angular-docs-cn/aio/content/guide/form-validation.md

@title Form Validation

@intro Validate user's form entries.

@description

{@a top}

Improve overall data quality by validating user input for accuracy and completeness.

This cookbook shows how to validate user input in the UI and display useful validation messages using first the template-driven forms and then the reactive forms approach.

Read more about these choices in the Forms and the Reactive Forms guides.

{@a toc}

Contents

• Simple template-driven forms

• Template-driven forms with validation messages in code

  • Component Class
  • The benefits of messages in code
  • FormModule and template-driven forms

• Reactive forms with validation in code

  • Switch to the ReactiveFormsModule

  • Component template

  • Component class

    • FormBuilder declaration
    • Committing hero value changes

• Custom validation

  • Custom validation directive

• Testing considerations

{@a live-example}

Try the live example to see and download the full cookbook source code.

{@a template1}

Simple template-driven forms

In the template-driven approach, you arrange form elements in the component's template.

You add Angular form directives (mostly directives beginning ng...) to help Angular construct a corresponding internal control model that implements form functionality. In template-drive forms, the control model is implicit in the template.

To validate user input, you add HTML validation attributes to the elements. Angular interprets those as well, adding validator functions to the control model.

Angular exposes information about the state of the controls including whether the user has "touched" the control or made changes and if the control values are valid.

In this first template validation example, notice the HTML that reads the control state and updates the display appropriately. Here's an excerpt from the template HTML for a single input control bound to the hero name:

Note the following:

• The <input> element carries the HTML validation attributes: required, minlength, and maxlength.

• The name attribute of the input is set to "name" so Angular can track this input element and associate it with an Angular form control called name in its internal control model.

• The [(ngModel)] directive allows two-way data binding between the input box to the hero.name property.

• The template variable (#name) has the value "ngModel" (always ngModel). This gives you a reference to the Angular NgModel directive associated with this control that you can use in the template to check for control states such as valid and dirty.

• The *ngIf on the <div> element reveals a set of nested message divs but only if there are "name" errors and the control is either dirty or touched.

• Each nested <div> can present a custom message for one of the possible validation errors. There are messages for required, minlength, and maxlength.

The full template repeats this kind of layout for each data entry control on the form.

{@a why-check}

Why check dirty and touched?

The app shouldn't show errors for a new hero before the user has had a chance to edit the value. The checks for dirty and touched prevent premature display of errors.

Learn about dirty and touched in the Forms guide.

The component class manages the hero model used in the data binding as well as other code to support the view.

Use this template-driven validation technique when working with static forms with simple, standard validation rules.

Here are the complete files for the first version of HeroFormTemplateCompononent in the template-driven approach:

{@a template2}

Template-driven forms with validation messages in code

While the layout is straightforward, there are obvious shortcomings with the way it's handling validation messages:

• It takes a lot of HTML to represent all possible error conditions. This gets out of hand when there are many controls and many validation rules.

• There's a lot of JavaScript logic in the HTML.

• The messages are static strings, hard-coded into the template. It's easier to maintain dynamic messages in the component class.

In this example, you can move the logic and the messages into the component with a few changes to the template and component.

Here's the hero name again, excerpted from the revised template (Template 2), next to the original version:

The <input> element HTML is almost the same. There are noteworthy differences:

• The hard-code error message <divs> are gone.

• There's a new attribute, forbiddenName, that is actually a custom validation directive. It invalidates the control if the user enters "bob" in the name <input>(try it). See the custom validation section later in this cookbook for more information on custom validation directives.

• The #name template variable is gone because the app no longer refers to the Angular control for this element.

• Binding to the new formErrors.name property is sufficent to display all name validation error messages.

{@a component-class}

Component class

The original component code for Template 1 stayed the same; however, Template 2 requires some changes in the component. This section covers the code necessary in Template 2's component class to acquire the Angular form control and compose error messages.

The first step is to acquire the form control that Angular created from the template by querying for it.

Look back at the top of the component template at the #heroForm template variable in the <form> element:

The heroForm variable is a reference to the control model that Angular derived from the template. Tell Angular to inject that model into the component class's currentForm property using a @ViewChild query:

Some observations:

• Angular @ViewChild queries for a template variable when you pass it the name of that variable as a string ('heroForm' in this case).

• The heroForm object changes several times during the life of the component, most notably when you add a new hero. Periodically inspecting it reveals these changes.

• Angular calls the ngAfterViewChecked lifecycle hook method when anything changes in the view. That's the right time to see if there's a new heroForm object.

• When there is a new heroForm model, formChanged() subscribes to its valueChanges Observable property. The onValueChanged handler looks for validation errors after every keystroke.

The onValueChanged handler interprets user data entry. The data object passed into the handler contains the current element values. The handler ignores them. Instead, it iterates over the fields of the component's formErrors object.

The formErrors is a dictionary of the hero fields that have validation rules and their current error messages. Only two hero properties have validation rules, name and power. The messages are empty strings when the hero data are valid.

For each field, the onValueChanged handler does the following:

• Clears the prior error message, if any.
• Acquires the field's corresponding Angular form control.
• If such a control exists and it's been changed ("dirty") and it's invalid, the handler composes a consolidated error message for all of the control's errors.

Next, the component needs some error messages of course—a set for each validated property with one message per validation rule:

Now every time the user makes a change, the onValueChanged handler checks for validation errors and produces messages accordingly.

{@a improvement}

The benefits of messages in code

Clearly the template got substantially smaller while the component code got substantially larger. It's not easy to see the benefit when there are just three fields and only two of them have validation rules.

Consider what happens as the number of validated fields and rules increases. In general, HTML is harder to read and maintain than code. The initial template was already large and threatening to get rapidly worse with the addition of more validation message <div> elements.

After moving the validation messaging to the component, the template grows more slowly and proportionally. Each field has approximately the same number of lines no matter its number of validation rules. The component also grows proportionally, at the rate of one line per validated field and one line per validation message.

Both trends are manageable.

Now that the messages are in code, you have more flexibility and can compose messages more efficiently. You can refactor the messages out of the component, perhaps to a service class that retrieves them from the server. In short, there are more opportunities to improve message handling now that text and logic have moved from template to code.

{@a formmodule}

FormModule and template-driven forms

Angular has two different forms modules—FormsModule and ReactiveFormsModule—that correspond with the two approaches to form development. Both modules come from the same @angular/forms library package.

You've been reviewing the "Template-driven" approach which requires the FormsModule. Here's how you imported it in the HeroFormTemplateModule.

This guide hasn't talked about the SharedModule or its SubmittedComponent which appears at the bottom of every form template in this cookbook.

They're not germane to the validation story. Look at the live example if you're interested.

{@a reactive}

Reactive forms with validation in code

In the template-driven approach, you markup the template with form elements, validation attributes, and ng... directives from the Angular FormsModule. At runtime, Angular interprets the template and derives its form control model.

Reactive Forms takes a different approach. You create the form control model in code. You write the template with form elements and form... directives from the Angular ReactiveFormsModule. At runtime, Angular binds the template elements to your control model based on your instructions.

This approach requires a bit more effort. You have to write the control model and manage it.

This allows you to do the following:

• Add, change, and remove validation functions on the fly.
• Manipulate the control model dynamically from within the component.
• Test validation and control logic with isolated unit tests.

The following cookbook sample re-writes the hero form in reactive forms style.

{@a reactive-forms-module}

Switch to the ReactiveFormsModule

The reactive forms classes and directives come from the Angular ReactiveFormsModule, not the FormsModule. The application module for the reactive forms feature in this sample looks like this:

The reactive forms feature module and component are in the src/app/reactive folder. Focus on the HeroFormReactiveComponent there, starting with its template.

{@a reactive-component-template}

Component template

Begin by changing the <form> tag so that it binds the Angular formGroup directive in the template to the heroForm property in the component class. The heroForm is the control model that the component class builds and maintains.

Next, modify the template HTML elements to match the reactive forms style. Here is the "name" portion of the template again, revised for reactive forms and compared with the template-driven version:

Key changes are:

• The validation attributes are gone (except required) because validating happens in code.

• required remains, not for validation purposes (that's in the code), but rather for css styling and accessibility.

A future version of reactive forms will add the required HTML validation attribute to the DOM element (and perhaps the aria-required attribute) when the control has the required validator function.

Until then, apply the required attribute and add the Validator.required function to the control model, as you'll see below.

• The formControlName replaces the name attribute; it serves the same purpose of correlating the input with the Angular form control.

• The two-way [(ngModel)] binding is gone. The reactive approach does not use data binding to move data into and out of the form controls. That's all in code.

The retreat from data binding is a principle of the reactive paradigm rather than a technical limitation.

{@a reactive-component-class}

Component class

The component class is now responsible for defining and managing the form control model.

Angular no longer derives the control model from the template so you can no longer query for it. You can create the Angular form control model explicitly with the help of the FormBuilder class.

Here's the section of code devoted to that process, paired with the template-driven code it replaces:

• Inject FormBuilder in a constructor.

• Call a buildForm method in the ngOnInit lifecycle hook method because that's when you'll have the hero data. Call it again in the addHero method.

A real app would retrieve the hero asynchronously from a data service, a task best performed in the ngOnInit hook.

• The buildForm method uses the FormBuilder, fb, to declare the form control model. Then it attaches the same onValueChanged handler (there's a one line difference) to the form's valueChanges event and calls it immediately to set error messages for the new control model.

{@a formbuilder}

FormBuilder declaration

The FormBuilder declaration object specifies the three controls of the sample's hero form.

Each control spec is a control name with an array value. The first array element is the current value of the corresponding hero field. The optional second value is a validator function or an array of validator functions.

Most of the validator functions are stock validators provided by Angular as static methods of the Validators class. Angular has stock validators that correspond to the standard HTML validation attributes.

The forbiddenNames validator on the "name" control is a custom validator, discussed in a separate section below.

Learn more about FormBuilder in the Introduction to FormBuilder section of Reactive Forms guide.

{@a committing-changes}

Committing hero value changes

In two-way data binding, the user's changes flow automatically from the controls back to the data model properties. Reactive forms do not use data binding to update data model properties. The developer decides when and how to update the data model from control values.

This sample updates the model twice:

1. When the user submits the form.
2. When the user adds a new hero.

The onSubmit() method simply replaces the hero object with the combined values of the form:

This example is lucky in that the heroForm.value properties just happen to correspond exactly to the hero data object properties.

The addHero() method discards pending changes and creates a brand new hero model object.

Then it calls buildForm() again which replaces the previous heroForm control model with a new one. The <form> tag's [formGroup] binding refreshes the page with the new control model.

Here's the complete reactive component file, compared to the two template-driven component files.

Run the live example to see how the reactive form behaves, and to compare all of the files in this cookbook sample.

{@a custom-validation}

Custom validation

This cookbook sample has a custom forbiddenNamevalidator() function that's applied to both the template-driven and the reactive form controls. It's in the src/app/shared folder and declared in the SharedModule.

Here's the forbiddenNamevalidator() function:

The function is actually a factory that takes a regular expression to detect a specific forbidden name and returns a validator function.

In this sample, the forbidden name is "bob"; the validator rejects any hero name containing "bob". Elsewhere it could reject "alice" or any name that the configuring regular expression matches.

The forbiddenNameValidator factory returns the configured validator function. That function takes an Angular control object and returns either null if the control value is valid or a validation error object. The validation error object typically has a property whose name is the validation key, 'forbiddenName', and whose value is an arbitrary dictionary of values that you could insert into an error message ({name}).

{@a custom-validation-directive}

Custom validation directive

In the reactive forms component, the 'name' control's validator function list has a forbiddenNameValidator at the bottom.

In the template-driven example, the <input> has the selector (forbiddenName) of a custom attribute directive, which rejects "bob".

The corresponding ForbiddenValidatorDirective is a wrapper around the forbiddenNameValidator.

Angular forms recognizes the directive's role in the validation process because the directive registers itself with the NG_VALIDATORS provider, a provider with an extensible collection of validation directives.

Here is the rest of the directive to help you get an idea of how it all comes together:

If you are familiar with Angular validations, you may have noticed that the custom validation directive is instantiated with useExisting rather than useClass. The registered validator must be this instance of the ForbiddenValidatorDirective—the instance in the form with its forbiddenName property bound to “bob". If you were to replace useExisting with useClass, then you’d be registering a new class instance, one that doesn’t have a forbiddenName.

To see this in action, run the example and then type “bob” in the name of Hero Form 2. Notice that you get a validation error. Now change from useExisting to useClass and try again. This time, when you type “bob”, there's no "bob" error message.

For more information on attaching behavior to elements, see Attribute Directives.

{@a testing}

Testing Considerations

You can write isolated unit tests of validation and control logic in Reactive Forms.

Isolated unit tests probe the component class directly, independent of its interactions with its template, the DOM, other dependencies, or Angular itself.

Such tests have minimal setup, are quick to write, and easy to maintain. They do not require the Angular TestBed or asynchronous testing practices.

That's not possible with template-driven forms. The template-driven approach relies on Angular to produce the control model and to derive validation rules from the HTML validation attributes. You must use the Angular TestBed to create component test instances, write asynchronous tests, and interact with the DOM.

While not difficult, this takes more time, work and skill—factors that tend to diminish test code coverage and quality.