) { 11 collection.find( 12 {}, 13 {'sort':'_id'} 14 ).each(function (err, document){ 15 if (document === null){ 16 callback(); 17 } else { 18 console.dir(document); 19 } 20 }); 21 }; 22 23 var doInsert = function (i) { 24 if (i < 20){ 25 var value = Math.floor(Math.random() * 10); 26 collection.insert( 27 {'n':'#' + i, 'v': value}, 28 function (err, count) { 29 doInsert(i + 1); 30 }); 31 } else { 32 console.log(); 33 console.log('Inserted', i, 'documents:'); 34 doFind(function () { 35 doUpdate(); 36 }); 37 } 38 }; 39 40 var doUpdate = function () { 41 collection.update( 42 {'n':/^#1/}, Node.js and MongoDB 98 43 {'$mul':{'v': 2} }, 44 {'multi': true}, 45 function (err, count) { 46 console.log(); 47 console.log('Updated', count, 'documents:'); 48 doFind(function () { 49 collection.remove({}, function () { 50 connection.close(); 51 }); 52 }); 53 }); 54 }; 55 56 doInsert(0); 57 58 }); The mongodb driver also provides a convenient way to access the records of a result set through a streaming API. The following code excerpt shows a version of the doFind function which has been rewritten to make use of this API: var doFind = function (callback) { var stream = collection.find( {}, {'sort':'_id'} ).stream(); stream.on('data', function(document){ console.dir(document); }); stream.on('close', function() { callback(); }); }; As you can see, it’s quite simple: The stream method of the cursor object returned by collection.find() returns a stream object which emits data and close events, among others. We can then attach callback functions to these events where we either print the incoming document, or call the provided callback function in order to continue our script. Node.js and MongoDB 99 The stream object returned by collection.find().stream() implements the stream.Readable interface. See the corresponding Node.js documentation chapter⁴ for more information. Summary In this chapter we learned about the workings of MongoDB and the mongodb driver, which enables us to insert, update, retrieve and remove documents from this NoSQL database. We learned about query filters and complex update operations, demonstrated how indexes can significantly accelerate query operations, and showed how to handle large result sets efficiently. ⁴http://nodejs.org/api/stream.html#stream_class_stream_readable Part 2: Building a complete web application with Node.js and AngularJS Introduction I want to admit it right away: Looking at the amount of code, this part is probably more about AngularJS than Node.js - but I think it’s important to see how Node.js can be used as one part in a complete application. Also, I’m a huge fan of AngularJS and am firmly convinced that knowing Node.js and AngularJS in combination is extremely valuable. Nevertheless, besides teaching you a bit about building AngularJS frontends, the chapter does teach you a lot about Node.js, too. We will learn how to create a full-fledged RESTful Webservice API, and probably even more important, we will learn how to do so in a fully test-driven manner using end-to-end test specifications. Additionally, we will talk about using a database abstraction layer, and we will learn how to integrate database migrations into our development and testing workflow. All in all, we will learn a lot about creating Node.js applications for the real world. The code of the final application is also available at https://github.com/manuelkiessling/nodecraftsman/tree/master/keyword-wrangler The requirements from a user’s perspective What we are going to build is an application which allows its users to comfortably manage a set of structured data from the world of food and cooking. The data basically is a collection of keywords like Aubergine, Knife, or Paul Bocuse, and these keywords are mapped into categories, like Vegetable, Utility, or Famous chef. What’s needed now is an application where users can browse and edit this data: create new keywords, delete keywords, change the value of existing keywords, and change the category of keywords. Thus, at the heart of it we need a CRUD application, with a little bit of extra functionality. It has already been decided that the users want a web-based solution. The data will be stored in a relational database. This is how the user interface will look like: Introduction 102 +---------------+ +---------------+ New keyword: | | | Category V Create +---------------+ +---------------+ Keyword Category Ops +----------------+------------------------------+--------------+ | | | | | Aubergine | Vegetable | Save Delete | | | | | +--------------------------------------------------------------+ | | | | | Knife | Utility | Save Delete | | | | | +--------------------------------------------------------------+ | | | | | Paul Bocuse | Star chef | Save Delete | | | | | +--------------------------------------------------------------+ What this depicts is a screen with a table of keywords and the categories they are mapped to. Each row also has an operations column which allows to Save and Delete a keyword. Above this table is a form which allows to Create a new keyword. Being the JavaScript aficionados that we are, we are going to create the application as an AngularJS frontend that talks to a RESTful webservice API on the backend written in Node.js. High level architecture overview The following is an overview about the most important architectonical parts of our application: Introduction 103 Application Tests Tools +-----------------+ +-----------------+ | | | Bower | | AngularJS | +-----------------+ | Frontend | Frontend | | +-----------------+ +-----------------+ | ng-grid | +-----------------+ +-----------------+ | Restangular | +-----------------+ ^ | | v +-----------------+ +-----------------+ +-----------------+ | | | | | NPM | | Node.js |<----| End-to-End | +-----------------+ | Backend |---->| API Tests | Backend | | | | +-----------------+ +-----------------+ +-----------------+ +-----------------+ | Percolator | | Jasmine | +-----------------+ +-----------------+ +-----------------+ +-----------------+ | node-dbi |<------------+ | db-migrate | +-----------------+ | +-----------------+ ^ | | | | | | | | v v | +-----------------+ +-----------------+ | | MySQL or sqlite | | sqlite | | Persistence +-----------------+ +-----------------+ | ^----------------------^---------------------+ We will make use of the AngularJS plugin Restangular which will help us manage our webservice calls. Our backend server will be built using Percolator, a library which streamlines the creation of RESTful APIs with Node.js. Introduction 104 We are going to use different database solutions for testing & development (sqlite) and production (MySQL), which is why we will use node-dbi, a database access abstraction layer. In order to ease database schema management in different environments, we will use the db-migrate tool that handles the details of versioning the structural changes of our project’s database. Once again Jasmine will be our testing framework - this time, however, we are going to create end-to-end tests instead of unit tests. As always, NPM will be our trusty companion when it comes to dependency management on the backend - on the frontend side, bower will take that role. Setting up the development environment This is going to be our largest application to date, so let’s make sure we can work comfortably. Let’s start by creating a project folder - let’s call our project Keyword Wrangler: $ mkdir keyword-wrangler Next, we initialize a Node.js project in this folder by running npm init and answering the upcoming questions as follows: $ npm init This utility will walk you through creating a package.json file. It only covers the most common items, and tries to guess sane defaults. See `npm help json` for definitive documentation on these fields and exactly what they do. Use `npm install --save` afterwards to install a package and save it as a dependency in the package.json file. Press ^C at any time to quit. name: (keyword-wrangler) version: (1.0.0) 0.0.1 description: A tool to manage keywords entry point: (index.js) src/backend/index.js test command: git repository: keywords: author: license: (ISC) Introduction 105 About to write to keyword-wrangler/package.json: { "name": "keyword-wrangler", "version": "0.0.1", "description": "A tool to manage keywords", "main": "src/backend/index.js", "scripts": { "test": "echo \"Error: no test specified\" && exit 1" }, "author": "", "license": "ISC" } Is this ok? (yes) yes Make sure to execute the npm init command within the keyword-wrangler folder! This gives us a sensible package.json file to start with. We can now start to pull in all the dependencies we need. We are going to need the following Node.js packages: • percolator - that’s the REST framework we are going to use • node-dbi - an SQL abstraction layer • mysql - The MySQL db driver, used by node-dbi • sqlite3 - The sqlite3 db driver, used by node-dbi • db-migrate - an SQL migrations manager • async - we are going to write some asynchronous orchestration • jasmine - the test framework • request - this enables us to talk to the API we will build from within our tests • bower - the dependency and package manager for the frontend application Running $ npm install percolator node-dbi mysql db-migrate async bower --save $ npm install jasmine-node request sqlite3 --save-dev will install the most recent stable versions of these packages and add them as dependencies to the package.json file. We only need Jasmine and request during development, which is why we install them separately with --save-dev. This allows to selectively install only the non-dev packages for production (per default, however, npm always installs all dependencies). Milestone 1: A first passing test against the server We will approach the requirements through a set of milestones. The first milestone we will try to reach is a very first passing end-to-end Jasmine test case against our webservice API. Let’s see what we need to reach that milestone: • We need to be able to execute Jasmine test cases • We need a Jasmine spec that requests our http backend • We need an http backend that responds to the request In order to verify that we are good to go in regards to executing Jasmine test cases, let’t create a subfolder spec/e2e in our project directory: mkdir -p spec/e2e Next, we create the file that will hold the end-to-end tests for our yet-to-be-written API server. We name it apiSpec.js and store it within the spec/e2e folder we just created: 1 'use strict'; 2 3 describe('The API', function () { 4 5 it('should just work', function () { 6 expect(true).toBe(true); 7 }); 8 9 }); As long as the test infrastructure itself works correctly, this will of course pass. Let’s verify that: Milestone 1: A first passing test against the server 107 $ ./node_modules/.bin/jasmine-node --verbose ./spec/ The API - 1 ms should just work - 1 ms Finished in 0.004 seconds 1 test, 1 assertion, 0 failures, 0 skipped ./node_modules/.bin is a folder with shortcuts to all the executables within node_modules. The actual executable in this case is ./node_- modules/jasmine-node/bin/jasmine-node Ok, great, we can build on that. The goal of our first milestone is to have a test case which successfully requests a server backend and checks for the correct response. Let’s rewrite apiSpec.js accordingly: 1 'use strict'; 2 3 var request = require('request'); 4 5 describe('The API', function () { 6 7 it('should respond to a GET request at /api/keywords/', function (done) { 8 request.get( 9 { 10 'url':'http://localhost:8080/api/keywords/', 11 'json': true 12 }, 13 function (err, res, body) { 14 expect(res.statusCode).toBe(200); 15 expect(body.foo).toEqual('bar'); 16 done(); 17 }); 18 }); 19 20 }); If we re-execute the test command, we get - nothing. Not only a failed test run, but literally no output to the console whatsoever. The problem is that there isn’t any server yet that could respond to our GET request; however, Jasmine remains silent about that. The reason is that requesting the missing HTTP server leads to an exception, and we need to explicitly tell Jasmine that we want to see exceptions: Milestone 1: A first passing test against the server 108 $ ./node_modules/.bin/jasmine-node --verbose --captureExceptions ./spec/ TypeError: Cannot read property 'statusCode' of undefined at Request._callback (keyword-wrangler/spec/e2e/apiSpec.js:14:19) at self.callback (keyword-wrangler/node_modules/request/request.js:373:22) at Request.emit (events.js:95:17) at Request.onRequestError (keyword-wrangler/node_modules/request/request.js:\ 971:8) at ClientRequest.emit (events.js:95:17) at Socket.socketErrorListener (http.js:1552:9) at Socket.emit (events.js:95:17) at net.js:441:14 at process._tickCallback (node.js:442:13) Ok, it’s time to create a web server. As said, we will use a library for this, namely Percolator. The spec we defined expects a server at port 8080 that responds with {"foo": "bar"} when receiving a GET request at /api/keywords/. Let’s see how to implement this behaviour with Percolator. Put the following into file src/backend/index.js: 1 'use strict'; 2 3 var Percolator = require('percolator').Percolator; 4 5 var port = 8080; 6 var server = Percolator({'port': port}); 7 8 server.route('/api/keywords', 9 { 10 GET: function (req, res) { 11 res.object({'foo':'bar'}).send() 12 } 13 } 14 ); 15 16 server.listen(function() { 17 console.log('Server started and listening on port', port); 18 }); Then, start this server in a new console window: Milestone 1: A first passing test against the server 109 $ node src/backend/index.js Server started and listening on port 8080 Now, running the test command again should look much better: $ ./node_modules/.bin/jasmine-node --verbose --captureExceptions ./spec/ The API - 19 ms should respond to a GET request at /api/keywords/ - 18 ms Finished in 0.021 seconds 1 test, 2 assertions, 0 failures, 0 skipped Great, our first milestone has been reached. Milestone 2: The API responds with actual database content Let’s now look at how to integrate database access into our server application. We defined that within our local development and test environment, we would like to use sqlite. Let’s start with this. Abstracting database access In order to connect to the database, we are going to use an abstraction layer by utilizing node-dbi. This allows us to transparently use MySQL at a later point in time, without the need to change the code that does database operations. First, we should change our spec in order to have it expect actual database content. A request to /api/keywords/ should return a JSON structure that carries an array of objects, where each object represents a keyword with its ID, value, and the ID of its category: [ {"id": 1,"value":"Aubergine","categoryID": 1}, {"id": 2,"value":"Onion","categoryID": 1}, {"id": 3,"value":"Knife","categoryID": 2} ] A data structure like this would be stored in a relational database like this: Table: keyword +----------------------------------------------------------+ | id | value | categoryID | +==========================================================+ | 1 | Aubergine | 1 | +----------------------------------------------------------+ | 2 | Onion | 1 | +----------------------------------------------------------+ | 3 | Knife | 2 | +----------------------------------------------------------+ Milestone 2: The API responds with actual database content 111 Table: category +----------------------------------------------------------+ | id | name | +==========================================================+ | 1 | Vegetable | +----------------------------------------------------------+ | 2 | Utility | +----------------------------------------------------------+ In this structure, keyword.categoryID is a foreign key to category.id. For now, however, we will only work with the keyword table. Assuming the above example database content during a test run, our spec expectation needs to look like this (in file spec/e2e/apiSpec.js): 1 'use strict'; 2 3 var request = require('request'); 4 5 describe('The API', function () { 6 7 it('should respond to a GET request at /api/keywords/', function (done) { 8 var expected = { 9 "_items":[ 10 {'id': 1,'value':'Aubergine','categoryID': 1}, 11 {'id': 2,'value':'Onion','categoryID': 1}, 12 {'id': 3,'value':'Knife','categoryID': 2} 13 ] 14 }; 15 16 request.get( 17 { 18 'url':'http://localhost:8080/api/keywords/', 19 'json': true 20 }, 21 function (err, res, body) { 22 expect(res.statusCode).toBe(200); 23 expect(body).toEqual(expected); 24 done(); 25 }); 26 }); Milestone 2: The API responds with actual database content 112 27 28 }); Note how expected is an object with a field _items, which contains the actual list of keywords. Percolator doesn’t allow to return an array of things directly. There is still some work to do in order to make this succeed. Ensuring a clean slate for test runs We are going to make our lives a lot easier if we write some helper code for our tests which ensures that at the start of each test run, the database is empty. Let’s put this into spec/resetDatabase.js: 1 'use strict'; 2 3 var async = require('async'); 4 5 var resetDatabase = function (dbSession, callback) { 6 7 async.series( 8 [ 9 10 function (callback) { 11 dbSession.remove('keyword','1', function (err) { 12 callback(err) 13 }); 14 }, 15 16 function (callback) { 17 dbSession.remove('category','1', function (err) { 18 callback(err) 19 }); 20 } 21 22 ], 23 24 function (err, results) { 25 callback(err); 26 } 27 ); Milestone 2: The API responds with actual database content 113 28 29 }; 30 31 module.exports = resetDatabase; As you can see, we make use of a dbSession object. We need to create a module where this is set up, in file src/backend/dbSession.js: 1 'use strict'; 2 3 var DBWrapper = require('node-dbi').DBWrapper; 4 5 var dbWrapper = new DBWrapper('sqlite3',{'path':'/var/tmp/keyword-wrangler.tes\ 6 t.sqlite'}); 7 dbWrapper.connect(); 8 module.exports = dbWrapper; This is a very simplistic implementation - we will later refine it in order to allow for different database setups depending on the environment. Let’s now create the sqlite database file with the keyword and category table: $ sqlite3 /var/tmp/keyword-wrangler.test.sqlite sqlite> CREATE TABLE keyword (id INTEGER PRIMARY KEY, value TEXT, categoryID INT\ EGER); sqlite> CREATE TABLE category (id INTEGER PRIMARY KEY, name TEXT); sqlite> .quit Install the sqlite3 package if it is not available on your system! Completing the first spec We now have the boilerplate code and database structures needed to further complete our spec: Milestone 2: The API responds with actual database content 114 1 'use strict'; 2 3 var request = require('request'); 4 var dbSession = require('../../src/backend/dbSession.js'); 5 var resetDatabase = require('../resetDatabase.js'); 6 var async = require('async'); 7 8 describe('The API', function () { 9 10 it('should respond to a GET request at /api/keywords/', function (done) { 11 var expected = { 12 "_items":[ 13 {'id': 1,'value':'Aubergine','categoryID': 1}, 14 {'id': 2,'value':'Onion','categoryID': 1}, 15 {'id': 3,'value':'Knife','categoryID': 2} 16 ] 17 }; 18 19 async.series( 20 [ 21 22 function(callback) { 23 resetDatabase(dbSession, callback); 24 }, 25 26 function(callback) { 27 dbSession.insert( 28 'keyword', 29 {'value':'Aubergine','categoryID': 1}, 30 function(err) { callback(err) }); 31 }, 32 33 function(callback) { 34 dbSession.insert( 35 'keyword', 36 {'value':'Onion','categoryID': 1}, 37 function(err) { callback(err) }); 38 }, 39 40 function(callback) { 41 dbSession.insert( 42 'keyword', Milestone 2: The API responds with actual database content 115 43 {'value':'Knife','categoryID': 2}, 44 function(err) { callback(err) }); 45 } 46 47 ], 48 49 function(err, results) { 50 request.get( 51 { 52 'url':'http://localhost:8080/api/keywords/', 53 'json': true 54 }, 55 function (err, res, body) { 56 expect(res.statusCode).toBe(200); 57 expect(body).toEqual(expected); 58 done(); 59 } 60 ); 61 } 62 63 ); 64 65 }); 66 67 }); With this, our spec is complete - we empty the database, insert our test data, and expect the web service API to respond with the data we just inserted. This is not yet the behaviour of the API - we need to rewrite src/backend/index.js: 1 'use strict'; 2 3 var Percolator = require('percolator').Percolator; 4 var dbSession = require('../../src/backend/dbSession.js'); 5 6 var port = 8080; 7 var server = Percolator({'port': port, 'autoLink': false}); 8 9 server.route('/api/keywords', 10 { 11 GET: function (req, res) { 12 dbSession.fetchAll('SELECT id, value, categoryID FROM keyword ORDER BY id'\ Milestone 2: The API responds with actual database content 116 13 , function(err, rows) { 14 if (err) { 15 console.log(err); 16 res.status.internalServerError(err); 17 } else { 18 res.collection(rows).send(); 19 } 20 }); 21 } 22 } 23 ); 24 25 server.listen(function() { 26 console.log('Server started and listening on port', port); 27 }); Don’t forget to restart the server after making changes to it! Our tests now pass again, and thus our second milestone has been reached. Milestone 3: Setting the stage for a continuous delivery workflow We are already growing our application driven by tests, however, there are still some inconveniences that hinder us from going fully continuous. If the goal is to put every change we make into production automatically (safeguarded by our tests, of course), then we need to automate the things we still do manually: starting the server and changing the database structure. Let’s tackle the server startup first. It would be great if each test case would start and stop the server process as needed. This is possible by refactoring the server code a bit. We need to split into two files what is currently handled in src/backend/index.js. The goal here is to have the server logic in one file, and the server startup in another one. This way, our tests can start the server logic up by themselves, and for production, we still have a file we can execute. To achieve this, let’s create a file src/backend/server.js with the following content: 1 'use strict'; 2 3 var Percolator = require('percolator').Percolator; 4 var dbSession = require('../../src/backend/dbSession.js'); 5 6 var Server = function(port) { 7 var server = Percolator({'port': port, 'autoLink': false}); 8 9 server.route('/api/keywords', 10 { 11 GET: function (req, res) { 12 dbSession.fetchAll('SELECT id, value, categoryID FROM keyword ORDER BY i\ 13 d', function(err, rows) { 14 if (err) { 15 console.log(err); 16 res.status.internalServerError(err); 17 } else { 18 res.collection(rows).send(); 19 } 20 }); 21 } 22 } Milestone 3: Setting the stage for a continuous delivery workflow 118 23 ); 24 25 return server; 26 }; 27 28 module.exports = {'Server': Server}; This isn’t that much different from what we had in index.js until now. But instead of configuring the server and starting it up directly, we now merely provide a Server function that sets up a Percolator server object for us. It’s then our business to get that server going - we do so in src/backend/index.js: 1 'use strict'; 2 3 var Server = require('./server.js').Server; 4 5 var server = Server('8080'); 6 7 server.listen(function() { 8 console.log('Server started and listening on port', server.options.port); 9 }); This way, we can still start the server via node src/backend/index.js. However, we can now also manage the server from other parts of our code - for example, from our tests in spec/e2e/apiSpec.js. 1 'use strict'; 2 3 var request = require('request'); 4 var dbSession = require('../../src/backend/dbSession.js'); 5 var Server = require('../../src/backend/server.js').Server; 6 var resetDatabase = require('../resetDatabase.js'); 7 var async = require('async'); 8 9 describe('The API', function () { 10 11 var server; 12 13 beforeEach(function (done) { 14 server = Server('8081'); 15 server.listen(function (err) { 16 resetDatabase(dbSession, function() { 17 done(err); 18 }); Milestone 3: Setting the stage for a continuous delivery workflow 119 19 }); 20 }); 21 22 afterEach(function (done) { 23 server.close(function () { 24 resetDatabase(dbSession, function() { 25 done(); 26 }); 27 }); 28 }); 29 30 it('should respond to a GET request at /api/keywords/', function (done) { 31 var expected = { 32 "_items":[ 33 {'id': 1,'value':'Aubergine','categoryID': 1}, 34 {'id': 2,'value':'Onion','categoryID': 1}, 35 {'id': 3,'value':'Knife','categoryID': 2} 36 ] 37 }; 38 39 async.series( 40 [ 41 42 function(callback) { 43 dbSession.insert( 44 'keyword', 45 {'value':'Aubergine','categoryID': 1}, 46 function(err) { callback(err) }); 47 }, 48 49 function(callback) { 50 dbSession.insert( 51 'keyword', 52 {'value':'Onion','categoryID': 1}, 53 function(err) { callback(err) }); 54 }, 55 56 function(callback) { 57 dbSession.insert( 58 'keyword', 59 {'value':'Knife','categoryID': 2}, 60 function(err) { callback(err) }); Milestone 3: Setting the stage for a continuous delivery workflow 120 61 } 62 63 ], 64 65 function(err, results) { 66 if (err) throw(err); 67 request.get( 68 { 69 'url':'http://localhost:8081/api/keywords/', 70 'json': true 71 }, 72 function (err, res, body) { 73 expect(res.statusCode).toBe(200); 74 expect(body).toEqual(expected); 75 done(); 76 } 77 ); 78 } 79 80 ); 81 82 }); 83 84 }); Now, our specs no longer require an already running server: before each test case, the server is started, and after each case, it is shut down. We also moved the database reset step into these blocks. Also, we are now using port 8081 for our tests, versus port 8080 which is used for production (do not forget to change the port on line 14). With this, you can now stop the running server, and you can execute test runs without the need to start the server. Introducing automatic database migrations Let’s take the final step in automation by setting up database migrations. A migration is a change to the database structure of our application. E.g., if we change our code and it now expects a new field in one of our tables, the migration would be the code that alters the table structure and adds the missing field. Making these migrations part of our code base allows us to automate this - when the tests or the application is started, all migrations that are not yet applied to the database are applied automatically; we no longer need to ensure manually that our code and our database structures are in sync. Milestone 3: Setting the stage for a continuous delivery workflow 121 The Node.js module that takes care of this for us is db-migrate. We already installed it via npm. This modules expects a database.json file where it can find the database connection parameters for different environments. An environment is, for example, test, or production. This way, we can use different database configurations for our different environments. We declared that we would like to use sqlite for our tests and MySQL in production; the database.json file makes this kind of setup simple (and, as we will see, we can use it for db-migrate and from our own code). Here is the database.json file we need: 1 { 2 "test":{ 3 "driver":"sqlite3", 4 "filename":"/var/tmp/keyword-wrangler.test.sqlite" 5 }, 6 7 "dev":{ 8 "driver":"sqlite3", 9 "filename":"/var/tmp/keyword-wrangler.dev.sqlite" 10 }, 11 12 "production":{ 13 "driver":"mysql", 14 "host":"localhost", 15 "user":"root", 16 "password":"root", 17 "database":"keyword_wrangler" 18 } 19 } The file needs to reside in the root folder of our project. We also need a place for our migrations. By default, the migration files are placed in a folder called migrations, again in the root folder of our project. We can now create our first migration; it will programmatically create the keyword and category tables we already created manually. Migrations are files with JavaScript code, but we don’t need to create them manually. db-migrate does this for us: $ ./node_modules/.bin/db-migrate create createKeywordAndCategoryTable --env test This gives you a file with a name like migrations/20150127080251-createKeywordAndCategoryTable.js and the following content: Milestone 3: Setting the stage for a continuous delivery workflow 122 1 var dbm = require('db-migrate'); 2 var type = dbm.dataType; 3 4 exports.up = function(db, callback) { 5 6 }; 7 8 exports.down = function(db, callback) { 9 10 }; Two functions, one up, one down. The up function carries the code that is to be executed if a migration needs to be applied. The down function carries the reverse, i.e., the code that is needed to un-apply or rollback the migration. Given our database structure, here is how this looks in practice: 1 'use strict'; 2 3 var dbm = require('db-migrate'); 4 var type = dbm.dataType; 5 var async = require('async'); 6 7 exports.up = function(db, callback) { 8 async.series( 9 [ 10 11 db.createTable.bind(db, 'keyword',{ 12 id:{ type:'int', primaryKey: true, autoIncrement: true, notNull: true \ 13 }, 14 value:{ type:'string', length:'128', notNull: true, unique: true }, 15 categoryID:{ type:'int', notNull: true } 16 }), 17 18 db.createTable.bind(db, 'category',{ 19 id:{ type:'int', primaryKey: true, autoIncrement: true, notNull: true \ 20 }, 21 name:{ type:'string', length:'128', notNull: true } 22 }) 23 24 ], callback); 25 }; 26 Milestone 3: Setting the stage for a continuous delivery workflow 123 27 exports.down = function(db, callback) { 28 async.series( 29 [ 30 db.dropTable.bind(db, 'keyword'), 31 db.dropTable.bind(db, 'category') 32 ], callback); 33 }; The up step will create our tables, the down step will remove them. This is the equivalent of the manual SQL operations we did earlier. And because we already did that manually, we need to remove them before our migrations can work. The most simple way to do this is by deleting the sqlite database file: $ rm /var/tmp/keyword-wrangler.test.sqlite We are now able to run our migrations: $ ./node_modules/.bin/db-migrate up --env test [INFO] Processed migration 20150127080251-createKeywordAndCategoryTable [INFO] Done Looks good. This creates the tables, and additionally creates a table named migrations, where db- migrate keeps track of the migrations already applied: $ sqlite3 /var/tmp/keyword-wrangler.test.sqlite sqlite> SELECT * FROM migrations; 1|20150127080251-createKeywordAndCategoryTable|1422435911741 So, from now on we need to combine our migration runs with our test runs; whenever we change the code and its tests in a way that makes them expect a different database structure, we need to create a new migrations file, program the required database changes into them, and can then apply them, followed by a test run: $ ./node_modules/.bin/db-migrate up --env test && \ ./node_modules/.bin/jasmine-node --verbose --captureExceptions ./spec/ The great thing is that now all development steps can be automatized, and we can ensure that code and database are always in sync. Imagine that another developer clones the repository with your code. Given that she has a working Node.js environment, all she needs to do is: Milestone 3: Setting the stage for a continuous delivery workflow 124 • check out the code • run npm install • run the migrations and she is immediately up-to-date with her development environment for your project. Milestone 3 has been reached! Milestone 4: Giving users a frontend We have now reached a point where our backend service deserves a frontend. As said, we want this to be based on AngularJS. Furthermore, we will use the Restangular library (which will ease the communication with our backend webservice), and ngGrid, which makes building dynamic tables straightforward. Setting up frontend dependencies through bower The very first step is to utilize Bower, the package manager for frontend JavaScript projects, to pull in the dependencies of our project. We need to create a directory src/frontent, and place a bower.json file into it with the following contents: 1 { 2 "name":"keyword-wrangler-client", 3 "dependencies":{ 4 "angular":"1.3.15", 5 "angular-route":"1.3.15", 6 "angular-grid":"2.0.14", 7 "restangular":"1.4.0", 8 "bootstrap":"3.3.4", 9 "async":"0.9.2" 10 }, 11 "analytics": false 12 } Then, while in folder keyword-wrangler/src/frontend, run: $ ../../node_modules/.bin/bower install This will install all dependecies into src/frontend/bower_components. Serving the frontend through the backend server Before we start building our frontend app, let’s make sure we can browse to it. To do so, change line 7 in src/backend/server.js from Milestone 4: Giving users a frontend 126 var server = Percolator({'port': port, 'autoLink': false}); to var server = Percolator({'port': port, 'autoLink': false,'staticDir':__dirname\ + '/../frontend'}); With this change, Percolator will serve everything in src/frontend as static files (while still serving the API at the /api route). Adding the frontend code Time to start building the app. Create src/frontend/index.html as follows: 1 2 3 4 5 6 Keyword Wrangler 7 8 10 12 18 19 20 21
22
23
24 25 26 27 Milestone 4: Giving users a frontend 127 28 29 30 31 32 33 34 35 36 37 38 39 When running the backend server locally, this will be served at http://localhost:8080/, and will host our single-page app. As you can see on line 32 and following, we need to provide several JavaScript files - these are the components of our AngularJS application. Let’s start with src/frontend/app/app.js: 1 'use strict'; 2 3 (function() { 4 5 var app = angular.module('app',['ngRoute','ngGrid','restangular']); 6 7 app.config(['$routeProvider', 8 function($routeProvider) { 9 10 // This makes app/keywords/KeywordsController.js handle the http://localho\ 11 st:8080/#/ URL 12 $routeProvider. 13 when('/',{ 14 templateUrl:'app/keywords/partials/editor.html', 15 controller:'KeywordsController' 16 }); 17 18 }]); 19 20 })(); This defines the main entry point for the AngularJS applications. It sets up dependencies and in-app routing. Milestone 4: Giving users a frontend 128 The key component here is the KeywordController. Put this into src/frontend/app/keywords/Key- wordsController.js: 1 'use strict'; 2 3 (function() { 4 5 var app = angular.module('app'); 6 7 app.controller('KeywordsController', function($scope, RepositoryFactory, resol\ 8 veEntity) { 9 10 /* == Frontend Initialization == */ 11 12 /* All of this happens as soon as the page loads */ 13 14 /* resolveEntity is a helper function which is used in partials/keywordCateg\ 15 oryGridCell.html 16 in order to display the name of a keyword category based on its id */ 17 $scope.resolveEntity = resolveEntity; 18 19 /* A repository is the connection between this controller and the REST Api. 20 We use one for keyword categories... */ 21 var KeywordCategoriesRepository = new RepositoryFactory({ 22 endpoint:'keywords/categories', 23 retrieveItems: function (data) { 24 return data._items; 25 } 26 }); 27 28 /* ...and one for keywords */ 29 var KeywordsRepository = new RepositoryFactory({ 30 endpoint:'keywords', 31 retrieveItems: function(data) { 32 return data._items; 33 } 34 }); 35 36 /* When the frontend loads, we want the controller to immediately 37 load all keyword categories and categories from the API */ 38 KeywordCategoriesRepository.readAll().then(function(keywordCategories) { 39 $scope.keywordCategories = keywordCategories; 40 KeywordsRepository.readAll().then(function(keywords) { Milestone 4: Giving users a frontend 129 41 $scope.keywords = keywords; 42 }); 43 }); 44 45 /* The grid. This part is best coded while listening to the Tron: Legacy sou\ 46 ndtrack. */ 47 $scope.keywordsGridOptions = { 48 data:'keywords',// This makes the grid use the data in $scope.keywords 49 enableCellSelection: false,// breaks edit of cells with select element if\ 50 true 51 enableCellEdit: true, 52 keepLastSelected: false, 53 enableRowSelection: false, 54 multiSelect: false, 55 enableSorting: true, 56 enableColumnResize: true, 57 enableColumnReordering: true, 58 showFilter: false, 59 rowHeight:'40', 60 columnDefs:[ 61 { 62 field:'id', 63 displayName:'ID', 64 enableCellEdit: false, 65 width:'80px' 66 }, 67 { 68 field:'value', 69 displayName:'Value' 70 }, 71 { 72 /* The keyword category field does not use the build-in cell template,\ 73 but our own */ 74 field:'keywordCategoryID', 75 displayName:'Category', 76 cellTemplate:'app/keywords/partials/keywordCategoryGridCell.html', 77 editableCellTemplate:'app/keywords/partials/keywordCategoryGridCellEd\ 78 itor.html' 79 }, 80 { 81 /* Same goes for the operations column */ 82 field:'', Milestone 4: Giving users a frontend 130 83 displayName:'Operations', 84 cellTemplate:'app/keywords/partials/operationsGridCell.html', 85 enableCellEdit: false, 86 sortable: false 87 } 88 ] 89 }; 90 91 92 /* == Frontend Operations == */ 93 94 /* These functions are called when the frontend is operated, e.g., if a butt\ 95 on is clicked */ 96 97 $scope.createKeyword = function(newKeyword) { 98 $scope.$broadcast('ngGridEventEndCellEdit'); 99 if (newKeyword.value.length > 0){ 100 KeywordsRepository.createOne(newKeyword).then(function () { 101 KeywordsRepository.readAll().then(function (keywords) { 102 $scope.keywords = keywords; 103 }); 104 }); 105 } 106 }; 107 108 $scope.updateKeyword = function(keyword) { 109 $scope.$broadcast('ngGridEventEndCellEdit'); 110 KeywordsRepository.updateOne(keyword); 111 }; 112 113 $scope.deleteKeyword = function(keyword) { 114 $scope.$broadcast('ngGridEventEndCellEdit'); 115 KeywordsRepository.deleteOne(keyword).then(function() { 116 KeywordsRepository.readAll().then(function(keywords) { 117 $scope.keywords = keywords; 118 }); 119 }); 120 }; 121 122 /* These are here to make the grid behave cleanly in regards to the keyword \ 123 category select */ 124 $scope.stopEditingKeywordCategory = function() { Milestone 4: Giving users a frontend 131 125 $scope.$broadcast('ngGridEventEndCellEdit'); 126 }; 127 128 $scope.$on('ngGridEventRows', function(newRows) { 129 $scope.$broadcast('ngGridEventEndCellEdit'); 130 }); 131 132 }); 133 134 })(); This controller has two dependencies, RepositoryFactory and resolveEntity. Let’s define those. In file src/frontend/app/lib/RepositoryFactory.js: 1 'use strict'; 2 3 (function() { 4 5 var app = angular.module('app'); 6 7 app.factory( 8 'RepositoryFactory', 9 ['Restangular','$q', RepositoryFactory] 10 ); 11 12 function RepositoryFactory(Restangular, $q) { 13 14 Restangular.setBaseUrl('/api/'); 15 16 var Repository = function(options) { 17 this.endpoint = options.endpoint; 18 this.retrieveItems = options.retrieveItems; 19 }; 20 21 Repository.prototype.readAll = function() { 22 var self = this; 23 var deferred = $q.defer(); 24 Restangular.all(self.endpoint + '/').doGET().then(function(data) { 25 var items = self.retrieveItems(data); 26 deferred.resolve(items); 27 }); 28 return deferred.promise; Milestone 4: Giving users a frontend 132 29 }; 30 31 Repository.prototype.createOne = function(newItem) { 32 var self = this; 33 var deferred = $q.defer(); 34 Restangular.one(self.endpoint + '/','').post('', newItem).then(function(r\ 35 esponse) { 36 deferred.resolve(response); 37 }); 38 return deferred.promise; 39 }; 40 41 Repository.prototype.updateOne = function(item) { 42 var self = this; 43 var deferred = $q.defer(); 44 Restangular.one(self.endpoint, item.id).post('', item).then(function(respo\ 45 nse) { 46 deferred.resolve(response); 47 }); 48 return deferred.promise; 49 }; 50 51 Repository.prototype.deleteOne = function(item) { 52 var self = this; 53 var deferred = $q.defer(); 54 Restangular.one(self.endpoint, item.id).remove().then(function(response) { 55 deferred.resolve(response); 56 }); 57 return deferred.promise; 58 }; 59 60 return Repository; 61 62 } 63 64 })(); This is basically a convenience wrapper for all the different API operations we are going to need for keywords and keyword categories: Reading, creating, updating, deleting, and merging. Note how each operation is asynchronous and therefore returns promises. You can see how these operations are used in the controller on lines 37, 39, 95, 96, 105, 110, 111, 123 and 126. Milestone 4: Giving users a frontend 133 Of course, the webservice routing and logic for these operations does not yet exist in our backend server (with the exception of GET /api/keywords. We will implement those soon. We also need to create src/frontend/app/lib/resolveEntity.js, which is a helper function that is needed to get from category IDs to category names in the view of our app: 1 'use strict'; 2 3 (function() { 4 5 var app = angular.module('app'); 6 7 app.factory( 8 'resolveEntity', 9 [resolveEntity] 10 ); 11 12 function resolveEntity() { 13 14 return function(queryParams) { 15 for (var i=0; i < queryParams.from.length; i++){ 16 if (queryParams.from[i][queryParams.where] === queryParams.is) { 17 return queryParams.from[i][queryParams.what]; 18 } 19 } 20 return false; 21 }; 22 23 } 24 25 })(); Adding AngularJS view templates With this, the JavaScript code for our AngularJS app is in place, but we aren’t done yet. We also need four HTML views for the app: The main view with the editor, plus three custom grid cell templates, one for the operations column that holds the Save and Delete controls, one for the Category column cells (because there we need to translate from category IDs to category names), and one for the Category cells in edit mode, because when we edit the category of a keyword, we don’t want to edit its name as a string, but rather choose from a list of existing categories through a dropdown control. We start with src/frontend/app/keywords/partials/editor.html: Milestone 4: Giving users a frontend 134 1
2
3 10 16
21 22
This provides us with the form for creating new keywords, plus it provides the div element where ngGrid will render the table. Note how on line 15 the createKeyword method of our KeywordsController is called upon clicking the Create button. Now we create src/frontend/app/keywords/partials/operationsGridCell.html: 1