{"rowid": 310, "title": "Fairytale of new Promise", "contents": "There are only four good Christmas songs.\nI know, yeah, JavaScript or whatever. We\u2019ll get to that in a minute, I promise.\nFirst\u2014and I cannot stress this enough\u2014 there are four good Christmas songs. You\u2019re free to disagree with me here, of course, but please try to understand that you will be wrong.\nThey don\u2019t all have the most safe-for-work titles; I can\u2019t list all of them here, but if you choose to let your fingers do the walkin\u2019 to your nearest search engine, I will say that one was released by the band FEAR way back in 1982 and one was on Run the Jewels\u2019 self-titled debut album. The lyrics are a hell of a lot worse than the titles, so maybe wait until you get home from work before you queue them up. Wear headphones, if you\u2019ve got thin walls.\nFor my money, though, the two I can reference by name are the top of that small heap: Tom Waits\u2019 Christmas Card from a Hooker in Minneapolis, and The Pogues\u2019 Fairytale of New York. The former once held the honor of being the only good Christmas song\u2014about which which I was also unequivocally correct, right up until I changed my mind. It\u2019s not the song up for discussion today, but feel free to familiarize yourself just the same\u2014I\u2019ll wait.\nFairytale of New York\u2014the top of the list\u2014starts out by hinting at some pretty standard holiday fare; dreams and cheer and whatnot. Typical seasonal stuff, so long as you ignore that the story seems to be recounted as a drunken flashback in a jail cell. You can probably make a few guesses at the underlying spirit of the song based on that framing: following a lucky break, our bright-eyed protagonists move to New York in search of fame and fortune, only to quickly descend into bad decisions, name-calling, and vaguely festive chaos.\nThis song speaks to me on a couple of levels, not the least of which is as a retelling of my day-to-day interactions with JavaScript. Each day\u2019s melody might vary a little bit, granted, but the lyrics almost always follow a pretty clear arc toward \u201cPARENTAL ADVISORY: EXPLICIT CONTENT.\u201d You might have heard a similar tune yourself; it goes a little somethin\u2019 like setTimeout(function() { console.log( \"this should be happening last\" ); }, 1000); . Callbacks are calling callbacks calling callbacks and something is happening somewhere, as the JavaScript interpreter plods through our code start-to-finish, line-by-line, step-by-step. If we need to take actions based on the results of something that could take its sweet time resolving, well, we\u2019d better fiddle with the order of things to make sure those actions don\u2019t happen too soon.\n\u201cBut I can see a better time,\u201d as the song says, \u201cwhen all our dreams come true.\u201d So, with that Pogues brand of holiday spirit squarely in mind\u2014by which I mean that your humble narrator is almost certainly drunk, and may be incarcerated at the time of publication\u2014gather \u2019round for a story of hope, of hardships, of semi-asynchronous JavaScript programming, and ultimately: of Promise unfulfilled.\nThe Main Thread\nJavaScript is single-minded, in a manner of speaking. Anything we tell the JavaScript runtime to do goes into a single-file queue; you\u2019ll see it referred to as the \u201cmain thread,\u201d or \u201cUI thread.\u201d That thread can be shared by a number of critical browser processes, like rendering and re-rendering parts of the page, and user interactions ranging from the simple\u2014say, highlighting text\u2014to the more complex\u2014interacting with form elements.\nIf that sounds a little scary to you, well, that\u2019s because it is. The more complex our scripts, the more we\u2019re cramming into that single-file main thread, to be processed along with\u2014say\u2014some of our CSS animations. Too much JavaScript clogging up the main thread means a lot of user-facing performance jankiness. Getting away from that single thread is a big part of all the excitement around Web Workers, which allow us to offload entire scripts into their own dedicated background threads\u2014though not without limitations of their own. Outside of Web Workers, that everything-thread is the only game in town: scripts executed one thing at a time, functions calling functions calling functions, taking numbers and crowding up the same deli counter as a user\u2019s interactions\u2014which, in this already strained metaphor, would be ham, I guess?\nAsynchronous JavaScript\nNow, those queued actions may include asynchronous things. For example: AJAX callbacks, setTimeout/setInterval, and addEventListener won\u2019t block the main thread while we\u2019re waiting for a request to come back, a timer to tick away, or an event to trigger. Once those things do kick in, though, the actions they\u2019re meant to perform will get shuffled right back into that single-thread queue.\nThere are a couple of places you might have written asynchronously-fired JavaScript, even if you\u2019re not super familiar with the overarching concept: XMLHttpRequest\u2014\u201cAJAX,\u201d if ya nasty\u2014or just kicking off a function once a user triggers a click or mouseenter event. Event-driven development is writ a little larger, with the overall flow of the script dictated by events, both internal and external. Writing event-driven JavaScript applications is a step in the right direction for sure\u2014it won\u2019t cure what ails the main thread, but it does work with the medium in a reasonable way. Event-driven development allows us to manage our use of the main thread in a way that makes sense. If any of this rings a bell for you, the motivation for Promises should feel familiar.\nFor example, a custom init event might kick things off, and fire a create event that applies our classes and restructures our markup which, on completion, fires a bindEvents event to handle all the event listeners for user interaction. There might not sound like much difference between that and one big function that kicks off, manipulates the DOM, and binds our events line-by-line\u2014but in a script of sufficient size and complexity we\u2019re not only provided with a decoupled flow through the script, but obvious touchpoints for future updates and a predictable structure for ongoing maintenance. \nThis pattern falls apart a little where we were still creating, binding, and listening for events in the same top-to-bottom, one-item-at-a-time way\u2014we had to set a listener on a given object before the event fires, or nothing would happen:\n// Create the event:\nvar event = document.createEvent( \"Event\" );\n\n// Name the event:\nevent.initEvent( \"doTheStuff\", true, true );\n\n// Listen for the custom `doTheStuff` event on `window`:\nwindow.addEventListener( \"doTheStuff\", initializeEverything );\n\n// Fire the custom event\nwindow.dispatchEvent( event );\nThis example is a little contrived, and this stuff is a lot more manageable for sure with the addition of a framework, but that\u2019s the basic gist: create and name the event, add a listener for the event, and\u2014after setting our listener\u2014dispatch the event.\nEvents and callbacks aren\u2019t the only game in town for weaving our way in and out of the main thread, though\u2014at least, not anymore. \nPromises\nA Promise is, at the risk of sounding sentimental, pure potential\u2014an empty container into which a value eventually results. A Promise can exist in several states: \u201cpending,\u201d while the computation they contain is being performed or \u201cresolved\u201d once that computation is complete. Once resolved, a Promise is \u201cfulfilled\u201d if it gave us back something we expect, or \u201crejected\u201d if it didn\u2019t.\nThe Promise constructor accepts a callback with two arguments: resolve and reject. We perform an action\u2014asynchronous or otherwise\u2014within that callback. If everything in there has gone according to plan, we call resolve. If something has gone awry, we call reject\u2014with an error, conventionally. To illustrate, let\u2019s tack something together with a pretty decent chance of doing what we don\u2019t want: a promise meant only to give us the number 1, but has a chance of giving us back a 2. No reasonable person would ever do this, of course, but I wouldn\u2019t necessarily put it past me.\nvar promisedOne = new Promise( function( resolve, reject ) {\n  var coinToss = Math.floor( Math.random() * 2 ) + 1;\n\n  if( coinToss === 1 ) {\n    resolve( coinToss );\n  } else {\n    reject( new Error( \"That ain\u2019t a one.\" ) );\n  }\n});\nThere\u2019s nothing too surprising in there, after you boil it all down. It\u2019s a little return-y, with the exception that we\u2019re flagging results as \u201cas expected\u201d or \u201csomething went wrong.\u201d\nTapping into that Promise uses another new keyword: then\u2014and as someone who attempts to make sense of JavaScript by breaking it down to plain ol\u2019 human-language, I\u2019m a big fan of this syntax. then is tacked onto our Promise identifier, and does just what it says on the tin: once the Promise is resolved, then do one of two things, both supplied as callbacks: the first in the case of a fulfilled promise, and the second in the case of a rejected one. Those two callbacks will have, as arguments, the results we specified with resolve orreject, respectively. It sounds like a lot in prose, but in code it\u2019s a pretty simple pattern:\npromisedOne.then( function( result ) {\n  console.log( result );\n}, function( error ) {\n  console.error( error );\n});\nIf you\u2019ve spent any time working with AJAX\u2014jQuery-wise, in particular\u2014you\u2019ve seen something like this pattern before: a success callback and an error callback. The state of a promise, once fulfilled or rejected, cannot be changed\u2014any reference we make to promisedOne will have a single, fixed result.\nIt may not look like too much the way I\u2019m using it here, but it\u2019s powerful stuff\u2014a pattern for asynchronously resolving anything. I\u2019ve recently used Promises alongside a script that emulates Font Load Events, to apply webfonts asynchronously and avoid a potential performance hit. Font Face Observer allows us to, as the name implies, determine when the files referenced by our @font-face rules have finished loading. \nvar fontObserver = new FontFaceObserver( \"Fancy Font\" );\n\nfontObserver.check().then(function() {\n  document.documentElement.className += \" fonts-loaded\";\n}, function( error ) {\n  console.error( error );\n});\nfontObserver.check() gives us back a Promise, allowing us to chain on a then containing our callbacks for success and failure. We use the fulfilled callback to bolt a class onto the page once the font file has been fully transferred. We don\u2019t bother including an argument in the first function, since we don\u2019t care about the result itself so much as we care that the promise resolved without error\u2014we\u2019re not doing anything with the resolved value, just adding a class to the page. We do include the error argument, since we\u2019ll want to know what happened should something go wrong.\nNow, this isn\u2019t the tidiest syntax around\u2014at least to my eyes\u2014with those two functions just kinda floating in a then. Luckily there\u2019s an similar alternative syntax; one that I find a bit easier to parse at-a-glance:\nfontObserver.check()\n  .then(function() {\n    document.documentElement.className += \" fonts-loaded\";\n  })\n  .catch(function( error ) {\n    console.log( error );\n  });\nThe first callback inside then provides us with our success state, while the catch provides us with a single, explicit \u201csomething went wrong\u201d callback. The two syntaxes aren\u2019t completely identical in all situations, but for a simple case like this, I find it a little neater.\nThe Common Thread\nI guess I still owe you an explanation, huh. Not about the JavaScript-whatever; I think I\u2019ve explained that plenty. No, I mean Fairytale of New York, and why it\u2019s perched up there at the top of the four (4) song heap.\nFairytale is a sad song, ostensibly. If you follow the main thread\u2014start to finish, line-by-line, step by step\u2014 Fairytale is a sad song. And I can see you out there, visions of Die Hard dancing in your heads: \u201cbut is it a Christmas song?\u201d\nWell, for my money, nothing says \u201cholidays\u201d quite like unreliable narration.\nShane MacGowan, the song\u2019s author, has placed the first verse about \u201cChristmas Eve in the drunk tank\u201d as happening right after the \u201clucky one, came in eighteen-to-one\u201d\u2014not at the chronological end of the story. That means the song might not be mostly drunken flashback, but all of it a single, overarching flashback including a Christmas Eve in protective custody. It could be that the man and woman are, together, recounting times long past\u2014good times and bad times\u2014maybe not even in chronological order. Hell, the \u201cNYPD Choir\u201d mentioned in the chorus? There\u2019s no such thing.\nWe\u2019re not big Christmas folks, my family and I. But just the same, every year, the handful of us get together, and every year\u2014like clockwork\u2014there\u2019s a lull in conversation, there\u2019s a sharp exhale, and Ma says \u201cwe all made it.\u201d Not to a house, not to a dinner, but through another year, to another Christmas. At this point, without fail, someone starts telling a story\u2014and one begets another, and so on. Sometimes the stories are happy, sometimes they\u2019re sad, more often than not they\u2019re both. Some are about things we were lucky to walk away from, some are about a time when another one of us didn\u2019t.\nStart-to-finish, line-by-line, step-by-step, the main thread through the year doesn\u2019t change, and maybe there isn\u2019t a whole lot we can do to change it. But by carefully weaving our way in and out of that thread\u2014stories all out of sync and resolving one way or the other, with the results determined by questionably reliable narrators\u2014we can change the way we interact with it and, little by little, we can start making sense of it.", "year": "2016", "author": "Mat Marquis", "author_slug": "matmarquis", "published": "2016-12-19T00:00:00+00:00", "url": "https://24ways.org/2016/fairytale-of-new-promise/", "topic": "code"}
{"rowid": 309, "title": "HTTP/2 Server Push and Service Workers: The Perfect Partnership", "contents": "Being a web developer today is exciting! The web has come a long way since its early days and there are so many great technologies that enable us to build faster, better experiences for our users. One of these technologies is HTTP/2 which has a killer feature known as HTTP/2 Server Push.\nDuring this year\u2019s Chrome Developer Summit, I watched a really informative talk by Sam Saccone, a Software Engineer on the Google Chrome team. He gave a talk entitled Planning for Performance, and one of the topics that he covered immediately piqued my interest; the idea that HTTP/2 Server Push and Service Workers were the perfect web performance combination.\n\nIf you\u2019ve never heard of HTTP/2 Server Push before, fear not - it\u2019s not as scary as it sounds. HTTP/2 Server Push simply allows the server to send data to the browser without having to wait for the browser to explicitly request it first. In this article, I am going to run through the basics of HTTP/2 Server Push and show you how, when combined with Service Workers, you can deliver the ultimate in web performance to your users.\nWhat is HTTP/2 Server Push?\nWhen a user navigates to a URL, a browser will make an HTTP request for the underlying web page. The browser will then scan the contents of the HTML document for any assets that it may need to retrieve such as CSS, JavaScript or images. Once it finds any assets that it needs, it will then make multiple HTTP requests for each resource that it needs and begin downloading one by one. While this approach works well, the problem is that each HTTP request means more round trips to the server before any data arrives at the browser. These extra round trips take time and can make your web pages load slower. \nBefore we go any further, let\u2019s see what this might look like when your browser makes a request for a web page. If you were to view this in the developer tools of your browser, it might look a little something like this:\n\nAs you can see from the image above, once the HTML file has been downloaded and parsed, the browser then makes HTTP requests for any assets that it needs. \nThis is where HTTP/2 Server Push comes in. The idea behind HTTP/2 Server Push is that when the browser requests a web page from the server, the server already knows about all the assets that are needed for the web page and \u201cpushes\u201d it to browser. This happens when the first HTTP request for the web page takes place and it eliminates an extra round trip, making your site faster. \nUsing the same example above, let\u2019s \u201cpush\u201d the JavaScript and CSS files instead of waiting for the browser to request them. The image below gives you an idea of what this might look like.\n\nWhoa, that looks different - let\u2019s break it down a little. Firstly, you can see that the JavaScript and CSS files appear earlier in the waterfall chart. You might also notice that the loading times for the files are extremely quick. The browser doesn\u2019t need to make an extra HTTP request to the server, instead it receives the critical files it needs all at once. Much better! \nThere are a number of different approaches when it comes to implementing HTTP/2 Server Push. Adoption is growing and many commercial CDNs such as Akamai and Cloudflare already offer support for Server Push. You can even roll your own implementation depending on your environment. I\u2019ve also previously blogged about building a basic HTTP/2 Server Push example using Node.js. In this post, I\u2019m not going to dive into how to implement HTTP/2 Server Push as that is an entire post in itself! However, I do recommend reading this article to find out more about the inner workings.\nHTTP/2 Server Push is awesome, but it isn\u2019t a magic bullet. It is fantastic for improving the load time of a web page when it first loads for a user, but it isn\u2019t that great when they request the same web page again. The reason for this is that HTTP/2 Server Push is not cache \u201caware\u201d. This means that the server isn\u2019t aware about the state of your client. If you\u2019ve visited a web page before, the server isn\u2019t aware of this and will push the resource again anyway, regardless of whether or not you need it. HTTP/2 Server Push effectively tells the browser that it knows better and that the browser should receive the resources whether it needs them or not. In theory browsers can cancel HTTP/2 Server Push requests if they\u2019re already got something in cache but unfortunately no browsers currently support it. The other issue is that the server will have already started to send some of the resource to the browser by the time the cancellation occurs.\nHTTP/2 Server Push & Service Workers\nSo where do Service Workers fit in? Believe it or not, when combined together HTTP/2 Server Push and Service Workers can be the perfect web performance partnership. If you\u2019ve not heard of Service Workers before, they are worker scripts that run in the background of your website. Simply put, they act as middleman between the client and the browser and enable you to intercept any network requests that come and go from the browser. They are packed with useful features such as caching, push notifications, and background sync. Best of all, they are written in JavaScript, making it easy for web developers to understand.\nUsing Service Workers, you can easily cache assets on a user\u2019s device. This means when a browser makes an HTTP request for an asset, the Service Worker is able to intercept the request and first check if the asset already exists in cache on the users device. If it does, then it can simply return and serve them directly from the device instead of ever hitting the server.\nLet\u2019s stop for a second and analyse what that means. Using HTTP/2 Server Push, you are able to push critical assets to the browser before the browser requests them. Then, using Service Workers you are able to cache these resources so that the browser never needs to make a request to the server again. That means a super fast first load and an even faster second load!\nLet\u2019s put this into action. The following HTML code is a basic web page that retrieves a few images and two JavaScript files.\n<!DOCTYPE html>\n<html>\n<head>\n  <meta charset=\"UTF-8\">\n  <title>HTTP2 Push Demo</title>\n</head>\n<body>\n  <h1>HTTP2 Push</h1>\n  <img src=\"./images/beer-1.png\" width=\"200\" height=\"200\" />\n  <img src=\"./images/beer-2.png\" width=\"200\" height=\"200\" />\n  <br>\n  <br>\n  <img src=\"./images/beer-3.png\" width=\"200\" height=\"200\" />\n  <img src=\"./images/beer-4.png\" width=\"200\" height=\"200\" />\n  <!-- Scripts -->\n  <script async src=\"./js/promise.min.js\"></script>\n  <script async src=\"./js/fetch.js\"></script>\n  <script>\n  // Register the service worker\n  if ('serviceWorker' in navigator) {\n    navigator.serviceWorker.register('./service-worker.js').then(function(registration) {\n      // Registration was successful\n      console.log('ServiceWorker registration successful with scope: ', registration.scope);\n    }).catch(function(err) {\n      // registration failed :(\n      console.log('ServiceWorker registration failed: ', err);\n    });\n  }\n  </script>\n</body>\n</html>\nIn the HTML code above, I am registering a Service Worker file named service-worker.js. In order to start caching assets, I am going to use the Service Worker toolbox . It is a lightweight helper library to help you get started creating your own Service Workers. Using this library, we can actually cache the base web page with the path /push.\nThe Service Worker Toolbox comes with a built-in routing system which is based on the same routing as Express. With just a few lines of code, you can start building powerful caching patterns.\nI\u2019ve add the following code to the service-worker.js file.\n(global => {\n  'use strict';\n\n  // Load the sw-toolbox library.\n  importScripts('/js/sw-toolbox/sw-toolbox.js');\n\n  // The route for any requests\n  toolbox.router.get('/push', global.toolbox.fastest);\n\n  toolbox.router.get('/images/(.*)', global.toolbox.fastest);\n\n  toolbox.router.get('/js/(.*)', global.toolbox.fastest);\n\n  // Ensure that our service worker takes control of the page as soon as possible.\n  global.addEventListener('install', event => event.waitUntil(global.skipWaiting()));\n  global.addEventListener('activate', event => event.waitUntil(global.clients.claim()));\n})(self);\nLet\u2019s break this code down further. Around line 4, I am importing the Service Worker toolbox. Next, I am specifying a route that will listen to any requests that match the URL /push. Because I am also interested in caching the images and JavaScript for that page, I\u2019ve told the toolbox to listen to these routes too.\nThe best thing about the code above is that if any of the assets exist in cache, we will instantly return the cached version instead of waiting for it to download. If the asset doesn\u2019t exist in cache, the code above will add it into cache so that we can retrieve it when it\u2019s needed again.\nYou may also notice the code global.toolbox.fastest - this is important because gives you the compromise of fulfilling from the cache immediately, while firing off an additional HTTP request updating the cache for the next visit.\nBut what does this mean when combined with HTTP/2 Server Push? Well, it means that on the first load of the web page, you are able to \u201cpush\u201d everything to the user at once before the browser has even requested it. The Service Worker activates and starts caching the assets on the users device. The next time a user visits the page, the Service Worker will intercept the request and serve the asset directly from cache. Amazing! \nUsing this technique, the waterfall chart for a repeat visit should look like the image below.\n\nIf you look closely at the image above, you\u2019ll notice that the web page returns almost instantly without ever making an HTTP request over the network. Using the Service Worker library, we cached the base page for the route /push, which allowed us to retrieve this directly from cache.\nWhether used on their own or combined together, the best thing about these two features is that they are the perfect progressive enhancement. If your user\u2019s browser doesn\u2019t support them, they will simply fall back to HTTP/1.1 without Service Workers. Your users may not experience as fast a load time as they would with these two techniques, but it would be no different from their normal experience. HTTP/2 Server Push and Service Workers are really the perfect partners when it comes to web performance.\nSummary\nWhen used correctly, HTTP/2 Server Push and Service Workers can have a positive impact on your site\u2019s load times. Together they mean super fast first load times and even faster repeat views to a web page. Whilst this technique is really effective, it\u2019s worth noting that HTTP/2 push is not a magic bullet. Think about the situations where it might make sense to use it and don\u2019t just simply \u201cpush\u201d everything; it could actually lead to having slower page load times. If you\u2019d like to learn more about the rules of thumb for HTTP/2 Server Push, I recommend reading this article for more information. \nAll of the code in this example is available on my Github repo - if you have any questions, please submit an issue and I\u2019ll get back to you as soon as possible. \nIf you\u2019d like to learn more about this technique and others relating to HTTP/2, I highly recommend watching Sam Saccone\u2019s talk at this years Chrome Developer Summit. \nI\u2019d also like to say a massive thank you to Robin Osborne, Andy Davies and Jeffrey Posnick for helping me review this article before putting it live!", "year": "2016", "author": "Dean Hume", "author_slug": "deanhume", "published": "2016-12-15T00:00:00+00:00", "url": "https://24ways.org/2016/http2-server-push-and-service-workers/", "topic": "code"}
{"rowid": 308, "title": "How to Make a Chrome Extension to Delight (or Troll) Your Friends", "contents": "If you\u2019re like me, you grew up drawing mustaches on celebrities. Every photograph was subject to your doodling wrath, and your brilliance was taken to a whole new level with computer programs like Microsoft Paint. The advent of digital cameras meant that no one was safe from your handiwork, especially not your friends. And when you finally got your hands on Photoshop, you spent hours maniacally giggling at your artistic genius. \nBut today is different. You\u2019re a serious adult with important things to do and a reputation to uphold. You keep up with modern web techniques and trends, and have little time for fun other than a random Giphy on Slack\u2026 right? \nNope. \nIf there\u2019s one thing 2016 has taught me, it\u2019s that we\u2014the self-serious, world-changing tech movers and shakers of the universe\u2014haven\u2019t changed one bit from our younger, more delightable selves.\nHow do I know? This year I created a Chrome extension called Tabby Cat and watched hundreds of thousands of people ditch productivity for randomly generated cats. Tabby Cat replaces your new tab page with an SVG cat featuring a silly name like \u201cStinky Dinosaur\u201d or \u201cTiny Potato\u201d. Over time, the cats collect goodies that vary in absurdity from fishbones to lawn flamingos to Raybans. Kids and adults alike use this extension, and analytics show the majority of use happens Monday through Friday from 9-5. The popularity of Tabby Cat has convinced me there\u2019s still plenty of room in our big, grown-up hearts for fun.\n\n\nToday, we\u2019re going to combine the formula behind Tabby Cat with your intrinsic desire to delight (or troll) your friends, and create a web app that generates your friends with random objects and environments of your choosing. You can publish it as a Chrome extension to replace your new tab, or simply host it as a website and point to it with the New Tab Redirect extension. \nHere\u2019s a sneak peek at my final result featuring my partner, my cat, and I in cheerfully weird accessories. Your result will look however you want it to.\n\nAlong the way, we\u2019ll cover how to build a Chrome extension that replaces the new tab page, and explore ways to program randomness into your work to create something truly delightful. \nWhat you\u2019ll need\n\nAdobe Illustrator (or a similar illustration program to export PNG)\nSome images of your friends\nA text editor\n\nNote: This can be as simple or as complex as you want it to be. Most of the application is pre-built so you can focus on kicking back and getting in touch with your creative side. If you want to dive in deeper, you\u2019ll find ways to do it.\nGetting started\n\nDownload a local copy of the boilerplate for today\u2019s tutorial here, and open it in a text editor. Inside, you\u2019ll find a simple web app that you can run in Chrome. \nOpen index.html in Chrome. You should see a grey page that says \u201cNoname\u201d.\nOpen template.pdf in Adobe Illustrator or a similar program that can export PNG. The file contains an artboard measuring 800px x 800px, with a dotted blue outline of a face. This is your template.\n\nNote: We\u2019re using Google Chrome to build and preview this application because the end-result is a Chrome extension. This means that the application isn\u2019t totally cross-browser compatible, but that\u2019s okay.\nStep 1: Gather your friends\nThe first thing to do is choose who your muses are. Since the holidays are upon us, I\u2019d suggest finding inspiration in your family.\nCreate your artwork\nFor each person, find an image where their face is pointed as forward as possible. Place the image onto the Artwork layer of the Illustrator file, and line up their face with the template. Then, rename the artboard something descriptive like face_bob. Here\u2019s my crew:\nAs you can see, my use of the word \u201cfamily\u201d extends to cats. There\u2019s no judgement here.\nNotice that some of my photos don\u2019t completely fill the artboard\u2013that\u2019s fine. The images will be clipped into ovals when they\u2019re rendered in the application.\nNow, export your images by following these steps:\n\nTurn the Template layer off and export the images as PNGs. \nIn the Export dialog, tick the \u201cUse Artboards\u201d checkbox and enter the range with your faces. \nExport at 72ppi to keep things running fast. \nSave your images into the images/ folder in your project.\n\nAdd your images to config.js\nOpen scripts/config.js. This is where you configure your extension. \nAdd key value pairs to the faces object. The key should be the person\u2019s name, and the value should be the filepath to the image.\nfaces: {\n    leslie: 'images/face_leslie.png',\n    kyle: 'images/face_kyle.png',\n    beep: 'images/face_beep.png'\n}\nThe application will choose one of these options at random each time you open a new tab. This pattern is used for everything in the config file. You give the application groups of choices, and it chooses one at random each time it loads. The only thing that\u2019s special about the faces object is that person\u2019s name will also be displayed when their face is chosen.\nNow, when you refresh the project in Chrome, you should see one of your friends along with their name, like this:\n\nCongrats, you\u2019re off and running!\nStep 2: Add adjectives\nNow that you\u2019ve loaded your friends into the application, it\u2019s time to call them names. This step definitely yields the most laughs for the least amount of effort.\nAdd a list of adjectives into the prefixes array in config.js. To get the words flowing, I took inspiration from ways I might describe some of my relatives during a holiday gathering\u2026\nprefixes: [\n    'Loving',\n    'Drunk',\n    'Chatty',\n    'Merry',\n    'Creepy',\n    'Introspective',\n    'Cheerful',\n    'Awkward',\n    'Unrelatable',\n    'Hungry',\n    ...\n]\nWhen you refresh Chrome, you should see one of these words prefixed before your friend\u2019s name. Voila!\n\nStep 3: Choose your color palette\nReal talk: I\u2019m bad at choosing color palettes, so I have a trick up my sleeve that I want to share with you. If you\u2019ve been blessed with the gift of color aptitude, skip ahead.\nHow to choose colors\nTo create a color palette, I start by going to a Coolors.co, and I hit the spacebar until I find a palette that I like. We need a wide gamut of hues for our palette, so lock down colors you like and keep hitting the spacebar until you find a nice, full range. You can use as many or as few colors as you like.\nCopy these colors into your swatches in Adobe Illustrator. They\u2019ll be the base for any illustrations you create later.\nNow you need a set of background colors. Here\u2019s my trick to making these consistent with your illustration palette without completely blending in. Use the \u201cAdjust Palette\u201d tool in Coolors to dial up the brightness a few notches, and the saturation down just a tad to remove any neon effect. These will be your background colors.\n\nAdd your background colors to config.js\nCopy your hex codes into the bgColors array in config.js.\nbgColors: [\n    '#FFDD77',\n    '#FF8E72',\n    '#ED5E84',\n    '#4CE0B3',\n    '#9893DA',\n    ...\n]\nNow when you go back to Chrome and refresh the page, you\u2019ll see your new palette!\n\nStep 4: Accessorize\nThis is the fun part. We\u2019re going to illustrate objects, accessories, lizards\u2014whatever you want\u2014and layer them on top of your friends.\nYour objects will be categorized into groups, and one option from each group will be randomly chosen each time you load the page. Think of a group like \u201chats\u201d or \u201cglasses\u201d. This will allow combinations of accessories to show at once, without showing two of the same type on the same person.\nCreate a group of accessories\nTo get started, open up Illustrator and create a new artboard out of the template. Think of a group of objects that you can riff on. I found hats to be a good place to start. If you don\u2019t feel like illustrating, you can use cut-out images instead.\n\nNext, follow the same steps as you did when you exported the faces. Here they are again:\n\nTurn the Template layer off and export the images as PNGs. \nIn the Export dialog, tick the \u201cUse Artboards\u201d checkbox and enter the range with your hats. \nExport at 72ppi to keep things running fast. \nSave your images into the images/ folder in your project.\n\nAdd your accessories to config.js\nIn config.js, add a new key to the customProps object that describes the group of accessories that you just created. Its value should be an array of the filepaths to your images. This is my hats array:\ncustomProps: {\n    hats: [     \n        'images/hat_crown.png',\n        'images/hat_santa.png',\n        'images/hat_tophat.png',\n        'images/hat_antlers.png'\n    ]\n}\nRefresh Chrome and behold, accessories!\n\nCreate as many more accessories as you want\nRepeat the steps above to create as many groups of accessories as you want. I went on to make glasses and hairstyles, so my final illustrator file looks like this:\n\nThe last step is adding your new groups to the config object. List your groups in the order that you want them to be stacked in the DOM. My final output will be hair, then hats, then glasses:\ncustomProps: {\n    hair: [     \n        'images/hair_bowl.png',\n        'images/hair_bob.png'\n    ],  \n    hats: [     \n        'images/hat_crown.png',\n        'images/hat_santa.png',\n        'images/hat_tophat.png',\n        'images/hat_antlers.png'\n    ],\n    glasses: [\n        'images/glasses_aviators.png',\n        'images/glasses_monacle.png'\n    ]\n}\nAnd, there you have it! Randomly generated friends with random accessories. \n\nFeel free to go much crazier than I did. I considered adding a whole group of animals in celebration of the new season of Planet Earth, or even adding Sir David Attenborough himself, or doing a bit of role reversal and featuring the animals with little safari hats! But I digress\u2026\nStep 5: Publish it\nIt\u2019s time to put this in your new tabs! You have two options:\n\nPublish it as a Chrome extension in the Chrome Web Store.\nHost it as a website and point to it with the New Tab Redirect extension.\n\nToday, we\u2019re going to cover Option #1 because I want to show you how to make the simplest Chrome extension possible. However, I recommend Option #2 if you want to keep your project private. Every Chrome extension that you publish is made publicly available, so unless your friends want their faces published to an extension that anyone can use, I\u2019d suggest sticking to Option #2.\nHow to make a simple Chrome extension to replace the new tab page\nAll you need to do to make your project into a Chrome extension is add a manifest.json file to the root of your project with the following contents. There are plenty of other properties that you can add to your manifest file, but these are the only ones that are required for a new tab replacement:\n{\n    \"manifest_version\": 2,\n    \"name\": \"Your extension name\",\n    \"version\": \"1.0\",\n    \"chrome_url_overrides\" : {\n        \"newtab\": \"index.html\"\n    }\n}\nTo test your extension, you\u2019ll need to run it in Developer Mode. Here\u2019s how to do that:\n\nGo to the Extensions page in Chrome by navigating to chrome://extensions/.\nTick the checkbox in the upper-right corner labelled \u201cDeveloper Mode\u201d.\nClick \u201cLoad unpacked extension\u2026\u201d and select this project.\nIf everything is running smoothly, you should see your project when you open a new tab. If there are any errors, they should appear in a yellow box on the Extensions page.\n\nVoila! Like I said, this is a very light example of a Chrome extension, but Google has tons of great documentation on how to take things further. Check it out and see what inspires you.\nShare the love\nNow that you know how to make a new tab extension, go forth and create! But wield your power responsibly. New tabs are opened so often that they\u2019ve become a part of everyday life\u2013just consider how many tabs you opened today. Some people prefer to-do lists in their tabs, and others prefer cats. \nAt the end of the day, let\u2019s make something that makes us happy. Cheers!", "year": "2016", "author": "Leslie Zacharkow", "author_slug": "lesliezacharkow", "published": "2016-12-08T00:00:00+00:00", "url": "https://24ways.org/2016/how-to-make-a-chrome-extension/", "topic": "code"}
{"rowid": 307, "title": "Get the Balance Right: Responsive Display Text", "contents": "Last year in 24 ways I urged you to Get Expressive with Your Typography. I made the case for grabbing your readers\u2019 attention by setting text at display sizes, that is to say big. You should consider very large text in the same way you might a hero image: a picture that creates an atmosphere and anchors your layout.\nWhen setting text to be read, it is best practice to choose body and subheading sizes from a pre-defined scale appropriate to the viewport dimensions. We set those sizes using rems, locking the text sizes together so they all scale according to the page default and your reader\u2019s preferences.  You can take the same approach with display text by choosing larger sizes from the same scale.\nHowever, display text, as defined by its purpose and relative size, is text to be seen first, and read second. In other words a picture of text. When it comes to pictures, you are likely to scale all scene-setting imagery - cover photos, hero images, and so on - relative to the viewport. Take the same approach with display text: lock the size and shape of the text to the screen or browser window.\nIntroducing viewport units\nWith CSS3 came a new set of units which are locked to the viewport. You can use these viewport units wherever you might otherwise use any other unit of length such as pixels, ems or percentage. There are four viewport units, and in each case a value of 1 is equal to 1% of either the viewport width or height as reported in reference1 pixels:\n\nvw - viewport width,\nvh - viewport height,\nvmin - viewport height or width, whichever is smaller\nvmax - viewport height or width, whichever is larger\n\nIn one fell swoop you can set the size of a display heading to be proportional to the screen or browser width, rather than choosing from a scale in a series of media queries. The following makes the heading font size 13% of the viewport width:\nh1 {\n    font-size: 13 vw;\n}\nSo for a selection of widths, the rendered font size would be:\nRendered font size (px)\nViewport width\n13\u202fvw\n320\n42\n768\n100\n1024\n133\n1280\n166\n1920\n250\n\nA problem with using vw in this manner is the difference in text block proportions between portrait and landscape devices. Because the font size is based on the viewport width, the text on a landscape display is far bigger than when rendered on the same device held in a portrait orientation. \nLandscape text is much bigger than portrait text when using vw units.\nThe proportions of the display text relative to the screen are so dissimilar that each orientation has its own different character, losing the inconsistency and considered design you would want when designing to make an impression.\nHowever if the text was the same size in both orientations, the visual effect would be much more consistent. This where vmin comes into its own. Set the font size using vmin and the size is now set as a proportion of the smallest side of the viewport, giving you a far more consistent rendering.\nh1 {\n    font-size: 13vmin;\n}\nLandscape text is consistent with portrait text when using vmin units.\nComparing vw and vmin renderings for various common screen dimensions, you can see how using vmin keeps the text size down to a usable magnitude:\nRendered font size (px)\nViewport\n13\u202fvw\n13\u202fvmin\n320 \u00d7 480\n42\n42\n414 \u00d7 736\n54\n54\n768 \u00d7 1024\n100\n100\n1024 \u00d7 768\n133\n100\n1280 \u00d7 720\n166\n94\n1366 \u00d7 768\n178\n100\n1440 \u00d7 900\n187\n117\n1680 \u00d7 1050\n218\n137\n1920 \u00d7 1080\n250\n140\n2560 \u00d7 1440\n333\n187\n\nHybrid font sizing\nUsing vertical media queries to set text in direct proportion to screen dimensions works well when sizing display text. In can be less desirable when sizing supporting text such as sub-headings, which you may not want to scale upwards at the same rate as the display text. For example, we can size a subheading using vmin so that it starts at 16 px on smaller screens and scales up in the same way as the main heading:\nh1 {\n    font-size: 13vmin;\n}\nh2 {\n    font-size: 5vmin;\n}\nUsing vmin alone for supporting text can scale it too quickly\nThe balance of display text to supporting text on the phone works well, but the subheading text on the tablet, even though it has been increased in line with the main heading, is starting to feel disproportionately large and a little clumsy. This problem becomes magnified on even bigger screens.\nA solution to this is use a hybrid method of sizing text2. We can use the CSS calc() function to calculate a font size simultaneously based on both rems and viewport units. For example:\nh2 {\n    font-size: calc(0.5rem + 2.5vmin);\n}\nFor a 320 px wide screen, the font size will be 16 px, calculated as follows:\n(0.5 \u00d7 16) + (320 \u00d7 0.025) = 8 + 8 = 16px\nFor a 768 px wide screen, the font size will be 27 px:\n(0.5 \u00d7 16) + (768 \u00d7 0.025) = 8 + 19 = 27px\nThis results in a more balanced subheading that doesn\u2019t take emphasis away from the main heading:\n\nTo give you an idea of the effect of using a hybrid approach, here\u2019s a side-by-side comparison of hybrid and viewport text sizing:\ntable.ex--scale{width:100%;overflow: hidden;} table.ex--scale td{vertical-align:baseline;text-align:center;padding:0} tr.ex--scale-key{color:#666} tr.ex--scale-key td{font-size:.875rem;padding:0 0.125em} .ex--scale-2 tr.ex--scale-size{color:#ccc} tr.ex--scale-size td{font-size:1em;line-height:.34em;padding-bottom:.5rem} td.ex--scale-step{color:#000} td.ex--scale-hilite{color:red} .ex--scale-3 tr.ex--scale-size td{line-height:.9em}\n\ntop: calc() hybrid method; bottom: vmin only\n16\n20\n27\n32\n35\n40\n44\n16\n24\n38\n48\n54\n64\n72\n320\n480\n768\n960\n1080\n1280\n1440\n\nOver this festive period, try experiment with the proportion of rem and vmin in your hybrid calculation to see what feels best for your particular setting.\n\n\n\n\nA reference pixel is based on the logical resolution of a device which takes into account double density screens such as Retina displays.\u00a0\u21a9\ufe0e\n\n\nFor even more sophisticated uses of hybrid text sizing see the work of Mike Riethmuller.\u00a0\u21a9\ufe0e", "year": "2016", "author": "Richard Rutter", "author_slug": "richardrutter", "published": "2016-12-09T00:00:00+00:00", "url": "https://24ways.org/2016/responsive-display-text/", "topic": "code"}
{"rowid": 306, "title": "What next for CSS Grid Layout?", "contents": "In 2012 I wrote an article for 24 ways detailing a new CSS Specification that had caught my eye, at the time with an implementation only in Internet Explorer. What I didn\u2019t realise at the time was that CSS Grid Layout was to become a theme on which I would base the next four years of research, experimentation, writing and speaking. \nAs I write this article in December 2016, we are looking forward to CSS Grid Layout being shipped in Chrome and Firefox. What will ship early next year in those browsers is expanded and improved from the early implementation I explored in 2012. Over the last four years the spec has been developed as part of the CSS Working Group process, and has had input from browser engineers, specification writers and web developers. Use cases have been discussed, and features added.\nThe CSS Grid Layout specification is now a Candidate Recommendation. This status means the spec is to all intents and purposes, finished. The discussions now happening are on fine implementation details, and not new feature ideas. It makes sense to draw a line under a specification in order that browser vendors can ship complete, interoperable implementations. That approach is good for all of us, it makes development far easier if we know that a browser supports all of the features of a specification, rather than working out which bits are supported. However it doesn\u2019t mean that works stops here, and that new use cases and features can\u2019t be proposed for future levels of Grid Layout. Therefore, in this article I\u2019m going to take a look at some of the things I think grid layout could do in the future. I would love for these thoughts to prompt you to think about how Grid - or any CSS specification - could better suit the use cases you have.\nSubgrid - the missing feature of Level 1\nThe implementation of CSS Grid Layout in Chrome, Firefox and Webkit is comparable and very feature complete. There is however one standout feature that has not been implemented in any browser as yet - subgrid. Once you set the value of the display property to grid, any direct children of that element become grid items. This is similar to the way that flexbox behaves, set display: flex and all direct children become flex items. The behaviour does not apply to children of those items. You can nest grids, just as you can nest flex containers, but the child grids have no relationship to the parent.\n\nNesting Grids by Rachel Andrew (@rachelandrew) on CodePen.\nThe subgrid behaviour would enable the grid defined on the parent to be used by the children. I feel this would be most useful when working with a multiple column flexible grid - for example a typical 12 column grid. I could define a grid on a wrapper, then position UI elements on that grid - from the major structural elements of my page down through the child elements to a form where I wanted the field to line up with items above.\nThe specification contained an initial description of subgrid, with a value of subgrid for grid-template-columns and grid-template-rows, you can read about this in the August 2015 Working Draft. This version of the specification would have meant you could declare a subgrid in one dimension only, and create a different set of tracks in the other.\nIn an attempt to get some implementation of subgrid, a revised specification was proposed earlier this year. This gives a single subgrid value of the display property. As we now cannot specify a subgrid on rows OR columns this limits us to have a subgrid that works in two dimensions. At this point neither version has been implemented by anyone, and subgrids are marked as \u201cat risk\u201d in the Level 1 Candidate Recommendation. With regard to \u2018at-risk\u2019 this is explained as follows:\n\n\u201c\u2018At-risk\u2019 is a W3C Process term-of-art, and does not necessarily imply that the feature is in danger of being dropped or delayed. It means that the WG believes the feature may have difficulty being interoperably implemented in a timely manner, and marking it as such allows the WG to drop the feature if necessary when transitioning to the Proposed Rec stage, without having to publish a new Candidate Rec without the feature first.\u201d \n\nIf we lose subgrid from Level 1, as it looks likely that we will, this does give us a chance to further discuss and iterate on that feature. My current thoughts are that I\u2019m not completely happy about subgrids being tied to both dimensions and feel that a return to the earlier version, or something like it, would be preferable. \nFurther reading about subgrid\n\nMy post from 2015 detailing why I feel subgrid is important\nMy post based on the revised specification\nEric Meyer\u2019s thoughts on subgrid\nWrite-up of a discussion from Igalia who work on the Blink and Webkit browser implementations\n\nStyling cells, tracks and areas\nHaving defined a grid with CSS Grid Layout you can place child elements into that grid, however what you can\u2019t do is style the grid tracks or cells. Grid doesn\u2019t even go as far as multiple column layout, which has the column-rule properties.\nIn order to set a background colour on a grid cell at the moment you would have to add an empty HTML element or insert some generated content as in the below example. I\u2019m using a 1 pixel grid gap to fake lines between grid cells, and empty div elements, and some generated content to colour those cells.\n\nFaked backgrounds and borders by Rachel Andrew (@rachelandrew) on CodePen.\nI think it would be a nice addition to Grid Layout to be able to directly add backgrounds and borders to cells, tracks and areas. There is an Issue raised in the CSS WG Drafts repository for Decorative Grid Cell pseudo-elements, if you want to add thoughts to that.\nMore control over auto placement\nIf you haven\u2019t explicitly placed the direct children of your grid element they will be laid out according to the grid auto placement rules. You can see in this example how we have created a grid and the items are placing themselves into cells on that grid.\n\nItems auto-place on a defined grid by Rachel Andrew (@rachelandrew) on CodePen.\nThe auto-placement algorithm is very cool. We can position some items, leaving others to auto-place; we can set items to span more than one track; we can use the grid-auto-flow property with a value of dense to backfill gaps in our grid.\n\nWebsafe colors meet CSS Grid (auto-placement demo) by Rachel Andrew (@rachelandrew) on CodePen.\nI think however this could be taken further. In this issue posted to my CSS Grid AMA on GitHub, the question is raised as to whether it would be possible to ask grid to place items on the next available line of a certain name. This would allow you to skip tracks in the grid when using auto-placement, an issue that has also been raised by Emil Bj\u00f6rklund in this post to the www-style list prior to spec discussion moving to Github. I think there are probably similar issues, if you can think of one add a comment here.\nCreating non-rectangular grid areas\nA grid area is a collection of grid cells, defined by setting the start and end lines for columns and rows or by creating the area in the value of the grid-template-areas property as shown below. Those areas however must be rectangular - you can\u2019t create an L-shaped or otherwise non-regular shape.\n\nGrid Areas by Rachel Andrew (@rachelandrew) on CodePen.\n\nPerhaps in the future we could define an L-shape or other non-rectangular area into which content could flow, as in the below currently invalid code where a quote is embedded into an L-shaped content area.\n.wrapper {\n  display: grid;\n  grid-template-areas:\n      \"sidebar header  header\"\n      \"sidebar content quote\"\n      \"sidebar content content\";\n}\nFlowing content through grid cells or areas\nSome uses cases I have seen perhaps are not best solved by grid layout at all, but would involve grid working alongside other CSS specifications. As I detail in this post, there are a class of problems that I believe could be solved with the CSS Regions specification, or a revised version of that spec.\nBeing able to create a grid layout, then flow content through the areas could be very useful. Jen Simmons presented to the CSS Working Group at the Lisbon meeting a suggestion as to how this might work.\nIn a post from earlier this year I looked at a collection of ideas from specifications that include Grid, Regions and Exclusions. These working notes from my own explorations might prompt ideas of your own.\nSolving the keyboard/layout disconnect\nOne issue that grid, and flexbox to a lesser extent, raises is that it is very easy to end up with a layout that is disconnected from the underlying markup. This raises problems for people navigating using the keyboard as when tabbing around the document you find yourself jumping to unexpected places. The problem is explained by L\u00e9onie Watson with reference to flexbox in Flexbox and the keyboard navigation disconnect.\nThe grid layout specification currently warns against creating such a disconnect, however I think it will take careful work by web developers in order to prevent this. It\u2019s also not always as straightforward as it seems. In some cases you want the logical order to follow the source, and others it would make more sense to follow the visual. People are thinking about this issue, as you can read in this mailing list discussion.\nBringing your ideas to the future of Grid Layout\nWhen I\u2019m not getting excited about new CSS features, my day job involves working on a software product - the CMS that is serving this very website, Perch. When we launched Perch there were many use cases that we had never thought of, despite having a good idea of what might be needed in a CMS and thinking through lots of use cases. The additional use cases brought to our attention by our customers and potential customers informed the development of the product from launch. The same will be true for Grid Layout.\nAs a \u201cproduct\u201d grid has been well thought through by many people. Yet however hard we try there will be use cases we just didn\u2019t think of. You may well have one in mind right now. That\u2019s ok, because as with any CSS specification, once Level One of grid  is complete, work can begin on Level Two. The feature set of Level Two will be informed by the use cases that emerge as people get to grips with what we have now.\nThis is where you get to contribute to the future of layout on the web. When you hit up against the things you cannot do, don\u2019t just mutter about how the CSS Working Group don\u2019t listen to regular developers and code around the problem. Instead, take a few minutes and write up your use case. Post it to your blog, to Medium, create a CodePen and go to the CSS Working Group GitHub specs repository and post an issue there. Write some pseudo-code, draw a picture, just make sure that the use case is described in enough detail that someone can see what problem you want grid to solve. It may be that - as with any software development - your use case can\u2019t be solved in exactly the way you suggest. However once we have a use case, collected with other use cases, methods of addressing that class of problems can be investigated. \nI opened this article by explaining I\u2019d written about grid layout four years ago, and how we\u2019re only now at a point where we will have Grid Layout available in the majority of browsers. Specification development, and implementation into browsers takes time. This is actually a good thing, as it\u2019s impossible to take back CSS once it is out there and being used by production websites. We want CSS in the wild to be well thought through and that takes time. So don\u2019t feel that because you don\u2019t see your use case added to a spec immediately it has been ignored. Do your future self a favour and write down your frustrations or thoughts, and we can all make sure that the web platform serves the use cases we\u2019re dealing with now and in the future.", "year": "2016", "author": "Rachel Andrew", "author_slug": "rachelandrew", "published": "2016-12-12T00:00:00+00:00", "url": "https://24ways.org/2016/what-next-for-css-grid-layout/", "topic": "code"}
{"rowid": 305, "title": "CSS Writing Modes", "contents": "Since you may not have a lot of time, I\u2019m going to start at the end, with the dessert.\nYou can use a little-known, yet important and powerful CSS property to make text run vertically. Like this.\n\nOr instead of running text vertically, you can layout a set of icons or interface buttons in this way. Or, of course, with anything on your page.  \nThe CSS I\u2019ve applied makes the browser rethink the orientation of the world, and flow the layout of this element at a 90\u00b0 angle to \u201cnormal\u201d. Check out the live demo, highlight the headline, and see how the cursor is now sideways.\nSee the Pen Writing Mode Demo \u2014 Headline by Jen Simmons (@jensimmons) on CodePen.\n\nThe code for accomplishing this is pretty simple. \nh1 { \n  writing-mode: vertical-rl;\n}\nThat\u2019s all it takes to switch the writing mode from the web\u2019s default horizontal top-to-bottom mode to a vertical right-to-left mode. If you apply such code to the html element, the entire page is switched, affecting the scroll direction, too. \nIn my example above, I\u2019m telling the browser that only the h1 will be in this vertical-rl mode, while the rest of my page stays in the default of horizontal-tb.\nSo now the dessert course is over. Let me serve up this whole meal, and explain the the CSS Writing Mode Specification.\nWhy learn about writing modes?\nThere are three reasons I\u2019m teaching writing modes to everyone\u2014including western audiences\u2014and explaining the whole system, instead of quickly showing you a simple trick.\n\n\nWe live in a big, diverse world, and learning about other languages is fascinating. Many of you lay out pages in languages like Chinese, Japanese and Korean. Or you might be inspired to in the future.\n\n\nUsing writing-mode to turn bits sideways is cool. This CSS can be used in all kinds of creative ways, even if you are working only in English.\n\nMost importantly, I\u2019ve found understanding Writing Modes incredibly helpful when understanding Flexbox and CSS Grid. Before I learned Writing Mode, I felt like there was still a big hole in my knowledge, something I just didn\u2019t get about why Grid and Flexbox work the way they do. Once I wrapped my head around Writing Modes, Grid and Flexbox got a lot easier. Suddenly the Alignment properties, align-* and justify-*, made sense.\n\nWhether you know about it or not, the writing mode is the first building block of every layout we create. You can do what we\u2019ve been doing for 25 years \u2013 and leave your page set to the default left-to-right direction, horizontal top-to-bottom writing mode. Or you can enter a world of new possibilities where content flows in other directions.\nCSS properties\nI\u2019m going to focus on the CSS writing-mode property in this article. It has five possible options:\n  writing-mode: horizontal-tb;\n  writing-mode: vertical-rl;\n  writing-mode: vertical-lr;\n  writing-mode: sideways-rl;\n  writing-mode: sideways-lr;\nThe CSS Writing Modes Specification is designed to support a wide range of written languages in all our human and linguistic complexity. Which\u2014spoiler alert\u2014is pretty insanely complex. The global evolution of written languages has been anything but simple. \nSo I\u2019ve got to start with explaining some basic concepts of web page layout and writing systems. Then I can show you what these CSS properties do. \nInline Direction, Block Direction, and Character Direction\nIn the world of the web, there\u2019s a concept of \u2018block\u2019 and \u2018inline\u2019 layout. If you\u2019ve ever written display: block or display: inline, you\u2019ve leaned on these concepts.  \nIn the default writing mode, blocks stack vertically starting at the top of the page and working their way down. Think of how a bunch of block-levels elements stack\u2014like a bunch of a paragraphs\u2014that\u2019s the block direction. \n\nInline is how each line of text flows. The default on the web is from left to right, in horizontal lines. Imagine this text that you are reading right now, being typed out one character at a time on a typewriter. That\u2019s the inline direction. \n\nThe character direction is which way the characters point. If you type a capital \u201cA\u201d for instance, on which side is the top of the letter? Different languages can point in different directions. Most languages have their characters pointing towards the top of the page, but not all.\n\nPut all three together, and you start to see how they work as a system. \nThe default settings for the web work like this.\nNow that we know what block, inline, and character directions mean, let\u2019s see how they are used in different writing systems from around the world.\nThe four writing systems of CSS Writing Modes\nThe CSS Writing Modes Specification handles all the use cases for four major writing systems; Latin, Arabic, Han and Mongolian. \nLatin-based systems\nOne writing system dominates the world more than any other, reportedly covering about 70% of the world\u2019s population. \n\nThe text is horizontal, running from left to right, or LTR. The block direction runs from top to bottom. \nIt\u2019s called the Latin-based system because it includes all languages that use the Latin alphabet, including English, Spanish, German, French, and many others. But there are many non-Latin-alphabet languages that also use this system, including Greek, Cyrillic (Russian, Ukrainian, Bulgarian, Serbian, etc.), and Brahmic scripts (Devanagari, Thai, Tibetan), and many more.\nYou don\u2019t need to do anything in your CSS to trigger this mode. This is the default. \nBest practices, however, dictate that you declare in your opening <html> element which language and which direction (LTR or RTL) you are using. This website, for instance, uses <html lang='en-gb' dir='ltr'> to let the browser know this content is published in Great Britian\u2019s version of English, in a left to right direction. \nArabic-based systems\nArabic, Hebrew and a few other languages run the inline direction from right to left. This is commonly known as RTL.  \nNote that the inline direction still runs horizontally. The block direction runs from top to bottom. And the characters are upright.\n\nIt\u2019s not just the flow of text that runs from right to left, but everything about the layout of the website. The upper right-hand corner is the starting position. Important things are on the right. The eyes travel from right to left. So, typically RTL websites use layouts that are just like LTR websites, only flipped.\nOn websites that support both LTR and RTL, like the United Nations\u2019 site at un.org, the two layouts are mirror images of each other.\nFor many web developers, our experiences with internationalization have focused solely on supporting Arabic and Hebrew script. \nCSS layout hacks for internationalization & RTL\nTo prepare an LTR project to support RTL, developers have had to create all sorts of hacks. For example, the Drupal community started a convention of marking every margin-left and -right, every padding-left and -right, every float: left and float: right with the comment /* LTR */. Then later developers could search for each instance of that exact comment, and create stylesheets to override each left with right, and vice versa. It\u2019s a tedious and error prone way to work. CSS itself needed a better way to let web developers write their layout code once, and easily switch language directions with a single command.\nOur new CSS layout system does exactly that. Flexbox, Grid and Alignment use start and end instead of left and right. This lets us define everything in relationship to the writing system, and switch directions easily. By writing justify-content: flex-start, justify-items: end, and eventually margin-inline-start: 1rem we have code that doesn\u2019t need to be changed. \nThis is a much better way to work. I know it can be confusing to think through start and end as replacements for left and right. But it\u2019s better for any multiligual project, and it\u2019s better for the web as a whole.\nSadly, I\u2019ve seen CSS preprocessor tools that claim to \u201cfix\u201d the new CSS layout system by getting rid of start and end and bringing back left and right. They want you to use their tool, write justify-content: left, and feel self-righteous. It seems some folks think the new way of working is broken and should be discarded. It was created, however, to fulfill real needs. And to reflect a global internet. As Bruce Lawson says, WWW stands for the World Wide Web, not the Wealthy Western Web. Please don\u2019t try to convince the industry that there\u2019s something wrong with no longer being biased towards western culture. Instead, spread the word about why this new system is here.  \nSpend a bit of time drilling the concept of inline and block into your head, and getting used to start and end. It will be second nature soon enough. \nI\u2019ve also seen CSS preprocessors that let us use this new way of thinking today, even as all the parts aren\u2019t fully supported by browsers yet. Some tools let you write text-align: start instead of text-align: left, and let the preprocessor handle things for you. That is terrific, in my opinion. A great use of the power of a preprocessor to help us switch over now. \nBut let\u2019s get back to RTL. \nHow to declare your direction\nYou don\u2019t want to use CSS to tell the browser to switch from an LTR language to RTL. You want to do this in your HTML. That way the browser has the information it needs to display the document even if the CSS doesn\u2019t load.\nThis is accomplished mainly on the html element. You should also declare your main language. As I mentioned above, the 24 ways website is using <html lang='en-gb' dir='ltr'> to declare the LTR direction and the use of British English. The UN Arabic website uses <html lang='ar' dir='rtl'>to declare the site as an Arabic site, using a RTL layout. \nThings get more complicated when you\u2019ve got a page with a mix of languages. But I\u2019m not going to get into all of that, since this article is focused on CSS and layouts, not explaining everything about internationalization. \nLet me just leave direction here by noting that much of the heavy work of laying out the characters which make up each word is handled by Unicode. If you are interested in learning more about LTR, RTL and bidirectional text, watch this video: Introduction to Bidirectional Text, a presentation by Elika Etemad. \nMeanwhile, let\u2019s get back to CSS.\nThe writing mode CSS for Latin-based and Arabic-based systems\nFor both of these systems\u2014Latin-based and Arabic-based, whether LTR or RTL\u2014the same CSS property applies for specifying the writing mode: writing-mode: horizontal-tb. That\u2019s because in both systems, the inline text flow is horizontal, while the block direction is top-to-bottom. This is expressed as horizontal-tb.\nhorizontal-tb is the default writing mode for the web, so you don\u2019t need to specify it unless you are overriding something else higher up in the cascade. You can just imagine that every site you\u2019ve ever built came with:\nhtml {\n  writing-mode: horizontal-tb;\n}\nNow let\u2019s turn our attention to the vertical writing systems. \nHan-based systems\nThis is where things start to get interesting. \nHan-based writing systems include CJK languages, Chinese, Japanese, Korean and others. There are two options for laying out a page, and sometimes both are used at the same time.\nMuch of CJK text is laid out like Latin-based languages, with a horizontal top-to-bottom block direction, and a left-to-right inline direction. This is the more modern way to doing things, started in the 20th century in many places, and further pushed into domination by the computer and later the web. \nThe CSS to do this bit of the layouts is the same as above:\nsection {\n  writing-mode: horizontal-tb;\n}\nOr, you know, do nothing, and get that result as a default. \nAlternatively Han-based languages can be laid out in a vertical writing mode, where the inline direction runs vertically, and the block direction goes from right to left. \nSee both options in this diagram:\n\nNote that the horizontal text flows from left to right, while the vertical text flows from right to left. Wild, eh? \nThis Japanese issue of Vogue magazine is using a mix of writing modes. The cover opens on the left spine, opposite of what an English magazine does.  \n\nThis page mixes English and Japanese, and typesets the Japanese text in both horizontal and vertical modes. Under the title \u201cRichard Stark\u201d in red, you can see a passage that\u2019s horizontal-tb and LTR, while the longer passage of text at the bottom of the page is typeset vertical-rl. The red enlarged cap marks the beginning of that passage. The long headline above the vertical text is typeset LTR, horizontal-tb.\n\nThe details of how to set the default of the whole page will depend on your use case. But each element, each headline, each section, each article can be marked to flow the opposite of the default however you\u2019d like.\nFor example, perhaps you leave the default as horizontal-tb, and specify your vertical elements like this:\ndiv.articletext {\n  writing-mode: vertical-rl;\n}\nOr alternatively you could change the default for the page to a vertical orientation, and then set specific elements to horizontal-tb, like this:\nhtml { \n  writing-mode: vertical-rl;\n}\nh2, .photocaptions, section {\n  writing-mode: horizontal-tb;\n}\nIf your page has a sideways scroll, then the writing mode will determine whether the page loads with upper left corner as the starting point, and scroll to the right (horizontal-tb as we are used to), or if the page loads with the upper right corner as the starting point, scrolling to the left to display overflow. Here\u2019s an example of that change in scrolling direction, in a CSS Writing Mode demo by Chen Hui Jing. Check out her demo \u2014 you can switch from horizontal to vertical writing modes with a checkbox and see the difference. \n\nMongolian-based systems\nNow, hopefully so far all of this kind of makes sense. It might be a bit more complicated than expected, but it\u2019s not so hard. Well, enter the Mongolian-based systems.\nMongolian is also a vertical script language. Text runs vertically down the page. Just like Han-based systems. There are two major differences. First, the block direction runs the other way. In Mongolian, block-level elements stack from left to right. \nHere\u2019s a drawing of how Wikipedia would look in Mongolian if it were laid out correctly.\nPerhaps the Mongolian version of Wikipedia will be redone with this layout.\nNow you might think, that doesn\u2019t look so weird. Tilt your head to the left, and it\u2019s very familiar. The block direction starts on the left side of the screen and goes to the right. The inline direction starts on the top of the page and moves to the bottom (similar to RTL text, just turned 90\u00b0 counter-clockwise). But here comes the other huge difference. The character direction is \u201cupside down\u201d. The top of the Mongolian characters are not pointing to the left, towards the start edge of the block direction. They point to the right. Like this:\n\nNow you might be tempted to ignore all this. Perhaps you don\u2019t expect to be typesetting Mongolian content anytime soon. But here\u2019s why this is important for everyone \u2014 the way Mongolian works defines the results writing-mode: vertical-lr. And it means we cannot use vertical-lr for typesetting content in other languages in the way we might otherwise expect. \nIf we took what we know about vertical-rl and guessed how vertical-lr works, we might imagine this:\n\nBut that\u2019s wrong. Here\u2019s how they actually compare:\n\nSee the unexpected situation? In both writing-mode: vertical-rl and writing-mode: vertical-lr latin text is rotated clockwise. Neither writing mode let\u2019s us rotate text counter-clockwise. \nIf you are typesetting Mongolian content, apply this CSS in the same way you would apply writing-mode to Han-based writing systems. To the whole page on the html element, or to specific pages of the page like this:\nsection {\n  writing-mode: vertical-lr;\n}\nNow, if you are using writing-mode for a graphic design effect on a language that is otherwise typesets horizontally, I don\u2019t think writing-mode: vertical-lr is useful. If the text wraps onto two lines, it stacks in a very unexpected way. So I\u2019ve sort of obliterated it from my toolkit. I find myself using writing-mode: vertical-rl a lot. And never using -lr. Hm.\nWriting modes for graphic design\nSo how do we use writing-mode to turn English headlines sideways? We could rely on transform: rotate()\nHere are two examples, one for each direction. (By the way, each of these demos use CSS Grid for their overall layout, so be sure to test them in a browser that supports CSS Grid, like Firefox Nightly.)\n\nIn this demo 4A, the text is rotated clockwise using this code: \nh1 {\n  writing-mode: vertical-rl;\n}\n\nIn this demo 4B, the text is rotated counter-clockwise using this code: \nh1 {\n  writing-mode: vertical-rl;\n  transform: rotate(180deg);\n  text-align: right;\n}\nI use vertical-rl to rotate the text so that it takes up the proper amount of space in the overall flow of the layout. Then I rotate it 180\u00b0 to spin it around to the other direction. And then I use text-align: right to get it to rise up to the top of it\u2019s container. This feels like a hack, but it\u2019s a hack that works.\nNow what I would like to do instead is use another CSS value that was designed for this use case \u2014 one of the two other options for writing mode.\nIf I could, I would lay out example 4A with:\nh1 {\n  writing-mode: sideways-rl;\n}\nAnd layout example 4B with: \nh1 {\n  writing-mode: sideways-lr;\n}\nThe problem is that these two values are only supported in Firefox. None of the other browsers recognize sideways-*. Which means we can\u2019t really use it yet. \nIn general, the writing-mode property is very well supported across browsers. So I\u2019ll use writing-mode: vertical-rl for now, with the transform: rotate(180deg); hack to fake the other direction. \nThere\u2019s much more to what we can do with the CSS designed to support multiple languages, but I\u2019m going to stop with this intermediate introduction. \nIf you do want a bit more of a taste, look at this example that adds text-orientation: upright; to the mix \u2014 turning the individual letters of the latin font to be upright instead of sideways.\n\nIt\u2019s this demo 4C, with this CSS applied: \nh1 {\n  writing-mode: vertical-rl;\n  text-orientation: upright;\n  text-transform: uppercase;\n  letter-spacing: -25px;\n}\nYou can check out all my Writing Modes demos at labs.jensimmons.com/#writing-modes. \n\nI\u2019ll leave you with this last demo. One that applies a vertical writing mode to the sub headlines of a long article. I like how small details like this can really bring a fresh feeling to the content. \nSee the Pen Writing Mode Demo \u2014 Article Subheadlines by Jen Simmons (@jensimmons) on CodePen.", "year": "2016", "author": "Jen Simmons", "author_slug": "jensimmons", "published": "2016-12-23T00:00:00+00:00", "url": "https://24ways.org/2016/css-writing-modes/", "topic": "code"}
{"rowid": 303, "title": "We Need to Talk About Technical Debt", "contents": "In my work with clients, a lot of time is spent assessing old, legacy, sprawling systems and identifying good code, bad code, and technical debt.\nOne thing that constantly strikes me is the frequency with which bad code and technical debt are conflated, so let me start by saying this:\nNot all technical debt is bad code, and not all bad code is technical debt.\nSometimes your bad code is just that: bad code. Calling it technical debt often feels like a more forgiving and friendly way of referring to what may have just been a poor implementation or a substandard piece of work.\nIt is an oft-misunderstood phrase, and when mistaken for meaning \u2018anything legacy or old hacky or nasty or bad\u2019, technical debt is swept under the carpet along with all of the other parts of the codebase we\u2019d rather not talk about, and therein lies the problem.\nWe need to talk about technical debt.\nWhat We Talk About When We Talk About Technical Debt\nThe thing that separates technical debt from the rest of the hacky code in our project is the fact that technical debt, by definition, is something that we knowingly and strategically entered into. Debt doesn\u2019t happen by accident: debt happens when we choose to gain something otherwise-unattainable immediately in return for paying it back (with interest) later on.\nAn Example\nYou\u2019re a front-end developer working on a SaaS product, and your sales team is courting a large customer \u2013 a customer so large that you can\u2019t really afford to lose them. The customer tells you that as long as you can allow them to theme your SaaS application according to their branding, they are willing to sign on the dotted line\u2026 the problem being that your CSS architecture was never designed to incorporate theming at all, and there isn\u2019t currently a nice, clean way to incorporate a theme into the codebase.\nYou and the business make the decision that you will hack a theme into the product in two days. It\u2019s going to be messy, it\u2019s going to be ugly, but you can\u2019t afford to lose a huge customer just because your CSS isn\u2019t quite right, right now. This is technical debt.\nYou deliver the theme, the customer signs up, and everyone is happy. Except you (and the business, because you are one and the same) have a decision to make:\n\nDo we go back and build theming into the CSS architecture as a first-class citizen, porting the hacked theme back into a codified and formal framework?\nDo we carry on as we are? Things are working okay, and the customer paid up, so is there any reason to invest time and effort into things after we (and the customer) got what we wanted?\n\nOption 1 is choosing to pay off your debts; Option 2 is ignoring your repayments.\nWith Option 1, you\u2019re acknowledging that you did what you could given the constraints, but, free of constraints, you\u2019d have done something different. Now, you are choosing to implement that something different.\nWith Option 2, however, you are avoiding your responsibility to repay your debt, and you are letting interest accrue. The problem here is that\u2026\n\nyour SaaS product now offers theming to one of your customers;\nanother potential customer might also demand the ability to theme their instance of your product;\nyou can\u2019t refuse them that request, nor can you quickly fulfil it;\nyou hack in another theme, thus adding to the balance of your existing debt;\nand so on (plus interest) for every subsequent theme you need to implement.\n\nHere you have increased entropy whilst making little to no attempt to address what you already knew to be problems.\nYour second, third, fourth, fifth request for theming will be hacked on top of your hack, further accumulating debt whilst offering nothing by way of a repayment. After a long enough period, the code involved will get so unwieldy, so hard to work with, that you are forced to tear it all down and start again, and the most painful part of this is that you\u2019re actually paying off even more than your debt repayments would have been in the first place. Two days of hacking plus, say, five days of subsequent refactoring, would still have been substantially less than the weeks you will now have to spend rewriting your CSS to fix and incorporate the themes properly. You\u2019ve made a loss; your strategic debt ultimately became a loss-making exercise.\nThe important thing to note here is that you didn\u2019t necessarily write bad code. You knew there were two options: the quick way and the correct way. The decision to take the quick route was a definite choice, because you knew there was a better way. Implementing the better way is your repayment.\nGood Debt and Bad Debt\nTechnical debt is acceptable as long as you have intentions to settle; it can be a valuable solution to a business problem, provided the right approach is taken afterwards. That doesn\u2019t, however, mean that all debt is born equal. Just as in real life, there is good debt and there is bad debt.\nGood debt might be\u2026\n\na mortgage;\na student loan, or;\na business loan.\n\nThese are types of debt that will secure you the means of repaying them. These are well considered debts whose very reason for being will allow you to make the money to pay them off\u2014they have real, tangible benefit.\nA business loan to secure some equipment and premises will allow you to start an enterprise whose revenue will allow you to pay that debt back; a student loan will allow you to secure the kind of job that has the ability to pay a student loan back.\nThese kinds of debt involve a considered and well-balanced decision to acquire something in the short term in the knowledge that you will have the means, in the long term, to pay it back.\nConversely, bad debt might be\u2026\n\nborrowing $1,000 from a loan shark so you can go to Vegas, or;\ntaking out a payday loan in order to buy a new television.\n\nBoth of these kinds of debt will leave you paying for things that didn\u2019t provide you a way of earning your own capital. That is to say, the loans taken did not secure anything that would help pay off said loans. These are bad debts that will usually provide a net loss. You really are only gaining the short term in exchange for a long term financial responsibility: i.e., was it worth it?\nA good litmus test for debt is to compare the gains of its immediate benefit with the cost of its long term commitment.\nThe earlier example of theming a site is a good debt, provided we are keeping up our repayments (all debt is bad debt if you don\u2019t). A calculated decision to do something \u2018wrong\u2019 in the short term with the promise of better payoffs later on.\nBad Technical Debt\nThe majority of my work is with front-end development teams\u2014CSS is what I do. To that end, the most succinct example of technical debt for that audience is simply:\n!important\nAll front-end developers know the horrors and dangers associated with using !important, yet we continue to use it. Why?\nIt\u2019s not necessarily because we\u2019re bad developers, but because we see a shortcut. !important is usually implemented as a quick way out of a sticky specificity situation. We could spend the rest of the day refactoring our CSS to fix the issue at its source, or we can spend mere seconds typing the word !important and patch over the symptoms.\nThis is us making an explicit decision to do something less than ideal now in exchange for immediate benefit. After all, refactoring our CSS will take a lot more time, and will still only leave us with the same outcome that the vastly quicker !important solution will, so it seems to make better business sense.\nHowever, this is a bad debt. !important takes seconds to implement but weeks to refactor. The cost of refactoring this back out later will be an order of magnitude higher than it would be to have done things properly the first time. The first !important usually sets a precedent, and subsequent developers are likely to have to use it themselves in order to get around the one that you left.\nSo many CSS projects deteriorate because of this one simple word, and rewrites become more and more imminent. That makes it possibly the most costly 10 bytes a CSS developer could ever write.\nBad Code\nNow we\u2019ve got a good idea of what constitutes technical debt, let\u2019s take a look at what constitutes bad code. Something I hear time and time again in my client work goes a little like this:\n\nWe\u2019ve amassed a lot of technical debt and we\u2019d like to get a strategy in place\nto begin dealing with it.\n\nWhilst I genuinely admire their willingness to identify and desire to fix problems in their code, sometimes they\u2019re not looking at technical debt at\nall\u2014sometimes they\u2019re just looking at bad code, plain and simple.\nWhere technical debt is knowing that there\u2019s a better way, but the quicker way makes more sense right now, bad code is not caring if there\u2019s a better way at all.\nAgain, looking at a CSS-specific world, a lot of bad code is contributed by non-front-end developers with little training, appreciation, or even respect for the front-end landscape. Writing code with reckless abandon should not be described as technical debt, because to do so would imply that\u2026\n\nthe developers knew they were implementing a sub-par solution, but\u2026\nthe developers also knew that a better solution was out there, which\u2026\nimplies that it can be tidied up relatively simply.\n\nDevelopers writing bad code is a larger and more cultural problem that requires a lot more effort to fix. Hopefully\u2014and usually\u2014bad code is in the minority, but it helps to be objective in identifying and solving it. Bad code usually doesn\u2019t happen for a good enough reason, and is therefore much harder to justify.\nTechnical debt often represents ability in judgement, whereas bad code often represents a gap in skills.\nTakeaway\nTake time to familiarise yourself with the true concepts underlying technical debt and why it exists. Understand that technical debt can be good or bad. Admit that sometimes code is just of poor quality.\nUnderstanding these points will allow you to make better calls around what you might need to refactor and when, and what skills gaps you might have in your team.\n\nSometimes it\u2019s okay to cut corners if there is a tangible gain to be had in the immediate term.\nTechnical debt is okay provided it is a sensible debt and you have intentions to pay it off.\nTechnical debt is not necessarily synonymous with bad code, and bad code isn\u2019t necessarily technical debt. Technical debt is code that was implemented given limited knowledge or resource, with the understanding that you would need to repay something in future.\nTechnical debt is not inherently bad\u2014failure to make repayments is. Periodically, it is justifiable\u2014encouraged, even\u2014to enter a debt in order to fulfil a more pressing matter. However, it is imperative that we begin making repayments as soon as we are capable, be that based on newly available time or knowledge.\nBad code is worse than technical debt as it represents a lack of knowledge or quality control within a team. It needs a much more fundamental fix.", "year": "2016", "author": "Harry Roberts", "author_slug": "harryroberts", "published": "2016-12-05T00:00:00+00:00", "url": "https://24ways.org/2016/we-need-to-talk-about-technical-debt/", "topic": "code"}
{"rowid": 300, "title": "Taking Device Orientation for a Spin", "contents": "When The Police sang \u201cDon\u2019t Stand So Close To Me\u201d they weren\u2019t talking about using a smartphone to view a panoramic image on Facebook, but they could have been. For years, technology has driven relentlessly towards devices we can carry around in our pockets, and now that we\u2019re there, we\u2019re expected to take the thing out of our pocket and wave it around in front of our faces like a psychotic donkey in search of its own dangly carrot.\nBut if you can\u2019t beat them, join them.\nA brave new world\nA couple of years back all sorts of specs for new HTML5 APIs sprang up much to our collective glee. Emboldened, we ran a few tests and found they basically didn\u2019t work in anything and went off disheartened into the corner for a bit of a sob.\nTurns out, while we were all busy boohooing, those browser boffins have actually being doing some work, and lo and behold, some of these APIs are even half usable. Mostly literally half usable\u2014we\u2019re still talking about browsers, after all.\nNow, of course they\u2019re all a bit JavaScripty and are going to involve complex methods and maths and science and probably about a thousand dependancies from Github that will fall out of fashion while we\u2019re still trying to locate the documentation, right? Well, no! \nSo what if we actually wanted to use one of these APIs, say to impress our friends with our ability to make them wave their phones in front of their faces (because no one enjoys looking hapless more than the easily-technologically-impressed), how could we do something like that? Let\u2019s find out.\nThe Device Orientation API\nThe phone-wavy API is more formally known as the DeviceOrientation Event Specification. It does a bunch of stuff that basically doesn\u2019t work, but also gives us three values that represent orientation of a device (a phone, a tablet, probably not a desktop computer) around its x, y and z axes. You might think of it as pitch, roll and yaw if you like to spend your weekends wearing goggles and a leather hat.\nThe main way we access these values is through an event listener, which can inform our code every time the value changes. Which is constantly, because you try and hold a phone still and then try and hold the Earth still too.\nThe API calls those pitch, roll and yaw values alpha, beta and gamma. Chocks away:\nwindow.addEventListener('deviceorientation', function(e) {\n    console.log(e.alpha);\n    console.log(e.beta);\n    console.log(e.gamma);\n});\nIf you look at this test page on your phone, you should be able to see the numbers change as you twirl the thing around your body like the dance partner you never had. Wrist strap recommended.\nOne important note\nLike may of these newfangled APIs, Device Orientation is only available over HTTPS. We\u2019re not allowed to have too much fun without protection, so make sure that you\u2019re working on a secure line. I\u2019ve found a quick and easy way to share my local dev environment over TLS with my devices is to use an ngrok tunnel.\nngrok http -host-header=rewrite mylocaldevsite.dev:80\nngrok will then set up a tunnel to your dev site with both HTTP and HTTPS URL options. You, of course, want the HTTPS option.\nRight, where were we?\nMake something to look at\nIt\u2019s all well and good having a bunch of numbers, but they\u2019re no use unless we do something with them. Something creative. Something to inspire the generations. Or we could just build that Facebook panoramic image viewer thing (because most of us are familiar with it and we\u2019re not trying to be too clever here). Yeah, let\u2019s just build one of those.\nOur basic framework is going to be similar to that used for an image carousel. We have a container, constrained in size, and CSS overflow property set to hidden. Into this we place our wide content and use positioning to move the content back and  forth behind the \u2018window\u2019 so that the part we want to show is visible.\nHere it is mocked up with a slider to set the position. When you release the slider, the position updates. (This actually tests best on desktop with your window slightly narrowed.)\nThe details of the slider aren\u2019t important (we\u2019re about to replace it with phone-wavy goodness) but the crucial part is that moving the slider results in a function call to position the image. This takes a percentage value (0-100) with 0 being far left and 100 being far right (or \u2018alt-nazi\u2019 or whatever).\nvar position_image = function(percent) {\n    var pos = (img_W / 100)*percent;\n    img.style.transform = 'translate(-'+pos+'px)'; \n};\nAll this does is figure out what that percentage means in terms of the image width, and set the transform: translate(\u2026); CSS property to move the image. (We use translate because it might be a bit faster to animate than left/right positioning.)\nOk. We can now read the orientation values from our device, and we can programatically position the image. What we need to do is figure out how to convert those raw orientation values into a nice tidy percentage to pass to our function and we\u2019re done. (We\u2019re so not done.)\nThe maths bit\nIf we go back to our raw values test page and make-believe that we have a fascinating panoramic image of some far-off beach or historic monument to look at, you\u2019ll note that the main value that is changing as we swing back and forth is the \u2018alpha\u2019 value. That\u2019s the one we want to track.\nAs our goal here is hey, these APIs are interesting and fun and not let\u2019s build the world\u2019s best panoramic image viewer, we\u2019ll start by making a few assumptions and simplifications:\n\nWhen the image loads, we\u2019ll centre the image and take the current nose-forward orientation reading as the middle.\nMoving left, we\u2019ll track to the left of the image (lower percentage).\nMoving right, we\u2019ll track to the right (higher percentage).\nIf the user spins round, does cartwheels or loads the page then hops on a plane and switches earthly hemispheres, they\u2019re on their own.\n\nNose-forward\nWhen the page loads, the initial value of alpha gives us our nose-forward position. In Safari on iOS, this is normalised to always be 0, whereas most everywhere else it tends to be bound to pointy-uppy north. That doesn\u2019t really matter to us, as we don\u2019t know which direction the user might be facing in anyway \u2014 we just need to record that initial state and then use it to compare any new readings.\nvar initial_position = null;\n\nwindow.addEventListener('deviceorientation', function(e) {\n    if (initial_position === null) {\n        initial_position = Math.floor(e.alpha);\n    };\n\n    var current_position = initial_position - Math.floor(e.alpha);\n});\n(I\u2019m rounding down the values with Math.floor() to make debugging easier - we\u2019ll take out the rounding later.)\nWe get our initial position if it\u2019s not yet been set, and then calculate the current position as a difference between the new value and the stored one.\nThese values are weird\nOne thing you need to know about these values, is that they range from 0 to 360 but then you also get weird left-of-zero values like -2 and whatever. And they wrap past 360 back to zero as you\u2019d expect if you do a forward roll.\nWhat I\u2019m interested in is working out my rotation. If 0 is my nose-forward position, I want a positive value as I turn right, and a negative value as I turn left. That puts the awkward 360-tipping point right behind the user where they can\u2019t see it.\nvar rotation  = current_position;\nif (current_position > 180) rotation = current_position-360;\nWhich way up?\nSince we\u2019re talking about orientation, we need to remember that the values are going to be different if the device is held in portrait on landscape mode. See for yourself - wiggle it like a steering wheel and you get different values. That\u2019s easy to account for when you know which way up the device is, but in true browser style, the API for that bit isn\u2019t well supported. The best I can come up with is:\nvar screen_portrait = false;\nif (window.innerWidth < window.innerHeight) {\n    screen_portrait = true;\n}\nIt works. Then we can use screen_portrait to branch our code:\nif (screen_portrait) {\n        if (current_position > 180) rotation = current_position-360;\n} else {\n        if (current_position < -180) rotation = 360+current_position;\n}\nHere\u2019s the code in action so you can see the values for yourself. If you change screen orientation you\u2019ll need to refresh the page (it\u2019s a demo!).\nLimiting rotation\nNow, while the youth of today are rarely seen without a phone in their hands, it would still be unreasonable to ask them to spin through 360\u00b0 to view a photo. Instead, we need to limit the range of movement to something like 60\u00b0-from-nose in either direction and normalise our values to pan the entire image across that 120\u00b0 range. -60 would be full-left (0%) and 60 would be full-right (100%).\nIf we set max_rotation = 60, that code ends up looking like this:\nif (rotation > max_rotation) rotation = max_rotation;\nif (rotation < (0-max_rotation)) rotation = 0-max_rotation;\n\nvar percent = Math.floor(((rotation + max_rotation)/(max_rotation*2))*100);\nWe should now be able to get a rotation from -60\u00b0 to +60\u00b0 expressed as a percentage. Try it for yourself.\nThe big reveal\nAll that\u2019s left to do is pass that percentage to our image positioning function and would you believe it, it might actually work.\nposition_image(percent);\nYou can see the final result and take it for a spin. Literally.\nSo what have we made here? Have we built some highly technical panoramic image viewer to aid surgeons during life-saving operations using only JavaScript and some slightly questionable mathematics? No, my friends, we have not. Far from it. \nWhat we have made is progress. We\u2019ve taken a relatively newly available hardware API and a bit of simple JavaScript and paired it with existing CSS knowledge and made something that we didn\u2019t have this morning. Something we probably didn\u2019t even want this morning. Something that if you take a couple of steps back and squint a bit might be a prototype for something vaguely interesting. But more importantly, we\u2019ve learned that our browsers are just a little bit more capable than we thought.\nThe web platform is maturing rapidly. There are new, relatively unexplored APIs for doing all sorts of crazy thing that are often dismissed as the preserve of native apps. Like some sort of app marmalade. Poppycock. \nThe web is an amazing, exciting place to create things. All it takes is some base knowledge of the fundamentals, a creative mind and a willingness to learn. We have those! So let\u2019s create things.", "year": "2016", "author": "Drew McLellan", "author_slug": "drewmclellan", "published": "2016-12-24T00:00:00+00:00", "url": "https://24ways.org/2016/taking-device-orientation-for-a-spin/", "topic": "code"}
{"rowid": 298, "title": "First Steps in VR", "contents": "The web is all around us. As web folk, it is our responsibility to consider the impact our work can have. Part of this includes thinking about the future; the web changes lives and if we are building the web then we are the ones making decisions that affect people in every corner of the world. I find myself often torn between wanting to make the right decisions, and just wanting to have fun. To fiddle and play. We all know how important it is to sometimes just try ideas, whether they will amount to much or not. \nI think of these two mindsets as production and prototyping, though of course there are lots of overlap and phases in between. I mention this because virtual reality is currently seen as a toy for rich people, and in some ways at the moment it is. But with WebVR we are able to create interesting experiences with a relatively low entry point. I want us to have open minds, play around with things, and then see how we can use the tools we have at our disposal to make things that will help people.\nEvery year we see articles saying it will be the \u201cyear of virtual reality\u201d, that was especially prevalent this year. 2016 has been a year of progress, VR isn\u2019t quite mainstream but with efforts like Playstation VR and Google Cardboard, we are definitely seeing much more of it. This year also saw the consumer editions of the Oculus Rift and HTC Vive. So it does seem to be a good time for an overview of how to get involved with creating virtual reality on the web.\nWebVR is an API for connecting to devices and retrieving continuous data such as the position and orientation. Unlike the Web Audio API and some other APIs, WebVR does not feel like a framework. You use it however you want, taking the data and using it as you wish. To make it easier, there are plenty of resources such as Three.js, A-Frame and ReactVR that help to make the heavy lifting a bit easier.\nGetting Started with A-Frame\nI like taking the opportunity to learn new things whenever I can. So while planning this article I thought that instead of trying to teach WebGL or even Three.js in a way that is approachable for all, I would create my first project using A-Frame and write about that. This is not a tutorial as such, I just want to show how to go about getting involved with VR. The beauty of A-Frame is that it is very similar to web components, you can just write HTML to build worlds that will automatically work on all the different types of devices. It uses WebGL and WebVR but in such a way that it quite drastically reduces the learning curve. That\u2019s not to say you can\u2019t build complex things, you have complete access to write JavaScript and shaders.\nI\u2019m lazy. Whenever I learn a new language or framework I have found that the best way, personally, for me to learn is to have a project and to copy the starting code from someone else. A project lets you have a good idea of what you want to produce and it means you can ignore a lot of the irrelevant documentation, focussing purely on what you need. That reduces the stress of figuring things out. Copying code also makes it easier, because you know your boilerplate code is working. There\u2019s nothing worse than getting stuck before anything actually works the first time. So I tinker. I take code and I modify it, I play around. It\u2019s fun.\nFor this project I wanted to keep things as simple as possible, so I can easily explain it without the classic \u201cdraw a circle then draw an owl\u201d. I wrote a list of requirements, with some stretch goals that you can give a try yourself if you fancy:\n\nMust work on Google Cardboard at a minimum, because of price\nTherefore, it must not rely on having a controller\nAuto-moving around a maze would be a good example\nMove in direction you look\nStretch goal: Scoring, time until you hit a wall or get stuck in maze\nStretch goal: Levels, so the map doesn\u2019t need to be random\nStretch goal: Snow!\n\nI decided to  base this project on an example, Platforms, by Don McCurdy who wrote the really useful aframe-extras. Platforms has random 3D blocks that you can jump onto, going up into the sky. So I took his code and reduced it so that the blocks are randomly spread on the ground. \n<!DOCTYPE html>\n<html>\n<head>\n    <meta charset=\"utf-8\">\n    <meta name=\"viewport\" content=\"width=device-width\">\n    <title>24 ways</title>\n    <script src=\"https://aframe.io/releases/0.3.2/aframe.js\"></script>\n    <script src=\"//cdn.rawgit.com/donmccurdy/aframe-extras/v2.6.1/dist/aframe-extras.min.js\"></script>\n</head>\n<body>\n<a-scene>\n    <a-entity id=\"player\"\n              camera\n              universal-controls\n              kinematic-body\n              position=\"0 1.8 0\">\n    </a-entity>\n\n    <a-entity id=\"walls\"></a-entity>\n\n    <a-grid id=\"ground\" static-body></a-grid>\n\n    <a-sky id=\"sky\" color=\"#AADDF0\"></a-sky>\n\n    <!-- Lighting -->\n    <a-light type=\"ambient\" color=\"#ccc\"></a-light>\n</a-scene>\n<script>\n    document.querySelector('a-scene').addEventListener('render-target-loaded', function () {\n        var MAP_SIZE = 10,\n                PLATFORM_SIZE = 5,\n                NUM_PLATFORMS = 50;\n        var platformsEl = document.querySelector('#walls');\n        var v, box;\n        for (var i = 0;  i < NUM_PLATFORMS; i++) {\n            // y: 0 is ground\n            v = {\n                x: (Math.floor(Math.random() * MAP_SIZE) - PLATFORM_SIZE) * PLATFORM_SIZE,\n                y: PLATFORM_SIZE / 2,\n                z: (Math.floor(Math.random() * MAP_SIZE) - PLATFORM_SIZE) * PLATFORM_SIZE\n            };\n            box = document.createElement('a-box');\n            platformsEl.appendChild(box);\n            box.setAttribute('color', '#39BB82');\n            box.setAttribute('width', PLATFORM_SIZE);\n            box.setAttribute('height', PLATFORM_SIZE);\n            box.setAttribute('depth', PLATFORM_SIZE);\n            box.setAttribute('position', v.x + ' ' + v.y + ' ' + v.z);\n            box.setAttribute('static-body', '');\n        }\n        console.info('Platforms loaded.');\n    });\n</script>\n</body>\n</html>\nAs you can see, this is very readable. Especially if you ignore the JavaScript that is used to create the maze. A-Frame (with A-Frame Extras) gives you a lot of power with relatively little to learn. We start with an <a-scene> which is the container for everything that is going to show up on the screen. There are a few <a-entity> which can be compared to <div> as they are essentially non-semantic containers, able to be used for any purpose. The attributes are used to define functionality, for example the camera attribute sets the entity to function as a camera and kinematic-body makes it collide instead of go through objects. Attributes are also used to set position and sizes, often using JavaScript to dynamically define them.\nStyling\nNow we\u2019ve got the HTML written, we need to style it. To do this we add A-Frame compatible attributes such as color and material. I recommend playing around, you can get some quite impressive effects fairly easily. Originally I wanted a light snowy maze but it ended up being dark and foggy, as I really liked the feeling it gave.\nNote, you will probably need a server running for images to work. You can do this by running python -m \"SimpleHTTPServer\" in the folder where the code is, then go to localhost:8000 in browser.\nTextures\nUnless you are going for a cartoony style, you probably want to find some textures. I found some on textures.com, one image worked well for the walls and the other for the floor.\n<a-assets>\n    <img id=\"texture-floor\" src=\"floor.jpg\">\n    <img id=\"texture-wall\" src=\"wall.jpg\">\n</a-assets>\nThe <a-assets> is used to define (as well as preload and cache) all assets, including images, audio and video. As you can see, images in the Asset Management System just use normal img tags. The ids are important here as we can use them later for using the textures. \nTo apply a texture to an object, you create a material. For a simple material where it just shows the image, you set the src to the id selector of the image.\nReplace: \n<a-grid id=\"ground\" static-body></a-grid>\nWith:\n<a-grid id=\"ground\" static-body material=\"src: #texture-floor\"></a-grid>\nThis will automatically make the image repeat over the entire floor, in my case filling it with bricks. The walls are pretty much identical, with the slight exception that it is set in JavaScript as they are dynamically defined.\nbox.setAttribute('material', 'src: #texture-wall');\nThat\u2019s it for the textures, for now at least. These will not look completely realistic, as the light will bump off the rectangular wall rather than texture itself. This can be improved by using maps, textures that are used to modify the shape and physical properties of the object. \nLighting\nThe next part of styling is lighting. By using fog and different types of lighting, we are able to add atmospheric details to the game to make it feel that bit more realistic and polished.\nThere are lots of types of light in A-Frame (most coming from Three.js). You can add a light either by using the <a-light> entity or by attaching a light attribute to any other entity. If there are no lights defined then A-Frame adds some by default so that the scene is always lit.\nTo start with I wanted to light up the scene with a general light, type=\"ambient\", so that the whole game felt slightly dark. I chose to set the light to a reddish colour #92455E. After playing around with intensity I chose 0.4, it added enough light to get the feeling I wanted without it being overly red. I also added a blue skybox (<a-sky>), as it looked a bit odd with a white sky.\n<a-light type=\"ambient\" color=\"#92455E\" intensity=\"0.4\"></a-light>\n<a-sky id=\"sky\" color=\"#0000ff\"></a-sky>\nI felt that the maze looked good with a red tinge but it was a bit flat, everything was the same colour and it was a bit dark. So I added a light within the #player entity, this could have been as an attribute but I set it as a child a-light instead. By using type=\"point\" with a high intensity and low distance, it showed close walls as being lighter. It also added a sort-of object to the player, it isn\u2019t a walking human or anything but by moving light where the player is it feels a bit more physical.\n<a-light color=\"#fff\" distance=\"5\" intensity=\"0.7\" type=\"point\"></a-light>\n\nBy this point it was starting to look decent, so I wanted to add the fog to really give some personality and depth to the maze. To do this I added the fog attribute to the <a-scene> with type=exponential so it looks thicker the further away it is and a mid intensity, so you feel a bit lost but can still see.\n\nI was very happy with this result. It took a lot of playing around with colours and values, which is fun in itself. I highly recommend you take the code (or write your own) and play around with the numbers.\nMovement\nOne of the reasons I decided to use aframe-extras is that it has a few different camera controls built in. As you saw earlier, I am using the universal-controls which gives WASD (keyboard) controls by default. I wanted to make it automatically move in the direction that you\u2019re looking, but I wasn\u2019t quite sure how without rewriting the controls. So I asked Don McCurdy for advice and he very nicely gave me a small snippet of code to get it working.\nAFRAME.registerComponent('automove-controls', {\n    init: function () {\n        this.speed = 0.1;\n        this.isMoving = true;\n        this.velocityDelta = new THREE.Vector3();\n    },\n    isVelocityActive: function () {\n        return this.isMoving;\n    },\n    getVelocityDelta: function () {\n        this.velocityDelta.z = this.isMoving ? -speed : 0;\n        return this.velocityDelta.clone();\n    }\n});\nReplace:\nuniversal-controls\nWith:\nuniversal-controls=\"movementControls: automove, gamepad, keyboard\"\nThis works by creating a component automove-controls that adds auto-move to the player without overriding movement completely. It doesn\u2019t even touch direction, it just checks if isMoving is true then moves the player by the set speed. Components can be creating for adding all kinds of functionality with relative ease. It makes it very powerful for people of all difficulty levels.\nBuilding a map\nCurrently the maze is created randomly, which is great but means there will often be walls that overlap or the player gets trapped with nowhere to go. So to solve this, I decided to use a map editor (Tiled) so that we can create the mazes ourselves. This is a great start towards one of the stretch goals, levels.\nI made the maze in Tiled by finding a random tileset online (we don\u2019t need to actually show the images), I used one tile for the wall and another for the player. Then I exported as a JavaScript file and modified it in my text editor to get rid of everything I didn\u2019t need. I made it so 0 is the path, 1 is the wall and 2 is the player. I then added the script to the HTML, as a separate file so it\u2019s easy to update in the future. \nvar map =\n{\n \"data\":[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],\n \"height\":10,\n \"width\":10\n}\n\nAs you can see, this gives a simple 10x10 maze with some dead ends. The player starts in the bottom right corner (my choice, could be anywhere). I rewrote the random platforms code (from Don\u2019s example) to instead loop over the map data and place walls where it is 1 and position the player where data is 2. I set the position so that the origin of the map would be 0,1.5,0. The y axis is in this case the height (ground being 0), but if a wall is positioned at 0 by its centre then some of it is underground. So the y needed to be the height divided by 2.\ndocument.querySelector('a-scene').addEventListener('render-target-loaded', function () {\n    var WALL_SIZE = 5,\n        WALL_HEIGHT = 3;\n    var el = document.querySelector('#walls');\n    var wall;\n\n    for (var x = 0; x < map.height; x++)  {\n        for (var y = 0; y < map.width; y++) {\n\n            var i = y*map.width + x;\n            var position = (x-map.width/2)*WALL_SIZE + ' ' + 1.5 + ' ' + (y-map.height/2)*WALL_SIZE;\n            if (map.data[i] === 1) {\n                // Create wall\n                wall = document.createElement('a-box');\n                el.appendChild(wall);\n                wall.setAttribute('color', '#fff');\n                wall.setAttribute('material', 'src: #texture-wall;');\n                wall.setAttribute('width', WALL_SIZE);\n                wall.setAttribute('height', WALL_HEIGHT);\n                wall.setAttribute('depth', WALL_SIZE);\n                wall.setAttribute('position', position);\n                wall.setAttribute('static-body', ');\n            }\n\n            if (map.data[i] === 2)  {\n                // Set player position\n                document.querySelector('#player').setAttribute('position', position);\n            }\n\n        }\n    }\n    console.info('Walls added.');\n});\n\nWith this added, it makes it nice and easy to change around the map as well as to add new features. Perhaps you want monsters or objects. Just set the number in the map data and add an if statement to the loop. In the future you could add layers, so multiple things can be in the same position. Or perhaps even make the maze go up the y axis too, with ramps or staircases. There\u2019s a lot you can do with relative ease. As you can see, A-Frame really does reduce the learning curve of 3D and VR on the web.\nIt\u2019s Not All Fun And Games\nA lot of examples of virtual reality are games, including this one. So it is understandable to think that VR is for gaming, but actually that\u2019s just a tiny subset. There are all sorts of applications for VR, including story telling, data visualisation and even meditation.\nThere have been a number of cases where it has been shown virtual reality can help as a tool for therapies:\n\nOxford study finds virtual reality can help treat severe paranoia\nVirtual Reality Therapy for Phobias at the Duke Faculty Practice\nBravemind: Virtual Reality Exposure Therapy at the University of Southern California\n\nThese are just a few examples of where virtual reality is being used around the world to help people feel better and get through some very tough times. There have also been examples of it being used for simulating war zones or medical situations, both as a teaching and journalism tool.\nWrapping Up\nTen years ago, on this very site, Cameron Moll wrote an article explaining the mobile web. He explained how mobile phones with data plans were becoming increasingly common, that WAP 2.0 included the XHTML Mobile Profile meaning it would be familiar with web folk. \u201cThe mobile web is rapidly becoming an XHTML environment, and thus you and I can apply our existing \u201cdesktop web\u201d skills to understand how to develop content for it.\u201d\nWe can look at that and laugh a little, we have come a very long way in the last decade. Even people in developing countries with very little money have mobile phones with access to a web that is far more capable than the \u201cdesktop web\u201d Cameron was referring to.\nSo while I am not saying virtual reality is going to change the world or replace our phones, who knows! We can use our skills as web folk to dabble, we don\u2019t need to learn any new languages. If on the 2026 edition of 24 ways, somebody references this article and looks at how far we have come\u2026 well, let\u2019s hope we have used our skills well and made the world just that little bit better. And if VR is a fad? Well it\u2019s fun\u2026 have a go anyway.", "year": "2016", "author": "Shane Hudson", "author_slug": "shanehudson", "published": "2016-12-11T00:00:00+00:00", "url": "https://24ways.org/2016/first-steps-in-vr/", "topic": "code"}
{"rowid": 296, "title": "Animation in Design Systems", "contents": "Our modern front-end workflow has matured over time to include design systems and component libraries that help us stay organized, improve workflows, and simplify maintenance. These systems, when executed well, ensure proper documentation of the code available and enable our systems to scale with reduced communication conflicts. \nBut while most of these systems take a critical stance on fonts, colors, and general building blocks, their treatment of animation remains disorganized and ad-hoc. Let\u2019s leverage existing structures and workflows to reduce friction when it comes to animation and create cohesive and performant user experiences. \nUnderstand the importance of animation\nPart of the reason we treat animation like a second-class citizen is that we don\u2019t really consider its power. When users are scanning a website (or any environment or photo), they are attempting to build a spatial map of their surroundings. During this process, nothing quite commands attention like something in motion. \nWe are biologically trained to notice motion: evolutionarily speaking, our survival depends on it. For this reason, animation when done well can guide your users. It can aid and reinforce these maps, and give us a sense that we understand the UX more deeply. We retrieve information and put it back where it came from instead of something popping in and out of place. \n\n\u201cWhere did that menu go? Oh it\u2019s in there.\u201d \n\nFor a deeper dive into how animation can connect disparate states, I wrote about the Importance of Context-Shifting in UX Patterns for CSS-Tricks.\nAn animation flow on mobile.\nAnimation also aids in perceived performance. Viget conducted a study where they measured user engagement with a standard loading GIF versus a custom animation. Customers were willing to wait almost twice as long for the custom loader, even though it wasn\u2019t anything very fancy or crazy. Just by showing their users that they cared about them, they stuck around, and the bounce rates dropped.\n\n14 second generic loading screen.22 second custom loading screen.\nThis also works for form submission. Giving your personal information over to an online process like a static form can be a bit harrowing. It becomes more harrowing without animation used as a signal that something is happening, and that some process is completing. That same animation can also entertain users and make them feel as though the wait isn\u2019t as long. \nEli Fitch gave a talk at CSS Dev Conf called: \u201cPerceived Performance: The Only Kind That Really Matters\u201d, which is one of my favorite talk titles of all time. In it, he discussed how we tend to measure things like timelines and network requests because they are more quantifiable\u2013and therefore easier to measure\u2013but that measuring how a user feels when visiting the site is more important and worth the time and attention. \nIn his talk, he states \u201cHumans over-estimate passive waits by 36%, per Richard Larson of MIT\u201d. This means that if you\u2019re not using animation to speed up how fast the wait time of a form submission loads, users are perceiving it to be much slower than the dev tools timeline is recording.\nReign it in\nUnlike fonts, colors, and so on, we tend to add animation in as a last step, which leads to disorganized implementations that lack overall cohesion. If you asked a designer or developer if they would create a mockup or build a UI without knowing the fonts they were working with, they would dislike the idea. Not knowing the building blocks they\u2019re working with means that the design can fall apart or the development can break with something so fundamental left out at the start. Good animation works the same way.\nThe first step in reigning in your use of animation is to perform an animation audit. Look at all the places you are using animation on your site, or the places you aren\u2019t using animation but probably should. (Hint: perceived performance of a loader on a form submission can dramatically change your bounce rates.) \nNot sure how to perform a good audit? Val Head has a great chapter on it in her book, Designing Interface Animations, which has of buckets of research and great ideas.\nEven some beautiful component libraries that have animation in the docs make this mistake. You don\u2019t need every kind of animation, just like you don\u2019t need every kind of font. This bloats our code. Ask yourself questions like: do you really need a flip 180 degree animation? I can\u2019t even conceive of a place on a typical UI where that would be useful, yet most component libraries that I\u2019ve seen have a mixin that does just this.\nWhich leads to\u2026\nHave an opinion\nMany people are confused about Material Design. They think that Material Design is Motion Design, mostly because they\u2019ve never seen anyone take a stance on animation before and document these opinions well. But every time you use Material Design as your motion design language, people look at your site and think GOOGLE. Now that\u2019s good branding.\nBy using Google\u2019s motion design language and not your own, you\u2019re losing out on a chance to be memorable on your own website.\nWhat does having an opinion on motion look like in practice? It could mean you\u2019ve decided that you never flip things. It could mean that your eases are always going to glide. In that instance, you would put your efforts towards finding an ease that looks \u201cgliding\u201d and pulling out any transform: scaleX(-1) animation you find on your site. Across teams, everyone knows not to spend time mocking up flipping animation (even if they\u2019re working on an entirely different codebase), and to instead work on something that feels like it glides. You save time and don\u2019t have to communicate again and again to make things feel cohesive.\nCreate good developer resources\nSometimes people don\u2019t incorporate animation into a design system because they aren\u2019t sure how, beyond the base hover states. All animation properties can be broken into interchangeable pieces. This allows developers and designers alike to mix and match and iterate quickly, while still staying in the correct language. Here are some recommendations (with code and a demo to follow):\nCreate timing units, similar to h1, h2, h3. In a system I worked on recently, I called these t1, t2, t3. T1 would be reserved for longer pieces, down to t5 which is a bit like h5 in that it\u2019s the default (usually around .25 seconds or thereabouts).\nKeep animation easings for entrance, exit, entrance emphasis and exit emphasis that people can commonly refer to. This, and the animation-fill-mode, are likely to be the only two properties that can be reused for the entrance and exit of the animation.\nUse the animation-name property to define the keyframes for the animation itself. I would recommend starting with 5 or 6 before making a slew of them, and see if you need more. Writing 30 different animations might seem like a nice resource, but just like your color palette having too many can unnecessarily bulk up your codebase, and keep it from feeling cohesive. Think critically about what you need here. \nSee the Pen Modularized Animation for Component Libraries by Sarah Drasner (@sdras) on CodePen.\n\nThe example above is pared-down, but you can see how in a robust system, having pieces that are interchangeable cached across the whole system would save time for iterations and prototyping, not to mention make it easy to make adjustments for different feeling movement on the same animation easily.\nOne low hanging fruit might be a loader that leads to a success dialog. On a big site, you might have that pattern many times, so writing up a component that does only that helps you move faster while also allowing you to really zoom in and focus on that pattern. You avoid throwing something together at the last minute, or using a GIF, which are really heavy and mushy on retina. You can make singular pieces that look really refined and are reusable. \nReact and Vue Implementations are great for reusable components, as you can create a building block with a common animation pattern, and once created, it can be a resource for all. Remember to take advantage of things like props to allow for timing and easing adjustments like we have in the previous example!\nResponsive\nAt the very least we should ensure that interaction also works well on mobile, but if we\u2019d like to create interactions that take advantage of all of the gestures mobile has to offer, we can use libraries like zingtouch or hammer to work with swipe or multiple finger detection. With a bit of work, these can all be created through native detection as well.\nResponsive web pages can specify initial-scale=1.0 in the meta tag so that the device is not waiting the required 300ms on the secondary tap before calling action. Interaction for touch events must either start from a larger touch-target (40px \u00d7 40px or greater) or use @media(pointer:coarse) as support allows.\nBuy-in\nSometimes people don\u2019t create animation resources simply because it gets deprioritized. But design systems were also something we once had to fight for, too. This year at CSS Dev Conf, Rachel Nabors demonstrated how to plot out animation wants vs. needs on a graph (reproduced with her permission) to help prioritize them:\n\n\nThis helps people you\u2019re working with figure out the relative necessity and workload of the addition of these animations and think more critically about it. You\u2019re also more likely to get something through if you\u2019re proving that what you\u2019re making is needed and can be reused. \nGood compromises can be made this way: \u201cwe\u2019re not going to go all out and create an animated \u2018About Us\u2019 page like you wanted, but I suppose we can let our users know their contact email went through with a small progress and success notification.\u201d \nSuccessfully pushing smaller projects through helps build trust with your team, and lets them see what this type of collaboration can look like. This builds up the type of relationship necessary to push through projects that are more involved. It can\u2019t be overstressed that good communication is key.\nGet started!\nWith these tools and good communication, we can make our codebases more efficient, performant, and feel better for our users. We can enhance the user experience on our sites, and create great resources for our teams to allow them to move more quickly while innovating beautifully.", "year": "2016", "author": "Sarah Drasner", "author_slug": "sarahdrasner", "published": "2016-12-16T00:00:00+00:00", "url": "https://24ways.org/2016/animation-in-design-systems/", "topic": "code"}
{"rowid": 295, "title": "Internet of Stranger Things", "contents": "This year I\u2019ve been running a workshop about using JavaScript and Node.js to work with all different kinds of electronics on the Raspberry Pi. So especially for 24 ways I\u2019m going to show you how I made a very special Raspberry Pi based internet connected project! And nothing says Christmas quite like a set of fairy lights connected to another dimension1.\nWhat you\u2019ll see\nYou can rig up the fairy lights in your home, with the scrawly letters written under each one. The people from the other side (i.e. the internet) will be able to write messages to you from their browser in real time. In fact why not try it now; check this web page. When you click the lights in your browser, my lights (and yours) will turn on and off in real life! (There may be a queue if there are lots of people accessing it, hit the \u201cSend a message\u201d button and wait your turn.)\n\n\n\n\nIt\u2019s all done with JavaScript, using Node.js running on both the Raspberry Pi and on the server. I\u2019m using WebSockets to communicate in real time between the browser, server and Raspberry Pi.\nWhat you\u2019ll need\n\nRaspberry Pi any of the following models: Zero (will need straight male header pins soldered2 and Micro USB OTG adaptor), A+, B+, 2, or 3\nMicro SD card at least 4Gb Class 10 speed3\nMicro USB power supply at least 2A\nUSB Wifi dongle (unless you have a Pi 3 - that has wifi built in). \nAddressable fairy lights\nLogic level shifter (with pins soldered unless you want to do it!)\nBreadboard\nJumper wires (3x male to male and 4x female to male)\n\nOptional but recommended\n\nBase board to hold the Pi and Breadboard (often comes with a breadboard!)\n\nFind links for where to buy all of these items that goes along with this tutorial. The total price should be around $1004.\nSetting up the Raspberry Pi\nYou\u2019ll need to install the SD card for the Raspberry Pi. You\u2019ll find a link to download a disk image on the support document, ready-made with the Raspbian version of Linux, along with Node.js and all the files you need. Download it and write it to the SD card using the fantastic free software Etcher5. \nNext up you have to configure the wifi details on the SD card. If you plug the card into your computer you should see a drive called BOOT. There\u2019s a text file on there called wpa_supplicant.conf. Open it up in your favourite text editor and replace mywifi and mypassword with your wifi details6.\nnetwork={\n   ssid=\"mywifi\"\n   psk=\"mypassword\"\n}\nSave the file, eject the card from your computer and plug it into the Raspberry Pi. \nIf you have a base board or holder for the Raspberry Pi, attach it now. Then connect the wifi USB dongle7 and power supply, but don\u2019t plug it in yet!\nWiring!\nTime to wire everything up! \nFirst of all, push the Logic Level Converter into the middle of the breadboard:\n\nLogic Level Converter\nThe logic level converter may be labelled differently from the one in the diagram but the pins are usually exactly the same internally. I would just make sure the pins marked HV (High Voltage) are on the bottom and LV (Low Voltage) are on the top. \n\nRaspberry Pi pins only output 3.3v but the lights need 5v. That\u2019s why we need the logic level converter in there to boost up the signal.\nConnect the first two wires between the Raspberry Pi pins and the breadboard:\n\nNote that the pins on the Raspberry Pi are male, so you need a female to male jumper wire to connect between them and the breadboard. The colours don\u2019t have to match but it\u2019s easier to follow (and check) if you use the same ones as in the diagram. \n\nThen the next two:\n\nThis is what you should have so far:\n\nLights\nNow to connect the lights! My ones have a connector with three holes in it that I can push jumper wires into, and hopefully yours will too! So I used the male-to-male jumper wires to connect them to the breadboard.\n\n\n\nMake sure that you connect the right end of the lights, mine has a male connector at the wrong end so it\u2019s impossible to do this, but double check.  \nAlso make sure that the holes in the light connector are the same as mine. To do this, follow the wires from the connector to the first light and look at the circuit board inside. You should just about be able to make out the connections labelled + (sometimes 5V, V+ or VCC), GND (or \u2018-\u2019 or G) and DI (sometimes DIN for data in).\n\nYou can just about make out the +, DI and GND on this picture. Note that on the other side of the board there is a DO for data out - that\u2019s what takes the data along to the chip in the next light. Make sure that you\u2019re plugging into the data-in and not the data-out! \nThat\u2019s it! Everything\u2019s plugged in and ready to go! But before you plug power into your Pi, double check all your wires and make sure they\u2019re exactly right! You could damage your Raspberry Pi if it is not wired correctly. So triple check!\n\nThe Moment of Truth!\nPlug in the Raspberry Pi and wait around a minute or two for it to boot up. If all is well, the lights should strobe rainbow colours for one second - that\u2019s your confirmation that it\u2019s connected to my WebSocket server and ready to receive messages from the upside-down! \n\nHowever, if the first light in the string is pulsing red, it means that you\u2019re not connected to the internet. So check the Troubleshooting section of the support document. If it\u2019s pulsing green then you\u2019re connected to the internet but can\u2019t connect to my server. It must have gone down. Sorry! The code will keep trying so leave it running and maybe it\u2019ll come back up.  \nRig up the lights!\nFix the lights up on the wall however you want, pins, nails, tape. I\u2019ve used cable clips. Just be careful! I\u2019m using a 50 light string so I\u2019ve programmed it to use the lights at the end for the letters. That way I have just under half the string to extend down to the floor where I can keep the Raspberry Pi. \nCheck the photo here to see how the lights line up, note that there are spare unused lights in-between each row:\n\nNow visit lights.seb.ly and you\u2019ll see this : \n\nIf you\u2019re the only one online you\u2019ll have direct connection to the lights and any letter you click on will light up both in the browser and in real life. If there are other people there, you\u2019ll need to click the button to join the queue and wait your turn. \nHow it works - the geeky details!\nElectronics:\nThe pins on the Raspberry Pi are known as GPIO pins, general-purpose input/output. You can connect a wide variety of electronic components to them, LED lights, buttons, switches, and sensors. You can turn the power to the pins on and off using Node.js (or Python, if you prefer). \nAddressable LEDs or \u201cNeopixels\u201d\nWe\u2019re only using one GPIO pin on the Raspberry Pi (the other connections are 5V, 3.3V and ground) and that single pin is controlling all of the lights in the string. The code turns the pin on and off really fast in strictly timed morse-code-like dots and dashes to transmit binary data. The chips attached to each LED decode the binary and adjust the output to the LED accordingly. That chip then sends the data on to the next light in the string. \nThe chips on each light are the WS2811, part of the WS281x family that come in a multitude of different form factors and are often packaged with tiny LEDs in a single component. They are commonly referred to as Neopixels8 and I used them on my Laser Light Synths project.\nNeopixels with the chip and the LED all in one - it\u2019s the white square shaped component and the darker square inside is the chip. These are only 5mm wide!\nA Laser Light Synth! Covered with around 800 super bright neopixels!\nLogic Level Converter\nThe logic level converter is a really cheap and easy way to change the level from 3.3v to 5v and back again. You must be careful that you do not connect 5v into a GPIO pin or you will most likely damage the Raspberry Pi processor chip. \nPower\nNeopixels can often draw a lot of current so you need to be careful how you power them. I\u2019ve measured the current draw from the string to be less than 800mA so you should be fine wired directly to the 5V output. But if you use more lights or have them all on really bright at once, you\u2019ll need to use a separate 5V power supply. If you want to learn more, check out Adafruit\u2019s Neopixel Uberguide. \nNode.js\nThere are two Node.js apps running here, one on the Raspberry Pi and one on my server. You can see the code on my GitHub at github.com/sebleedelisle/stranger-lights for the Raspberry Pi and github.com/sebleedelisle/stranger-lights-server for the server.  And they\u2019re hosted on npm as stranger-lights and stranger-lights-server.  \nThe server side code sets up a standard web server to deliver the HTML for the web interface. It also sets up a WebSocket server that allows for real-time communication between the browser and the server. This server code also manages the queue and who is in control of the lights at any given time.\nWebSockets\nI\u2019m using the excellent Socket.io library to manage the WebSocket connection. Both the browser and the Raspberry Pi Node.js app connects to my WebSocket server. \nWhen you click on a letter in the browser, a message is sent to the server, which forwards it to the connected Raspberry Pi clients and also all the web browsers9. \nThe Raspberry Pi code\nThe Node.js app runs automatically on startup, and I made this happen by adding this to the /etc/rc.local file: \nnode /home/pi/strangerthings/client.js > /dev/null &\nAnything in the rc.local file gets executed when the Pi boots up and this line of code runs the Node.js app and routes its output to nowhere (ie /dev/null). The & means that it runs it in the background and doesn\u2019t hold up the boot process. \nWorking with the Raspberry Pi headless\nYou might know that when a computer has no screen or keyboard, you would refer to it as \u201crunning headless\u201d. So just like most web servers, you need to configure it over the network with ssh10. If you\u2019re on a mac you can find your Pi on the network through the name raspberrypi.local11, otherwise you\u2019ll need to find its IP address. There\u2019s more on the guide to Remote Access instructions on the Raspberry Pi website. And if you\u2019re very new to the terminal, I highly recommend this great online Linux command line tutorial.\nImprovements\nThis is quite an early experiment and I\u2019m sure I\u2019ll discover lots of optimisations over the next few weeks, especially if the server gets a proper hammering today! But there are a few things you can do. Obviously I\u2019ve just rigged up my lights with Post-it notes. It\u2019d be a lot nicer to get a paint brush and try to recreate the Winona-in-a-manic-state text style. \nWhere next?\nFinding quality resources about Node.js for electronics on the Pi can be somewhat hit and miss, but this is getting better all the time. Alternatively I am thinking about running some online courses, please let me know if that\u2019s something you\u2019d be interested in, or sign up to my mailing list at st4i.com. \nThere are many many more resources for the Raspberry Pi with Python (gpiozero is a good place to start), so if that language works for you, you\u2019ll be spoilt for choice! \nAlso take a look at Arduino - it\u2019s an incredibly popular platform for electronics and the internet is literally bursting with resources. \nI hope you enjoyed this little foray into the world of JavaScript electronics on the Raspberry Pi! If you get this working at home please let me know! Tweet me at @seb_ly. \n\n\n\n\nNot a particularly original idea, but I don\u2019t think I\u2019ve seen anyone do it quite like this before, ie using WebSockets, and Node.js on a Raspberry Pi. Other examples: Internet of Stranger Things, Strangerlights.com, and loads of examples on Instructables\u00a0\u21a9\ufe0e\n\n\nVideo guide to soldering pins on to a Pi Zero and further soldering advice from Adafruit\u00a0\u21a9\ufe0e\n\n\nSlower cards will work but performance may suffer\u00a0\u21a9\ufe0e\n\n\nOr \u00a35,000 in UK money. Sorry, Brexit joke :)\u00a0\u21a9\ufe0e\n\n\nYou will need a card reader on your computer - most micro SD cards come with an adaptor that fits standard SD slots. \u00a0\u21a9\ufe0e\n\n\nSSID and password should be all that you need but you can see all the config options on this wpa supplicant guide\u00a0\u21a9\ufe0e\n\n\nRaspberry Pi Zero will require the OTG to USB adaptor to attach the wifi dongle\u00a0\u21a9\ufe0e\n\n\nThanks to Adafruit who invented the term neopixels so we don\u2019t have to refer to them as WS281x any more!\u00a0\u21a9\ufe0e\n\n\nSo you can see other people sending messages in the browser\u00a0\u21a9\ufe0e\n\n\nssh is short for Secure Shell and is a way to connect to a remote computer and type in it just like you would in the terminal.\u00a0\u21a9\ufe0e\n\n\nYou can change this default hostname using raspi-config\u00a0\u21a9\ufe0e", "year": "2016", "author": "Seb Lee-Delisle", "author_slug": "sebleedelisle", "published": "2016-12-01T00:00:00+00:00", "url": "https://24ways.org/2016/internet-of-stranger-things/", "topic": "code"}
{"rowid": 294, "title": "New Tricks for an Old Dog", "contents": "Much of my year has been spent helping new team members find their way around the expansive and complex codebase that is the TweetDeck front-end, trying to build a happy and productive group of people around a substantial codebase with many layers of legacy.\nI\u2019ve loved doing this. Everything from writing new documentation, drawing diagrams, and holding technical architecture sessions teaches you something you didn\u2019t know or exposes an area of uncertainty that you can go work on.\nIn this article, I hope to share some experiences and techniques that will prove useful in your own situation and that you can impress your friends in some new and exciting ways!\nHow do you do, fellow kids?\nTo start with I\u2019d like to introduce you to our JavaScript framework, Flight. Right now it\u2019s used by twitter.com and TweetDeck although, as a company, Twitter is largely moving to React.\nOver time, as we used Flight for more complex interfaces, we found it wasn\u2019t scaling with us.\nComposing components into trees was fiddly and often only applied for a specific parent-child pairing. It seems like an obvious feature with hindsight, but it didn\u2019t come built-in to Flight, and it made reusing components a real challenge.\nThere was no standard way to manage the state of a component; they all did it slightly differently, and the technique often varied by who was writing the code. This cost us in maintainability as you just couldn\u2019t predict how a component would be built until you opened it.\nMaking matters worse, Flight relied on events to move data around the application. Unfortunately, events aren\u2019t good for giving structure to complex logic. They jump around in a way that\u2019s hard to understand and debug, and force you to search your code for a specific string \u2014 the event name\u201a to figure out what\u2019s going on.\nTo find fixes for these problems, we looked around at other frameworks. We like React for it\u2019s simple, predictable state management and reactive re-render flow, and Elm for bringing strict functional programming to everyone.\nBut when you have lots of existing code, rewriting or switching framework is a painful and expensive option. You have to understand how it will interact with your existing code, how you\u2019ll test it alongside existing code, and how it will affect the size and performance of the application. This all takes time and effort!\nInstead of planning a rewrite, we looked for the ideas hidden within other frameworks that we could reapply in our own situation or bring to the tools we already were using.\nBoiled down, what we liked seemed quite simple:\n\nComponent nesting & composition\nEasy, predictable state management\nNormal functions for data manipulation\n\nMaking these ideas applicable to Flight took some time, but we\u2019re in a much better place now. Through persistent trial-and-error, we have well documented, testable and standard techniques for creating complex component hierarchies, updating and reacting to state changes, and passing data around the app.\nWhile the specifics of our situation and Flight aren\u2019t really important, this experience taught me something: \n\nDistill good tech into great ideas. You can apply great ideas anywhere.\n\nYou don\u2019t have to use cool kids\u2019 latest framework, hottest build tool or fashionable language to benefit from them. If you can identify a nugget of gold at the heart of it all, why not use it to improve what you have already?\nTimes, they are a changin\u2019\nApart from stealing ideas from the new and shiny, how can we keep make the most of improved tooling and techniques? Times change and so should the way we write code.\nGoing back in time a bit, TweetDeck used some slightly outmoded tools for building and bundling. Without a transpiler like Babel we were missing out new language features, and without a more advanced build tools like Webpack, every module\u2019s source was encased in AMD boilerplate.\nIn fact, we found ourselves with a mix of both AMD syntaxes:\ndefine([\"lodash\"], function (_) {\n  // . . .\n});\n\ndefine(function (require) {\n  var _ = require(\"lodash\");\n  // . . .\n});\nThis just wouldn\u2019t do. And besides, what we really wanted was CommonJS, or even ES2015 module syntax:\nimport _ from \"lodash\";\nThese days we\u2019re using Babel, Webpack, ES2015 modules and many new language features that make development just\u2026 better. But how did we get there?\nTo explain, I want to introduce you to codemods and jscodeshift.\nA codemod is a large-scale refactor of a whole codebase, often mechanical or repetitive. Think of renaming a module or changing an API like URL(\"...\") to new URL(\"...\").\njscodeshift is a toolkit for running automated codemods, where you express a code transformation using code. The automated codemod operates on each file\u2019s syntax tree \u2013 a data-structure representation of the code \u2014 finding and modifying in place as it goes.\nHere\u2019s an example that renames all instances of the variable foo to bar:\nmodule.exports = function (fileInfo, api) {\n  return api\n    .jscodeshift(fileInfo.source)\n    .findVariableDeclarators('foo')\n    .renameTo('bar')\n    .toSource();\n};\nIt\u2019s a seriously powerful tool, and we\u2019ve used it to write a series of codemods that:\n\nrename modules,\nunify our use of AMD to a single syntax,\ntransition from one testing framework to another, and\nswitch from AMD to CommonJS.\n\nThese changes can be pretty huge and far-reaching. Here\u2019s an example commit from when we switched to CommonJS:\ncommit 8f75de8fd4c702115c7bf58febba1afa96ae52fc\nDate:   Tue Jul 12 2016\n\n   Run AMD -> CommonJS codemod\n\n 418 files changed, 47550 insertions(+), 48468 deletions(-)\n\nYep, that\u2019s just under 50k lines changed, tested, merged and deployed without any trouble. AMD be gone!\n\nFrom this step-by-step approach,  using codemods to incrementally tweak and improve, we extracted a little codemod recipe for making significant, multi-stage changes:\n\nFind all the existing patterns\nChoose the two most similar\nUnify with a codemod\nRepeat.\n\nFor example:\n\nFor module loading, we had 2 competing AMD patterns plus some use of CommonJS\nThe two AMD syntaxes were the most similar\nWe used a codemod to move to unify the AMD patterns\nLater we returned to AMD to convert it to CommonJS\n\nIt\u2019s worked for us, and if you\u2019d like to know more about codemods then check out Evolving Complex Systems Incrementally by Facebook engineer, Christoph Pojer.\nWelcome aboard!\nAs TweetDeck has gotten older and larger, the amount of things a new engineer has to learn about has exploded. The myriad of microservices that manage our data and their layers of authentication, security and business logic around them make for an overwhelming amount of information to hand to a newbie.\nInspired by Amy\u2019s amazing Guide to the Care and Feeding of Junior Devs, we realised it was important to take time to design our onboarding that each of our new hires go through to make the most of their first few weeks.\nJoining a new company, team, or both, is stressful and uncomfortable. Everything you can do to help a new hire will be valuable to them. So please, take time to design your onboarding!\nAnd as you build up an onboarding process, you\u2019ll create things that are useful for more than just new hires; it\u2019ll force you to write documentation, for example, in a way that\u2019s understandable for people who are unfamiliar with your team, product and codebase. This can lead to more outside contributions: potential contributors feel more comfortable getting set up on your product without asking for help.\nThis is something that\u2019s taken for granted in open source, but somehow I think we forget about it in big companies.\nAfter all, better documentation is just a good thing. You will forget things from time to time, and you\u2019d be surprised how often the \u201cbeginner\u201d docs help!\nFor TweetDeck, we put together system and architecture diagrams, and one-pager explanations of important concepts:\n\nWhat are our dependencies?\nWhere are the potential points of failure?\nWhere does authentication live? Storage? Caching?\nWho owns \u201cX\u201d?\n\n\nOf course, learning continues long after onboarding. The landscape is constantly shifting; old services are deprecated, new APIs appear and what once true can suddenly be very wrong. Keeping up with this is a serious challenge, and more than any one person can track.\nTo address this, we\u2019ve thought hard about our knowledge sharing practices across the whole team. For example, we completely changed the way we do code review.\nIn my opinion, code review is the single most effective practice you can introduce to share knowledge around, and build the quality and consistency of your team\u2019s work. But, if you\u2019re not doing it, here\u2019s my suggestion for getting started:\n\nEvery pull request gets a +1 from someone else.\n\nThat\u2019s all \u2014 it\u2019s very light-weight and easy. Just ask someone to have a quick look over your code before it goes into master.\nAt Twitter, every commit gets a code review. We do a lot of reviewing, so small efficiency and effectiveness improvements make a big difference. Over time we learned some things:\n\nDon\u2019t review for more than hour 1\nKeep reviews smaller than ~400 lines 2\nCode review your own code first 2\n\nAfter an hour, and above roughly 400 lines, your ability to detect issues in a code review starts to decrease. So review little and often. The gaps around lunch, standup and before you head home are ideal. And remember, if someone\u2019s put code up for a review, that review is blocking them doing other work. It\u2019s your job to unblock them.\nOn TweetDeck, we actually try to keep reviews under 250 lines. It doesn\u2019t sound like much, but this constraint applies pressure to make smaller, incremental changes. This makes breakages easier to detect and roll back, and leads to a very natural feature development process that encourages learning and iteration.\nBut the most important thing I\u2019ve learned personally is that reviewing my own code is the best way to spot issues. I try to approach my own reviews the way I approach my team\u2019s: with fresh, critical eyes, after a break, using a dedicated code review tool.\nIt\u2019s amazing what you can spot when you put a new in a new interface around code you\u2019ve been staring at for hours!\nAnd yes, this list features science. The data backs up these conclusions, and if you\u2019d like to learn more about scientific approaches to software engineering then I recommend you buy Making Software: What Really Works, and Why We Believe It. It\u2019s ace.\nFor more dedicated information sharing, we\u2019ve introduced regular seminars for everyone who works on a specific area or technology. It works like this: a team-member shares or teaches something to everyone else, and next time it\u2019s someone else\u2019s turn. Giving everyone a chance to speak, and encouraging a wide range of topics, is starting to produce great results.\nIf you\u2019d like to run a seminar, one thing you could try to get started: run a point at the thing you least understand in our architecture session \u2014 thanks to James for this idea. And guess what\u2026 your onboarding architecture diagrams will help (and benefit from) this!\nMore, please!\nThere\u2019s a few ideas here to get you started, but there are even more in a talk I gave this year called Frontend Archaeology, including a look at optimising for confidence with front-end operations.\nAnd finally, thanks to Amy for proof reading this and to Passy for feedback on the original talk.\n\n\n\n\nDunsmore et al. 2000. Object-Oriented Inspection in the Face of Delocalisation. Beverly, MA: SmartBear Software.\u00a0\u21a9\n\n\nCohen, Jason. 2006. Best Kept Secrets of Peer Code Review. Proceedings of the 22nd ICSE 2000: 467-476.\u00a0\u21a9 \u21a9", "year": "2016", "author": "Tom Ashworth", "author_slug": "tomashworth", "published": "2016-12-18T00:00:00+00:00", "url": "https://24ways.org/2016/new-tricks-for-an-old-dog/", "topic": "code"}
{"rowid": 293, "title": "A Favor for Your Future Self", "contents": "We tend to think about the future when we build things. What might we want to be able to add later? How can we refactor this down the road? Will this be easy to maintain in six months, a year, two years? As best we can, we try to think about the what-ifs, and build our websites, systems, and applications with this lens. \nWe comment our code to explain what we knew at the time and how that impacted how we built something. We add to-dos to the things we want to change. These are all great things! Whether or not we come back to those to-dos, refactor that one thing, or add new features, we put in a bit of effort up front just in case to give us a bit of safety later.\nI want to talk about a situation that Past Alicia and Team couldn\u2019t even foresee or plan for. Recently, the startup I was a part of had to remove large sections of our website. Not just content, but entire pages and functionality. It wasn\u2019t a very pleasant experience, not only for the reason why we had to remove so much of what we had built, but also because it\u2019s the ultimate \u201cI really hope this doesn\u2019t break something else\u201d situation. It was a stressful and tedious effort of triple checking that the things we were removing weren\u2019t dependencies elsewhere. To be honest, we wouldn\u2019t have been able to do this with any amount of success or confidence without our test suite.\nWriting tests for code is one of those things that developers really, really don\u2019t want to do. It\u2019s one of the easiest things to cut in the development process, and there\u2019s often a struggle to have developers start writing tests in the first place. One of the best lessons the web has taught us is that we can\u2019t, in good faith, trust the happy path. We must make sure ourselves, and our users, aren\u2019t in a tough spot later on because we only thought of the best case scenarios.\nJavaScript\nRegardless of your opinion on whether or not everything needs to be built primarily with JavaScript, if you\u2019re choosing to build a JavaScript heavy app, you absolutely should be writing some combination of unit and integration tests.\nUnit tests are for testing extremely isolated and small pieces of code, which we refer to as the units themselves. Great for reused functions and small, scoped areas, this is the closest you examine your code with the testing microscope. For example, if we were to build a calculator, the most minute piece we could test could be the basic operations.\n/*\n * This example uses a test framework called Jasmine\n */\n\ndescribe(\"Calculator Operations\", function () {\n\n    it(\"Should add two numbers\", function () {\n\n    // Say we have a calculator\n        Calculator.init();\n\n    // We can run the function that does our addition calculation...\n        var result = Calculator.addNumbers(7,3);\n\n    // ...and ensure we're getting the right output\n        expect(result).toBe(10);\n\n    });\n});\nEven though these teeny bits work in isolation, we should ensure that connecting the large pieces work, as well. This is where integration tests excel. These tests ensure that two or more different areas of code, that may not directly know about each other, still behave in expected ways. Let\u2019s build upon our calculator - we may want the operations to be saved in memory after a calculation runs. This isn\u2019t as suited for a unit test because there are a few other moving pieces involved in the process (the calculations, checking if the result was an error, etc.).\n    it(\u201cShould remember the last calculation\u201d, function () {\n\n    // Run an operation\n        Calculator.addNumbers(7,10);\n\n    // Expect something else to have happened as a result\n        expect(Calculator.updateCurrentValue).toHaveBeenCalled();\n        expect(Calculator.currentValue).toBe(17);\n    });\nUnit and integration tests provide assurance that your hand-rolled JavaScript should, for the most part, never fail in a grand fashion. Although it still might happen, you could be able to catch problems way sooner than without a test suite, and hopefully never push those failures to your production environment.\nInterfaces\nRegardless of how you\u2019re building something, it most definitely has some kind of interface. Whether you\u2019re using a very barebones structure, or you\u2019re leveraging a whole design system, these things can be tested as well.\nAcceptance testing helps us ensure that users can get from point A to point B within our web things, which can provide assurance that major features are always functioning properly. By simulating user input and data entry, we can go through whole user workflows to test for both success and failure scenarios. These are not necessarily for simulating edge-case scenarios, but rather ensuring that our core offerings are stable.\nFor example, if your site requires signup, you want to make sure the workflow is behaving as expected - allowing valid information to go through signup, while invalid information does not let you progress.\n/*\n * This example uses Jasmine along with an add-on called jasmine-integration\n */\n\ndescribe(\"Acceptance tests\", function () {\n\n  // Go to our signup page\n  var page = visit(\"/signup\");\n\n  // Fill our signup form with invalid information\n  page.fill_in(\"input[name='email']\", \"Not An Email\");\n  page.fill_in(\"input[name='name']\", \"Alicia\");\n  page.click(\"button[type=submit]\");\n\n  // Check that we get an expected error message\n  it(\"Shouldn't allow signup with invalid information\", function () {\n    expect(page.find(\"#signupError\").hasClass(\"hidden\")).toBeFalsy();\n  });\n\n  // Now, fill our signup form with valid information\n  page.fill_in(\"input[name='email']\", \"thisismyemail@gmail.com\");\n  page.fill_in(\"input[name='name']\", \"Gerry\");\n  page.click(\"button[type=submit]\");\n\n  // Check that we get an expected success message and the error message is hidden\n  it(\"Should allow signup with valid information\", function () {\n    expect(page.find(\"#signupError\").hasClass(\"hidden\")).toBeTruthy();\n    expect(page.find(\"#thankYouMessage\").hasClass(\"hidden\")).toBeFalsy();\n  });\n});\nIn terms of visual design, we\u2019re now able to take snapshots of what our interfaces look like before and after any code changes to see what has changed. We call this visual regression testing. Rather than being a pass or fail test like our other examples thus far, this is more of an awareness test, intended to inform developers of all the visual differences that have occurred, intentional or not. Developers may accidentally introduce a styling change or fix that has unintended side effects on other areas of a website - visual regression testing helps us catch these sooner rather than later. These do require a bit more consistent grooming than other tests, but can be valuable in major CSS refactors or if your CSS is generally a bit like Jenga.\nTools like PhantomCSS will take screenshots of your pages, and do a visual comparison to check what has changed between two sets of images. The code would look something like this:\n/*\n * This example uses PhantomCSS\n */\n\ncasper.start(\"/home\").then(function(){\n\n  // Initial state of form\n  phantomcss.screenshot(\"#signUpForm\", \"sign up form\");\n\n  // Hit the sign up button (should trigger error)\n  casper.click(\"button#signUp\");\n\n  // Take a screenshot of the UI component\n  phantomcss.screenshot(\"#signUpForm\", \"sign up form error\");\n\n  // Fill in form by name attributes & submit\n  casper.fill(\"#signUpForm\", {\n    name: \"Alicia Sedlock\",\n    email: \"alicia@example.com\"\n  }, true);\n\n  // Take a second screenshot of success state\n  phantomcss.screenshot(\"#signUpForm\", \"sign up form success\");\n});\nYou run this code before starting any development, to create your baseline set of screen captures. After you\u2019ve completed a batch of work, you run PhantomCSS again. This will create a second batch of screenshots, which are then put through an image comparison tool to display any differences that occurred. Say you changed your margins on our form elements \u2013 your image diff would look something like this:\n\nThis is a great tool for ensuring not just your site retains its expected styling, but it\u2019s also great for ensuring nothing accidentally changes in the living style guide or modular components you may have developed. It\u2019s hard to keep eagle eyes on every visual aspect of your site or app, so visual regression testing helps to keep these things monitored.\nConclusion\nThe shape and size of what you\u2019re testing for your site or app will vary. You may not need lots of unit or integration tests if you don\u2019t write a lot of JavaScript. You may not need visual regression testing for a one page site. It\u2019s important to assess your codebase to see which tests would provide the most benefit for you and your team.\nWriting tests isn\u2019t a joy for most developers, myself included. But I end up thanking Past Alicia a lot when there are tests, because otherwise I would have introduced a lot of issues into codebases. Shipping code that\u2019s broken breaks trust with our users, and it\u2019s our responsibility as developers to make sure that trust isn\u2019t broken. Testing shouldn\u2019t be considered a \u201cnice to have\u201d - it should be an integral piece of our workflow and our day-to-day job.", "year": "2016", "author": "Alicia Sedlock", "author_slug": "aliciasedlock", "published": "2016-12-03T00:00:00+00:00", "url": "https://24ways.org/2016/a-favor-for-your-future-self/", "topic": "code"}
{"rowid": 292, "title": "Watch Your Language!", "contents": "I\u2019m bilingual. My first language is French. I learned English in my early 20s. Learning a new language later in life meant that I was able to observe my thought processes changing over time. It made me realize that some concepts can\u2019t be expressed in some languages, while other languages express these concepts with ease.\nIt also helped me understand the way we label languages. English: business. French: romance. Here\u2019s an example of how words, or the absence thereof, can affect the way we think:\nIn French we love everything. There\u2019s no straightforward way to say we like something, so we just end up loving everything. I love my sisters, I love broccoli, I love programming, I love my partner, I love doing laundry (this is a lie), I love my mom (this is not a lie). I love, I love, I love. It\u2019s no wonder French is considered romantic. When I first learned English I used the word love rather than like because I hadn\u2019t grasped the difference. Needless to say, I\u2019ve scared away plenty of first dates!\nLearning another language made me realize the limitations of my native language and revealed concepts I didn\u2019t know existed. Without the nuances a given language provides, we fail to express what we really think. The absence of words in our vocabulary gets in the way of effectively communicating and considering ideas.\nWhen I lived in Montr\u00e9al, most people in my circle spoke both French and English. I could switch between them when I could more easily express an idea in one language or the other. I liked (or should I say loved?) those conversations. They were meaningful. They were efficient.\n\nI\u2019m quadrilingual. I code in Ruby, HTML/CSS, JavaScript, Python. In the past couple of years I have been lucky enough to write code in these languages at a massive scale. In learning Ruby, much like learning English, I discovered the strengths and limitations of not only the languages I knew but the language I was learning. It taught me to choose the right tool for the job.\nWhen I started working at Shopify, making a change to a view involved copy/pasting HTML and ERB from one view to another. The CSS was roughly structured into modules, but those modules were not responsive to different screen sizes. Our HTML was complete mayhem, and we didn\u2019t consider accessibility. All this made editing views a laborious process.\nGrep. Replace all. Test. Ship it. Repeat.\nThis wasn\u2019t sustainable at Shopify\u2019s scale, so the newly-formed front end team was given two missions:\n\nMake the app responsive (AKA Let\u2019s Make This Thing Responsive ASAP)\nMake the view layer scalable and maintainable (AKA Let\u2019s Build a Pattern Library\u2026 in Ruby)\n\nLet\u2019s make this thing responsive ASAP\nThe year was 2015. The Shopify admin wasn\u2019t mobile friendly. Our browser support was set to IE10. We had the wind in our sails. We wanted to achieve complete responsiveness in the shortest amount of time. Our answer: container queries.\nIt seemed like the obvious decision at the time. We would be able to set rules for each component in isolation and the component would know how to lay itself out on the page regardless of where it was rendered. It would save us a ton of development time since we wouldn\u2019t need to change our markup, it would scale well, and we would achieve complete component autonomy by not having to worry about page layout. By siloing our components, we were going to unlock the ultimate goal of componentization, cutting the tie to external dependencies. We were cool.\nWriting the JavaScript handling container queries was my first contribution to Shopify. It was a satisfying project to work on. We could drop our components in anywhere and they would magically look good. It took us less than a couple weeks to push this to production and make our app mostly responsive.\nBut with time, it became increasingly obvious that this was not as performant as we had hoped. It wasn\u2019t performant at all. Components would jarringly jump around the page before settling in on first paint.\nIt was only when we started using the flex-wrap: wrap CSS property to build new components that we realized we were not using the right language for the job. So we swapped out JavaScript container queries for CSS flex-wrapping. Even though flex wasn\u2019t yet as powerful as we wanted it to be, it was still a good compromise. Our components stayed independent of the window size but took much less time to render. Best of all: they used CSS instead of relying on JavaScript for layout.\nIn other words: we were using the wrong language to express our layout to the browser, when another language could do it much more simply and elegantly.\nLet\u2019s build a pattern library\u2026 in Ruby\nIn order to make our view layer maintainable, we chose to build a comprehensive library of helpers. This library would generate our markup from a single source of truth, allowing us to make changes system-wide, in one place. No. More. Grepping.\nWhen I joined Shopify it was a Rails shop freshly wounded by a JavaScript framework (See: Batman.js). JavaScript was like Voldemort, the language that could not be named. Because of this baggage, the only way for us to build a pattern library that would get buyin from our developers was to use Rails view helpers. And for many reasons using Ruby was the right choice for us. The time spent ramping developers up on the new UI Components would be negligible since the Ruby API felt familiar. The transition would be simple since we didn\u2019t have to introduce any new technology to the stack. The components would be fast since they would be rendered on the server. We had a plan.\nWe put in place a set of Rails tools to make it easy to build components, then wrote a bunch of sweet, sweet components using our shiny new tools. To document our design, content and front end patterns we put together an interactive styleguide to demonstrate how every component works. Our research and development department loved it (and still do)! We continue to roll out new components, and generally the project has been successful, though it has had its drawbacks.\nSince the Shopify admin is mostly made up of a huge number of forms, most of the content is static. For this reason, using server-rendered components didn\u2019t seem like a problem at the time. With new app features increasing the amount of DOM manipulation needed on the client side, our early design decisions mean making requests to the server for each re-paint. This isn\u2019t going to cut it.\nI don\u2019t know the end of this story, because we haven\u2019t written it yet. We\u2019ve been exploring alternatives to our current system to facilitate the rendering of our components on the client, including React, Vue.js, and Web Components, but we haven\u2019t determined the winner yet. Only time (and data gathering) will tell.\nRuby is great but it doesn\u2019t speak the browser\u2019s language efficiently. It was not the right language for the job.\n\nLearning a new spoken language has had an impact on how I write code. It has taught me that you don\u2019t know what you don\u2019t know until you have the language to express it. Understanding the strengths and limitations of any programming language is fundamental to making good design decisions. At the end of the day, you make the best choices with the information you have. But if you still feel like you\u2019re unable to express your thoughts to the fullest with what you know, it might be time to learn a new language.", "year": "2016", "author": "Annie-Claude C\u00f4t\u00e9", "author_slug": "annieclaudecote", "published": "2016-12-10T00:00:00+00:00", "url": "https://24ways.org/2016/watch-your-language/", "topic": "code"}
{"rowid": 289, "title": "Front-End Developers Are Information Architects Too", "contents": "The theme of this year\u2019s World IA Day was \u201cInformation Everywhere, Architects Everywhere\u201d. This article isn\u2019t about what you may consider an information architect to be: someone in the user-experience field, who maybe studied library science, and who talks about taxonomies. This is about a realisation I had a couple of years ago when I started to run an increasing amount of usability-testing sessions with people who have disabilities: that the structure, labelling, and connections that can be made in front-end code is information architecture. People\u2019s ability to be successful online is unequivocally connected to the quality of the code that is written.\nPlaces made of information\nIn information architecture we talk about creating places made of information. These places are made of ones and zeros, but we talk about them as physical structures. We talk about going onto a social media platform, posting in blogs, getting locked out of an environment, and building applications. In 2002, Andrew Hinton stated:\n\nPeople live and work in these structures, just as they live and work in their homes, offices, factories and malls. These places are not virtual: they are as real as our own minds.\n25 Theses\n\nWe\u2019re creating structures which people rely on for significant parts of their lives, so it\u2019s critical that we carry out our work responsibly. This means we must use our construction materials correctly. Luckily, our most important material, HTML, has a well-documented specification which tells us how to build robust and accessible places. What is most important, I believe, is to understand the semantics of HTML.\nSemantics\nThe word \u201csemantic\u201d has its origin in Greek words meaning \u201csignificant\u201d, \u201csignify\u201d, and \u201csign\u201d. In the physical world, a structure can have semantic qualities that tell us something about it. For example, the stunning Westminster Abbey inspires awe and signifies much about the intent and purpose of the structure. The building\u2019s size; the quality of the stone work; the massive, detailed stained glass: these are all signs that this is a building meant for something the creators deemed important. Alternatively consider a set of large, clean, well-positioned, well-lit doors on the ground floor of an office block: they don\u2019t need an \u201centrance\u201d sign to communicate their use and to stop people trying to use a nearby fire exit to get into the building. The design of the doors signify their usage. Sometimes a more literal and less awe-inspiring approach to communicating a building\u2019s purpose happens, but the affect is similar: the building is signifying something about its purpose.\nHTML has over 115 elements, many of which have semantics to signify structure and affordance to people, browsers, and assistive technology. The HTML 5.1 specification mentions semantics, stating:\n\nElements, attributes, and attribute values in HTML are defined \u2026 to have certain meanings (semantics). For example, the <ol> element represents an ordered list, and the lang attribute represents the language of the content.\nHTML 5.1 Semantics, structure, and APIs of HTML documents\n\nHTML\u2019s baked-in semantics means that developers can architect their code to signify structure, create relationships between elements, and label content so people can understand what they\u2019re interacting with. Structuring and labelling information to make it available, usable, and understandable to people is what an information architect does. It\u2019s also what a front-end developer does, whether they realise it or not.\nA brief introduction to information architecture\nWe\u2019re going to start by looking at what an information architect is. There are many definitions, and I\u2019m going to quote Richard Saul Wurman, who is widely regarded as the father of information architecture. In 1976 he said an information architect is:\n\nthe individual who organizes the patterns inherent in data, making the complex clear; a person who creates the structure or map of information which allows others to find their personal paths to knowledge; the emerging 21st century professional occupation addressing the needs of the age focused upon clarity, human understanding, and the science of the organization of information.\nOf Patterns And Structures\n\nTo me, this clearly defines any developer who creates code that a browser, or other user agent (for example, a screen reader), uses to create a structured, navigable place for people.\nJust as there are many definitions of what an information architect is, there are for information architecture itself. I\u2019m going to use the definition from the fourth edition of Information Architecture For The World Wide Web, in which the authors define it as:\nThe structural design of shared information environments.\nThe synthesis of organization, labeling, search, and navigation systems within digital, physical, and cross-channel ecosystems.\nThe art and science of shaping information products and experiences to support usability, findability, and understanding.\nInformation Architecture For The World Wide Web, 4th Edition\nTo me, this describes front-end development. Done properly, there is an art to creating robust, accessible, usable, and findable spaces that delight all our users. For example, at 2015\u2019s State Of The Browser conference, Edd Sowden talked about the accessibility of <table>s. He discovered that by simply not using the semantically-correct <th> element to mark up <table> headings, in some situations browsers will decide that a <table> is being used for layout and essentially make it invisible to assistive technology. Another example of how coding practices can affect the usability and findability of content is shown by L\u00e9onie Watson in her How ARIA landmark roles help screen reader users video. By using ARIA landmark roles, people who use screen readers are quickly able to identify and jump to common parts of a web page.\nOur definitions of information architects and information architecture mention patterns, rules, organisation, labelling, structure, and relationships. There are numerous different models for how these elements get boiled down to their fundamentals. In his Understanding Context book, Andrew Hinton calls them Labels, Relationships, and Rules; Jorge Arango calls them Links, Nodes, And Order; and Dan Klyn uses Ontology, Taxonomy, and Choreography, which is the one we\u2019re going to use. Dan defines these terms as:\nOntology\nThe definition and articulation of the rules and patterns that govern the meaning of what we intend to communicate.\nWhat we mean when we say what we say.\nTaxonomy\nThe arrangements of the parts. Developing systems and structures for what everything\u2019s called, where everything\u2019s sorted, and the relationships between labels and categories\nChoreography\nRules for interaction among the parts. The structures it creates foster specific types of movement and interaction; anticipating the way users and information want to flow and making affordance for change over time.\n\nWe now have definitions of an information architect, information architecture, and a model of the elements of information architecture. But is writing HTML really creating information or is it just wrangling data and metadata? When does data turn into information? In his book Managing For The Future Peter Drucker states:\n\n\u2026 data is not information. Information is data endowed with relevance and purpose.\nManaging For The Future\n\nIf we use the correct semantic element to mark up content then we\u2019re developing with purpose and creating relevance. For example, if we follow the advice of the HTML 5.1 specification and mark up headings using heading rank instead of the outline algorithm, we\u2019re creating a structure where the depth of one heading is relevant to the previous one. Architected correctly, an <h2> element should be relevant to its parent, which should be the <h1>. By following the HTML specification we can create a structured, searchable, labeled document that will hopefully be relevant to what our users need to be successful. If you\u2019ve never used a screen reader, you might be wondering how the headings on a page are searchable. Screen readers give users the ability to interact with headings in a couple of ways:\n\nby creating a list of headings so users can quickly scan the page for information\nby using a keyboard command to cycle through one heading at a time\n\nIf we had a document for Christmas Day TV we might structure it something like this:\n<h1>Christmas Day TV schedule</h1>\n  <h2>BBC1</h2>\n    <h3>Morning</h3>\n      <h3>Evening</h3>\n  <h2>BBC2</h2>\n    <h3>Morning</h3>\n    <h3>Evening</h3>\n  <h2>ITV</h2>\n    <h3>Morning</h3>\n    <h3>Evening</h3>\n  <h2>Channel 4</h2>\n    <h3>Morning</h3>\n    <h3>Evening</h3>\nIf I use VoiceOver to generate a list of headings, I get this:\n\nOnce I have that list I can use keyboard commands to filter the list based on the heading level. For example, I can press 2 to hear just the <h2>s:\n\nIf we hadn\u2019t used headings, of if we\u2019d nested them incorrectly, our users would be frustrated.\nPutting this together\nLet\u2019s put this together with an example of a button that, when pressed, toggles the appearance of a panel of links. There are numerous ways we could create a button on a web page, but the best way is to just use a <button>. Every browser understands what a <button> is, how it works, and what keyboard shortcuts should be used with them. The HTML specification for the <button> element says:\n\nThe <button> element represents a button labeled by its contents.\n\nThe contents that a <button> can have include the type attribute, any relevant ARIA attributes, and the actual text label that the user sees. This information is more important than the visual design: it doesn\u2019t matter how beautiful or obtuse the design is, if the underlying code is non-semantic and poorly labelled, people are going to struggle to use it. Here are three buttons, each created with the same HTML but with different designs:\n\nRegardless of what they look like, because we\u2019ve used semantic HTML instead of a bunch of meaningless <div>s or <span>s, people who use assistive technology are going to benefit. Out of the box, without any extra development effort, a <button> is accessible and usable with a keyboard. We don\u2019t have to write event handlers to listen for people pressing the Enter key or the space bar, which we would have to do if we\u2019d faked a button with non-semantic elements. Our <button> can also be quickly findable: for example, in the same way it\u2019s possible to create a list of headings with a screen reader, I can also create a list of form elements and then quickly jump to the one I want.\nNow we have our <button>, let\u2019s add the panel we\u2019re toggling the appearance of. Here\u2019s our code:\n<button aria-controls=\"panel\" aria-expanded=\"false\" class=\"settings\" id=\"settings\" type=\"button\">Settings</button>\n\n<div class=\"panel hidden\" id=\"panel\">\n    <ul aria-labelledby=\"settings\">\n        <li><a href=\"\u2026\">Account</a></li>\n        <li><a href=\"\u2026\">Privacy</a></li>\n        <li><a href=\"\u2026\">Security</a></li>\n    </ul>\n</div>\nThere\u2019s quite a bit going on here. We\u2019re using the:\n\naria-controls attribute to architect a connection between the <button> element and the panel whose appearance it controls. When some assistive technology, for example the JAWS screen reader, encounters an element with aria-controls it audibly tells a user about the controlled expanded element and gives them the ability to move focus to it.\naria-expanded attribute to denote whether the panel is visible or not. We toggle this value using JavaScript to true when the panel is visible and false when it\u2019s not. This important attribute tells people who use screen readers about the state of the elements they\u2019re interacting with. For example, VoiceOver announces Settings expanded button when the panel is visible and Settings collapsed button when it\u2019s hidden.\naria-labelledby attribute to give the list a title of \u201cSettings\u201d. This can benefit some users of assistive technology. For example, screen readers can cycle through all the lists on a page, so being able to title them can improve findability. Being able to hear list Settings three items is, I\u2019d argue, more useful than list three items. By doing this we\u2019re supporting usability and findability.\n<ul> element to contain our list of links in our panel.\n\nLet\u2019s look at the choice of <ul> to contain our settings choices. Firstly, our settings are related items, so they belong in a structure that semantically groups things. This is something that a list can do that other elements or patterns can\u2019t. This pattern, for example, isn\u2019t semantic and has no structure:\n<div><a href=\"\u2026\">Account</a></div>\n<div><a href=\"\u2026\">Privacy</a></div>\n<div><a href=\"\u2026\">Security</a></div>\nAll we have there is three elements next to each other on the screen and in the DOM. That is not robust code that signifies anything.\nWhy are we using an unordered list as opposed to an ordered list or a definition list? A quick look at the HTML specification tells us why:\n\nThe <ul> element represents a list of items, where the order of the items is not important \u2014 that is, where changing the order would not materially change the meaning of the document.\nThe HTML 5.1 specification\u2019s description of the  element\n\nWill the meaning of our document materially change if we moved the order of our links around? Nope. Therefore, I\u2019d argue, we\u2019ve used the correct element to structure our content.\nThese coding decisions are information architecture\nI believe that what we\u2019ve done here is pure information architecture. Going back to Dan Klyn\u2019s model, we\u2019ve practiced ontology by looking at the meaning of what we\u2019re intending to communicate:\n\nwe want to communicate there is an interactive element that toggles the appearance of an element on a page so we\u2019ve used one, a <button>, with those semantics.\nprogrammatically we\u2019ve used the type='button' attribute to signify that the button isn\u2019t a menu, reset, or submit element.\nvisually we\u2019ve designed our <button> look like something that can be interacted with and, importantly, we haven\u2019t removed the focus ring.\nwe\u2019ve labelled the <button> with the word \u201cSettings\u201d so that our users will hopefully understand what the button is for.\nwe\u2019ve used an <ul> element to structure and communicate our list of related items.\n\nWe\u2019ve also practiced taxonomy by developing systems and structures and creating relationships between our elements:\n\nby connecting the <button> to the panel using the aria-controls attribute we\u2019ve programmatically created a relationship between two elements.\nwe\u2019ve developed a structure in our elements by labelling our <ul> with the same name as the <button> that controls its appearance.\n\nAnd finally we\u2019ve practiced choreography by creating elements that foster movement and interaction. We\u2019ve anticipated the way users and information want to flow:\n\nwe\u2019ve used a <button> element that is interactive and accessible out of the box.\nour aria-controls attribute can help some people who use screen readers move easily from the <button> to the panel it controls.\nby toggling the value of the aria-expanded attribute we\u2019ve developed a system that tells assistive technology about the status of the relationship between our elements: the panel is visible or the panel is hidden.\nwe\u2019ve made sure our information is more usable and findable no matter how our users want or need to interact with it. Regardless of how someone \u201csees\u201d our work they\u2019re going to be able to use it because we\u2019ve architected multiple ways to access our information.\n\nInformation architecture, robust code, and accessibility\nThe United Nations estimates that around 10% of the world\u2019s population has some form of disability which, at the time of writing, is around 740,000,000 people. That\u2019s a lot of people who rely on well-architected semantic code that can be interpreted by whatever assistive technology they may need to use.\nIf everyone involved in the creation of our places made of information practiced information architecture it would make satisfying the WCAG 2.0 POUR principles so much easier. Our digital construction practices directly affect the quality of life of millions of people, and we have a responsibility to make technology available to them. In her book How To Make Sense Of Any Mess, Abby Covert states:\nIf we\u2019re going to be successful in this new world, we need to see information as a workable material and learn to architect it in a way that gets us to our goals.\nHow To Make Sense Of Any Mess\nI believe that the world will be a better place if we start treating front-end development as information architecture.", "year": "2016", "author": "Francis Storr", "author_slug": "francisstorr", "published": "2016-12-17T00:00:00+00:00", "url": "https://24ways.org/2016/front-end-developers-are-information-architects-too/", "topic": "code"}