angular-hybrid-forms

Angular Hybrid Forms

A form library implementing a template driven form methodology using Angular's Reactive Forms

This library is under active development and did has not reached its final form yet

Why?

Angular offers two form modules: Template-driven and Model-driven (aka Reactive forms).
Which one is better? The Angular team claims neither one is. Instead, you're encouraged to choose the approach that works best for you, weighing out their strengths and weaknesses.

When looking at Template-driven, the caveats manifest mostly around be able to reuse the form model in a parent child way.
When looking at Model-driven, we had to repeatedly update the model based on the form model in cases where we needed two way binding. We also hated the fact we had needed to build the form model programmatically, match it to the form control by name, bind to the valueChanges event programmatically...

Angular Hybrid Forms offers an opinionated compromise between the two, which nevertheless can configured and customized.
Consider the following template:

<form [formGroup]="formGroup">
  <input formControlName="username"/>
  ...
  <div formGroupName="address">
    <input formControlName="city"/>
    <input formControlName="street"/>
    <input formControlName="zipcode"/>
  </div>
  ...
      
</form>

For this to work, one would have to build the formGroup and formControl matching with a name in their component like so:

class App {
  private formGroup: FormGroup = new FormGroup({});
  private user: User;
  private address: Address;
  
  constructor() {
    formGroup.addControl('username', this.user.name);
    const addressFormGroup = new FormGroup({});
    addressFormGroup.addControl('city', address.city);
    addressFormGroup.addControl('street', address.street);
    addressFormGroup.addControl('zipcode', address.zipcode);
    formGroup.addControl('afress', addressFormGroup);
  }
}

Now, yes, we can clean this up a bit by using FormGroupBuilder. But what if we could get away with only the template, like the template driven approach, while still enabling us to combine and reuse forms like the model-driven approach?

Usage

With angular-hybrid-forms you can use this template:

<form [formGroup]="''|formGroup">
  <input formControlName="username" [formControlValue]="user.name"/>
  ...
  <div formGroupName="address">
    <input formControlName="city" [formControlValue]="address.city"/>
    <input formControlName="street" [formControlValue]="address.street"/>
    <input formControlName="zipcode" [formControlValue]="address.zipcode"/>
  </div>
  ...
      
</form>

With this component:

class App {
  private user: User;
  private address: Address;
}

Angular Hybrid Form will handle all the wiring of the FormGroups and FormControls for you!

You can of course apply any validations:

 <input formControlName="username" [formControlValue]="user.name" required/>

You can use two way binding:

 <input formControlName="username" [(formControlValue)]="user.name" required/>

Or in a more "reactive" mode:

 <input formControlName="username" [formControlName]="user.name" (formControlValue)="usernameChange($event)" required/>

We also expose a way for you to know when a change was made AND the value is valid, so you can save it:

 <input formControlName="username" [formControlName]="user.name" (formControlValidValue)="saveUsername($event)" required/>

We denounce this particular event emission by default, as you usually don't want to rush off and save on every change. You can change that if you feel like it doesn't work for you.

 <input formControlName="username" [formControlName]="user.name" (formControlValidValue)="saveUsername($event)" [formControlValidValueDelay]="0" required/>

It works with nested formGroupNames too!
You can also reuse your nested forms rather easily:

@Component({selector: 'address-form-group', 
template: `
  <div formGroupName="address">
      <input formControlName="city" [formControlValue]="address.city"/>
      <input formControlName="street" [formControlValue]="address.street"/>
      <input formControlName="zipcode" [formControlValue]="address.zipcode"/>
  </div>
`
})
class AdressFormGroup {
  @Input() address: Address;
}

@Component({selector: 'address-form-group', 
template: `
  <form [formGroup]="''|formGroup">
    <input formControlName="username" [formControlValue]="user.name"/>
    <address-form-group [address]="user.address">
    </address-form-group>
  </div>
`
})
class UserForm {
  @Input() user: User;
}

Build

Steps

Common tasks are present as npm scripts:

• npm start to run a live-reload server with the demo app
• npm run test to test in watch mode, or npm run test:once to only run once
• npm run build to build the library
• npm run lint to lint
• npm run clean to clean
• npm run integration to run the integration e2e tests
• npm install ./relative/path/to/lib after npm run build to test locally in another app

If you need to debug the integration app, please check ./integration/README.md.

The build step

You can build the library by running npm run build. This will generate a dist/ directory with all the entry points described above.

All the logic for creating the build can be found in ./build.js. It consists of roughly 5 steps:

• Compile with the AOT Compiler (AOT compiler or ngc) for ES5 and ES2015.
• Inline html and css inside the generated JavaScript files.
• Copy typings, metatada, html and css.
• Create each bundle using Rollup.
• Copy LICENSE, package.json and README.md files

Testing libraries

While testing is always important, it's especially important in libraries because consumer applications might break due to bugs in libraries.

But the fact that a library is consumed by another application is also what makes it hard to test.

To properly test a library, you need to have an integration tests. An integration test is to libraries what an end-to-end test is to applications. It tests how an app would install and use your library.

The QuickStart Library seed includes a directory called integration containing a standalone app that consumes your built library in both AOT and JIT modes, with end-to-end tests to verify it works.

To run the integration tests, do npm run integration which does the following:

• Build your library.
• Enter the integration app's directory.
• Install dependencies.
• Build the app in AOT mode.
• Test the app in AOT mode.
• Test the app in JIT mode.

Running integration tests gives you greater confidence that your library is properly built.

In addition to integration tests, you can also run unit tests in watch mode via npm run test, or single-run via npm run test:once.

You can also test your library by installing it in another local repository you have. To do so, first build your lib via npm run build. Then install it from your other repo using a relative path to the dist folder: npm install relative/path/to/library/dist.

Publishing your library

Every package on NPM has a unique name, and so should yours. If you haven't already, now is the time to change the name of your library.

Use your editor to search the project for all instances of angular-quickstart-lib and change it to your intended name (also in dash-case format). The library name is mentioned on at least these files:

• integration/src/app/app.component.ts
• integration/src/app/app.module.ts
• integration/src/systemjs.config.js
• integrations/package.json
• src/demo/app/app.component.ts
• src/demo/app/app.module.ts
• src/demo/systemjs.config.js
• src/demo/tsconfig.json
• src/lib/tsconfig.json
• src/lib/tsconfig.es5.json
• bs-config.json
• package.json
• README.md

You'll also need to rename the folder your project is in.

After you have changed the package name, you can publish it to NPM (read this link for details).

Instead of following the Updating the package on that previous doc, here we use standard-version. Read their docs to see how to use it.

First you'll need to create a NPM account and login on your local machine. Then you can publish your package by running npm publish dist/.
Since your package is built on the dist/ folder this is the one you should publish.

Be a good library maintainer

Now that you've published a library, you need to maintain it as well. Below are some of the most important points:

• Document your library.
• Keep an eye on the issue tracker.
• Manage your dependencies properly
• Follow Semantic Versioning
• Setup a Continuous Integration solution to test your library (included is a .travis.yml file for Travis CI)!
• Choose an appropriate license.

Appendix: Supporting AOT

AOT plays an important role in optimizing Angular applications. It's therefore important that third party libraries be published in a format compatible with AOT compilation. Otherwise it will not be possible to include the library in an AOT compiled application.

Only code written in TypeScript can be AOT compiled.

Before publishing the library must first be compiled using the AOT compiler (ngc). ngc extends the tsc compiler by adding extensions to support AOT compilation in addition to regular TypeScript compilation.

AOT compilation outputs three files that must be included in order to be compatible with AOT.

Transpiled JavaScript

As usual the original TypeScript is transpiled to regular JavaScript.

Typings files

JavaScript has no way of representing typings. In order to preserve the original typings, ngc will generate .d.ts typings files.

Meta Data JSON files

ngc outputs a metadata.json file for every Component and NgModule. These meta data files represent the information in the original NgModule and Component decorators.

The meta data may reference external templates or css files. These external files must be included with the library.

NgFactories

ngc generates a series of files with an .ngfactory suffix as well. These files represent the AOT compiled source, but should not be included with the published library.

Instead the ngc compiler in the consuming application will generate .ngfactory files based on the JavaScript, Typings and meta data shipped with the library.

Why not publish TypeScript?

Why not ship TypeScript source instead? After all the library will be part of another TypeScript compilation step when the library is imported by the consuming application.

Generally it's discouraged to ship TypeScript with third party libraries. It would require the consumer to replicate the complete build environment of the library. Not only typings, but potentially a specific version of ngc as well.

Publishing plain JavaScript with typings and meta data allows the consuming application to remain agnostic of the library's build environment.

Appendix: Supporting JIT

AOT compiled code is the preferred format for production builds, but due to the long compilation time it may not be practical to use AOT during development.

To create a more flexible developer experience a JIT compatible build of the library should be published as well. The format of the JIT bundle is umd, which stands for Universal Module Definition. Shipping the bundle as umd ensures compatibility with most common module loading formats.

The umd bundle will ship as a single file containing ES5 JavaScript and inlined versions of any external templates or css.

Appendix: Dependency Management

As a library maintainer, it's important to properly manage your dependencies in package.json.

There are three kinds of dependencies: dependencies, devDependencies and peerDependencies.

• dependencies: here go all the other libraries yours depends on when being used. A good way to figure out these is to go through your library source code (in src/lib only) and list all the libraries there.
• devDependencies: libraries that you need while developing, testing and building your library go here. When a user installs your library, these won't be installed. Users don't need to develop, build or test your library, they just need to run it.
• peerDependencies: these are similar to dependencies since your library expects them to be there at runtime. The difference is that you don't want to install a new version of these, but instead use the one already available.

A good example of a peer dependency is @angular/core and all other main Angular libraries. If you listed these in dependencies, a new one - with a different version! - could be installed for your library to use. This isn't what you wanted though. You want your library to use the exact same @angular/core that the app is using.

You'll usually used @angular/* libraries listed in both devDependencies and peerDependencies. This is normal and expected, because when you're developing your library also need a copy of them installed.

Another thing to remember is to keep your dependencies from changing too much unexpectedly. Different versions of libraries can have different features, and if you inadvertently are too lenient with allowed versions your library might stop working because a dependency changed.

You can choose what versions you allow by using ranges.

A good rule of thumb is to have all dependencies specified with a tilde ~(~1.2.3), while your peerDependencies have a range ("@angular/core": ">=4.0.0 <5.0.0 || >=4.0.0-beta <5.0.0").

Any extra dependency or peer dependency that you add to package.json should also be added to the globals and external array in the rollupBaseConfig variable in ./build.js.

This ensures your library doesn't package extra libraries inside of it and instead uses the ones available in the consuming app.