Nodejs Native Add-Ons from zero to hero

Editor's Notes

• #2: Good afternoon everyone. Thank you to attend my talk, today I’m very honored to be here and share my ideas with all of you.
• #3: My name is Nicola Del Gobbo and I’m developer at Packly. In my every day work I take care of building Packly’s backend systems. In my spare time I try to give my contribution to all frameworks, libraries and modules that I use in my work and obviously I like creating new one. Today the talk is focused on Node.js native addon start with its definition.
• #4: In a very simple way Native Addons could be considered as C / C++ code called from JavaScript. They are a bridge between our Application programming language JavaScript and Native environment that is completely written in C / C++. This allow us to call C / C++ functions and methods directly from JavaScript.
• #5: If we take a look at Node.js documentation under Addons section Native Addons are defined as: dynamically-linked shared objects, written in C++, that can be loaded into Node.js using the require() function, and used just as if they were an ordinary Node.js module Their main purpose is to provide an interface between JavaScript running in Node.js and C / C++ libraries. This increase the developer experience because if you are using a Native Addon you don’t care of this, but you just use it as any other pure JavaScript module.
• #6: How is it possible? All the magic is behind the Node.js architecture. Node.js is built on some C / C++ libraries the most important for us as native addon developers are V8 and libuv. V8 is a JavaScript engine that take care to compile JavaScript code, execute it and give us great performance. Through the V8 APIs we can create function, object and do almost everything that we do on the JavaScript side. Libuv is a high performance evented I/O library and has the responsibility to handle the event loop (our main thread) and all the asynchronous tasks.
• #7: Why should you implement new Native Add-On? The first reason is for performance In general C / C++ code performs better then JavaScript code, but it’s really true for CPU bounds operations. Think for example at: Image processing | Video processing | Compression | Scientific calculus | Cyclic redundancy check and all algorithms that at one point execute some CPU heavy tasks. In all these cases you can gain more in performance.
• #8: Sometimes you want just integrate a legacy application So you have the source code of an old C / C++ application and want to expose something of its functionalities through new Node.js application
• #9: Even if npm is a very large repo where you can find almost everything sometimes you don’t find what fits your specific needs and in this case you have to take a choice. Completely reimplementing the module in JavaScript, but in my opinion it’s an insane idea to reimplement from the ground solutions like ImageMagick | Ghostscript | Ffmpeg and TensorFlow because you will lose all your time on develop a module and not to concentrate on business logic of your application. In these cases the You have to concentrate on business logic of your application and not on developing a single module
• #17: N-API is one of the most important features announced with Node.js 8 which is aimed at reducing the maintenance cost for native add-ons
• #18: When I start to develop a new native addon I use the boilerplate depicted on the slide. Here the most important elements are the SRC and LIB folders. SRC folder contains the C / C++ code and the native dependencies while the LIB folder contains the JavaScript code There are also the usual package.json file and the binging.gyp file that contains all the building configurations for the addon. At the end our addon will be composed by two parts one written in C++ and another written in JavaScript and usually the JavaScript part will use the native code to expose some features
• #19: When I start to develop a new native addon I use the boilerplate depicted on the slide. Here the most important elements are the SRC and LIB folders. SRC folder contains the C / C++ code and the native dependencies while the LIB folder contains the JavaScript code There are also the usual package.json file and the binging.gyp file that contains all the building configurations for the addon. At the end our addon will be composed by two parts one written in C++ and another written in JavaScript and usually the JavaScript part will use the native code to expose some features
• #20: When I start to develop a new native addon I use the boilerplate depicted on the slide. Here the most important elements are the SRC and LIB folders. SRC folder contains the C / C++ code and the native dependencies while the LIB folder contains the JavaScript code There are also the usual package.json file and the binging.gyp file that contains all the building configurations for the addon. At the end our addon will be composed by two parts one written in C++ and another written in JavaScript and usually the JavaScript part will use the native code to expose some features
• #21: In our binding.gyp we need to set the target name that will match the module name. The reported example is almost simple sometimes finding the right settings for binding.gyp is not easy so I suggest to take a look at GYP documentation or at nodey-gyp wiki where there are examples of existing binding.gyp and use them as source of inspiration to solve your problems with these kind of configurations.
• #22: What happen here? The macro NODE_API_MODULE creates code that will register a module named ”echo” and in addition it will ensure that a function init is called when the module is required. In the init we export our function or object. In the reported example we export echo and you can find the code executed by this function in the method Echo. As first thing we validate the input data, then transfer the JavaScript input to C++ data structure. C++ code perform the requested computation and before returning transfer the output data to JavaScript context. This is the most straightforward integration pattern that you can realize to pass data between JavaScript and C / C++ . The main reason for using this approach is simplicity and because you can integrate the C++ without modification. In addition we have a complete decoupling of the JavaScript and C++ code.
• #25: In an asynchronous addon function, the calling JavaScript code returns immediately. The calling code passes a callback function to the addon, and the addon does its work in a separate worker thread. This avoids locking up the Node.js event loop, as the addon function does not block.