Frameworks are artefacts of a conversation

In Six tips on front-end framework selection I talked a bit about how to go about choosing a front-end framework for a new project.

So, what is the bigger picture here. What are frameworks all about?

I think that frameworks are solidified artefacts from an ongoing conversation we are all having about how to develop products more effectively, easily, cheaply, or whateverly.

Artefacts from a ongoing conversation.

Over time, we develop stuff, and aim to improve the how of the development. In this process, people solidify ideas into frameworks at various points. They meet some shared pain or add an exciting capability and a frameworks becomes popular. People try it out. Their learning feeds the minds of those building the next framework: introducing new metaphors, tools, style and syntax as they go.

These artefacty frameworks also contain reactions to the environment around — available browser performance or particular browser features lead to new ideas being tested in a framework.

That is all I wanted to say: Using, testing, moaning, exclaiming about a framework are all part of the conversation we have to have. New frameworks follow old ones. The are built from previous experiments and ideas.

Why does this matter? Perspective helps make good decisions.

You could deny it, but every front-end framework is thus made of jQuery. From the jQuery experience, annoyances and inflexibility we all dealt with, we have the beginnings of the modern browser framework. jQuery informs AngularJS. AngularJS informs React. And so on.

Six tips on front-end framework selection

React or Angular, or VueJS, or whatever?  What do I choose for this next app I'm building?

Framework selection isn't easy. Here I want to put down my thoughts on what I think matters.

1. Complex things are complex

Firstly, let's be honest about trivial examples. They are trivial.  Nothing I've encountered in building web apps is anywhere near as simple as a To Do app.  TodoMVC is lovely for comparing framework features BUT any non-trivial app will push the boundaries of the framework and build process and comparing via a simple SPA comparison won't be that helpful.

No matter what you choose there will be always be too much boilerplate code and too many untidy corners where it doesn't work nicely. One or more of the framework's metaphors will suck.  That will all just happen.  So, rather than look to the framework to solve your complexity, look how it can support you through it.  

How does it scale to many many components and pages? How easy is it to test within and write tests for?

2. Fashion

Fashion drives a lot of our choices.  One of the wonderful things about the web industry is that there is plenty of new to play with, and there is an ongoing 'conversation' between all the developers on how to build web pages and web applications.  As a part of the conversation new tools are developed, languages are enhanced,  and new ideas emerge.

There is a massive temptation to try the latest thing, and the latest thing sells itself as the best yet.  However, these are all steps in an ongoing conversation we developers are having about how to do things. New frameworks and new versions of existing frameworks are really just a few more sentences in the conversation.  There will be something more amazing next week.  

So, keeping the choices in this sort of context really helps with decision making.  It is worth calling out the fashion aspect as you are deciding.  At least be honest if you just want to do the latest thing :-)

3. Nothing lasts that long

Ideas get old. Things stop working. Support evaporates.  As a part of this drive to the future with tools and frameworks, this will happen to your favourite framework, language, operating system.  When it happens it may require you to upgrade, mothball, re-build etc.  This will probably happen sooner then you expect and will be complicated by dependencies. 

Especially: don't expect stuff to just sit in npm for ever.  Stuff will evaporate.  Have a plan.

4. Skills

I've just finished a re-building an app into Angular 2 from earlier Angular 1 app with a weird backend Java CMS.  Moving to Angular 2+ pretty much mandated TypeScript instead of JavaScript.  Now, guess what happened?  TypeScript typing looks a lot like Java - and it became pretty easy for the Java devs to join in the front end dev using TypeScript. That was unexpected and really helped the project along.  

My point here is:  what skills do you have, and therefore what do you choose?   Might be fun to take on the new thing, but it you are therefore niche-ing into tech that has little support from the people around you, or if it is nice or fashionable enough it is going to be hard to hire people.    The easier and less exciting path (which includes leaving work earlier each night) might be to pick something better supported around you, rather than the bleeding edge.

5. Appreciate simplicity, plan for difficulty; read the source

Framework simplicity is valuable; a framework that simply solves your problem is golden.  Given that your non-trivial application will get complex or have some sort of twisty corner in the URL scheme that doesn't quite work with the router, you'll end up beyond the documentation reading the source.  So why not start now?   I encourage you to code review at least some of the whole framework before you start using it.  You'll end up looking in the source anyway, so why not use this as a part of your evaluation.

6. Or just don't

Finally, maybe you don't need much framework, and maybe none.  If your page is free of routing, relatively simple, then why not consider the no-framework approach?  However, you are going to need good patterns and coding standard and testing or you may well create an organic mess.  Remember, a framework is holding you to certain patterns, so you'll need to decide on and apply your own to keep stuff maintainable.

I hope this is useful. Feel free to comment below on your framework choices, or contact me if you'd like some help or chat about framework selection.

My recent framework selections (in reverse time order) look like this: Angular 2+, No framework, Angular 1.5 + material, No framework, Angular 1.x.

Review: Nearly six months with Angular 2

Angular 2 is at rc.5 as I write this.  There is not quite a 2.0 release.

After nearly six months working with the various release candidates, here's my review of using using Angular 2 to largely re-build an Angular 1 app and integrate it into a mixed Angular 1 and 2 environment.    This will give you a high level overview of what is different in building an app for Angular 2 and serve as a launch pad into your own research.

This article will be helpful if you are looking on starting a new project and trying to work out whether to start with Angular 2, migrate from 1 to 2, go React, or something else.


TLDR;   There’s lots of good stuff In Angular 2: Components are first class; JavaScript modules and classes are used everywhere; Promises are replaced by RxJS.  You can migrate from Angular 1 reasonably easily. Write in TypeScript, Dart or JavaScript.  Basically Solid. Expect a few small changes before release. More software engineering than browser hacking.


What is all this about Components?

Each generation of web framework brings something new to help with the more complex challenges we face in software engineering in the browser.  Where Angular 1 arose to answer the question “How do I build a complete fully-featured single-page app?”, both Angular 2 and React both arrive to answer the additional question  “How do I build a complete fully-featured single-page app in a maintainable way, using real software engineering?” Components just seem like the obvious next step as we build more complex interfaces made of many parts.  React does this in a less bundled chose-your-own-bits way; Angular 2 in a bit more complete framework kind of way.

I'd say you really want to build component-based if you are doing anything more than making a quick one page website.  You need a component-based approach If you are looking to re-use bits of your app, work with a style guide, or have more than trivial QA/testing.  

As implemented by Angular 2, a component can contain code (JavaScript, TypeScript or Dart), an HTML Template, and one or more CSS files. HTML and CSS can be inline text in the component source or referenced files that are imported at compile time.  The way compiling and bundling works, you end up with these assets specific to the component embedded in the compiled component itself.   Nice.  Makes bundling easy, makes it easy to work on a single bit.

So, your component ends up with an HTML template string compiled in and if you choose, an array of CSS strings compiled in.  When the component is rendered by Angular 2, the component’s CSS is written to a head style tag and targeted with a unique attribute for this component only.

This does make it nice and easy to pick up a component, satisfy its dependencies and test it as an individual item.

Less a Global Soup

Angular 1 always gave me an uneasy feeling. The lack of real modules in JavaScript coupled with all these named strings, plus the sometimes-dashed sometimeCamelcase naming system left a lot of room for duh! Errors, conflicts and confusion.

This has all changed.  You have to be explicit in component metadata about dependencies, directives to be used in templates, services to be provided, input and outputs.   Up until rc.5 that could mean quite a lot of metadata for a non-trivial component, however this latest release candidate provides (and requires use of) an angular Module abstraction to collect a bunch of services, directives etc together.  This looks good though I haven’t tried it yet.

Less global soup also means you can bootstrap multiple angular apps in one browser completely separately, or even bootstrap Angular 1 and 2 apps separately alongside each other.  It it easy to confident about this with Angular 2 because it just doesn’t appear to leak stuff all over the browser global space.

And, biggest of all, the need to be explicit, especially with the inputs and outputs of components, means the rendering is targeted on what has changed only.  Therefore it is faster.  We haven't measurd it, but it looks much quicker than Angular 1 for the same complex pages.

Using a language: TypeScript or Dart or JavaScript

Going from Angular 1 to 2 is a bit like going from jQuery to Angular 1.   There is a lot more formality in the framework and it feels a lot more like software and a lot less like hacking up something quick in the browser.  This just makes sense as the complexity of single page apps increase, and more and more logic moves into the front-end.  To handle this increased complexity, you have a choice of front-end languages: TypeScript, Google’s own Dart, or JavaScript.

Both TypeScript and Dart are typed languages.  This may be a shock for those in love with JavaScript’s nearly unlimited expressibility and type mutability, but y’know, it does make life easier for two reasons I can immediately see.  One: strong typing takes the pressure off unit tests -- you’ll catch a lot of errors simply by using types when compiling -- all those ‘this property should be in this object’ or 'this function should be called with this' are caught by the compiler.  This cuts down on a lot of boilerplate unit tests that you just don’t need. And secondly:  developers coming from, say, Java, and having a little JavaScript awareness will find particularly TypeScript a pretty comfortable language to move into the front-end with.

The Angular 2 docs samples have been delivered in the order of TypeScript, Dart, and then JavaScript.  So Typescript was an obvious place to start for me.   The Typescript tooling is solid.  It cross-compiles to JavaScript well, has a proper strict linter etc. There were the usual new-language annoyances in the beginning, but the results in the project are very very few in-code bugs we’re debugging in the page.

If you really do want to go with JavaScript -- I would not advise that -- you either need to be using ES6 or later or get very used to coding up module and class patterns by hand.

Going Reactive: Promises replaced by Reactive RxJS streams

The next step beyond promises has been taken up.  Most places where you’d expect a promise to be returned (eg http service calls) with an async. result will now provide an RxJS stream to subscribe to.  At the basic level, this is like a Promise that can return multiple values in a stream.  Like a then called multiple times.  There’s a brain wrench in here, a slight syntax change over promises, and done carefully you can end up with an app that has flexibility in the time domain. Connect it to websockets and it keeps changing as the inputs change.

An early design decision in this app we’re working on was to build a subject-based configuration system using RxJS Subjects.  This means that we can in real-time change configuration inside the app and it will change form as it needs to.   Change detection and publishing the change is all done by RxJS -- all our app has to do is subscribe to and listen for config changes.  We use this config to  propagate view changes, query results and display size changes amongst other things and the app just responds as it happens.   There’s no writing code to make things dynamic as such.   It will hurt the head initially if you haven't dealt with it before. This guide helps.

My Overall Review: The Gut Feel

Too many words already, so a bit of my gut feel to finish off.    This is a good framework and at rc.4 it is solid.  It feels maintainable. It is expressive, in that it is easy to get stuff done in the templates and in the component TypeScript itself.   There is a step up in seriousness -- be ready for that or go for something lighter-weight.  It is fast compared to Angular 1.  The reactive feel is enjoyable.

I’d pick it up for the next reasonably-sized project or maybe even a small one.

Where to go next?

I'd suggest from here you dive into the Angular 2 documentation over at angular.io, and try out the quickstart.


Angular 2 - useful links

Angular 2 is a release candidate: rc.1 as I write this.  The official documents still have some missing pieces.  Here are a collection of useful links and my own usage comments to help fill in the missing bits:

Official Docs

Official latest Angular 2 docs - fairly good. Some bits missing as of rc.1

Official TypeScript docs - very good - starts with basic types selected

Unit Testing

Unit Testing Recipes - useful recipes for testing Components, Directives, HTTP Mocking etc - though think this conflicts with the latest I've found on async injecting into unit tests.  I might be wrong.

Testing Angular 2 Components with Unit Tests and the TestComponentBuilder (RC1+) - -but see this repo for tests actually updated for RC1: https://github.com/krimple/angular2-unittest-samples-rc    --- note that TestComponentBuilder has been moved from @angular/core/testing to @angular/compiler/testing.

Touch events

Haven't tried this yet -- not sure what really works. It does seem that Hammer.js must be loaded before angular2.

Bootstrap Carousel

..I'll add more as I revisit links.. First week or two with Angular 2 involved a lot of reading source and searching for solutions..