What is the differences between these two and what are the use cases for each?
The docs aren't exactly helpful:
forRoot creates a module that contains all the directives, the given
routes, and the router service itself.
forChild creates a module that
contains all the directives and the given routes, but does not include
the router service.
My vague guess is that one is for the 'main' module and the other is for any imported modules (since they would already have the service available from the main module), but I can't really think of a use case.
I strongly suggest reading this article:
- Avoiding common confusions with modules in Angular
Module with providers
When you import a module you usually use a reference to the module class:
@NgModule({
providers: [AService]
})
export class A {}
-----------------------------------
@NgModule({
imports: [A]
})
export class B
In this way all providers registered on module A
will be added to the root injector and available for the entire application.
But there is another way to register a module with providers like this:
@NgModule({
providers: [AService]
})
class A {}
export const moduleWithProviders = {
ngModule: A,
providers: [AService]
};
----------------------
@NgModule({
imports: [moduleWithProviders]
})
export class B
This has the same implications as the previous one.
You probably know that lazy loaded modules have their own injector. So suppose you want to register AService
to be available for the entire application, but some BService
to be available to only lazy loaded modules. You can refactor your module like this:
@NgModule({
providers: [AService]
})
class A {}
export const moduleWithProvidersForRoot = {
ngModule: A,
providers: [AService]
};
export const moduleWithProvidersForChild = {
ngModule: A,
providers: [BService]
};
------------------------------------------
@NgModule({
imports: [moduleWithProvidersForRoot]
})
export class B
// lazy loaded module
@NgModule({
imports: [moduleWithProvidersForChild]
})
export class C
Now BService
will only be available for child lazy loaded modules and AService
will be available for the entire application.
You can rewrite the above as an exported module like this:
@NgModule({
providers: [AService]
})
class A {
forRoot() {
return {
ngModule: A,
providers: [AService]
}
}
forChild() {
return {
ngModule: A,
providers: [BService]
}
}
}
--------------------------------------
@NgModule({
imports: [A.forRoot()]
})
export class B
// lazy loaded module
@NgModule({
imports: [A.forChild()]
})
export class C
How is that relevant to RouterModule?
Suppose they are both accessed using the same token:
export const moduleWithProvidersForRoot = {
ngModule: A,
providers: [{provide: token, useClass: AService}]
};
export const moduleWithProvidersForChild = {
ngModule: A,
providers: [{provide: token, useClass: BService}]
};
With separate configurations when you request token
from a lazy loaded module you will get BService
just as planned.
RouterModule uses ROUTES
token to get all routes specific to a module. Since it wants routes specific to lazy loaded module to be available inside this module (analogues to our BService) it uses different configuration for the lazy loaded child modules:
static forChild(routes: Routes): ModuleWithProviders {
return {
ngModule: RouterModule,
providers: [{provide: ROUTES, multi: true, useValue: routes}}]
}
}
Documentation clearly states what is the purpose of this distinction here: https://angular.io/docs/ts/latest/guide/ngmodule.html#!#core-for-root
Call forRoot only in the root application module, AppModule. Calling it in any other module, particularly in a lazy loaded module, is contrary to the intent and is likely to produce a runtime error.
Remember to import the result; don't add it to any other @NgModule list.
Every application has exactly one starting point (root) where the main routing service should be initialized with forRoot
, while routes for particular "child" features should be registered additionaly with forChild
. It is extremely useful for submodules and lazy loaded modules which do not have to be loaded at the application start, and as @Harry Ninh said they are told to reuse RouterService instead of registration of the new service, which may cause a runtime error.
I think the answers are right but I think something is missing.
The thing which is missing is "why and what it solves ?".
Ok let's start.
First let's mention some info:
All modules has access to the root services.
So even lazy loaded modules can use a service which was provided in app.module
.
What will happen if a lazy loaded module will provide to itself a service which the app module already provided ? there will be 2 instances.
It's not a problem but sometimes it is.
How can we solve it ? simply don't import a module with that provider to lazy loaded modules.
End of story.
This ^ was just to show that lazy loaded modules has their own injection point ( as opposed to non-lazy-loaded modules).
But what happens when a shared(!) module has declared providers
, and that module is imported by lazy and app.module
? Again , like we said , two instances.
So how can we solve this in the shared module POV ? we need a way not to use providers:[]
! Why ? becuase they will be auto imported to both consuming lazy and app.module and we don't want that as we saw that each will have a different instance.
Well , it turns out that we can declare a shared module that won't have providers:[]
, but still , will provide prodivers ( sorry :))
How ? Like this :
Notice , no providers.
But
Entering Manual mechanism via convention :
You will notice that the providers in the pictures have service1
and service2
This allows us to import service2
for lazy loaded modules and service1
for non-lazy modules. ( cough...router....cough)
BTW , no one is stopping you to call forRoot
within a lazy module. but you will have 2 instances becuase app.module
should also do it - so don't do it in lazy modules.
Also - if app.module
calls forRoot
( and no one calls forchild
) - that's fine , but root injector will only have service1
. ( available to all app)
So why do we need it ? I'd say :
It allows a shared module , to be able to split its
different-providers to be used with eager modules and lazy modules -
via forRoot
and forChild
convention. I repeat : convention
That's it.
WAIT !! not a single word about singleton ?? so why do I read
singleton everywhere ?
Well - it's hidden in the sentence above ^
It allows a shared module , to be able to split its
different-providers to be used with eager modules and lazy modules -
via forRoot and forChild.
The convention (!!!) allows it to be singleton - or to be more precise - if you won't follow the convention - you will NOT get a singleton.
So if you only load forRoot
in the app.module
, then you get only one instance becuase you only should call forRoot
it in the app.module
.
BTW - at this point you can forget about forChild
. the lazy loaded module shouldn't / won't call forRoot
- so you're safe in POV of singleton.
forRoot and forChild are not one unbreakable package - it's just that there is no point of calling for Root which obviously will be loaded only in app.module
without giving the ability for lazy modules , have their own services , without creating new services-which-should-be-singleton.
This convention give you a nice ability called forChild
- to consume "services only for lazy loaded modules".
Here is a demo Root providers yields positive numbers , lazy loaded modules yields negative numbers.
https://www.freelancermap.com/freelancer-tips/12255-forroot-forchild-angular
This is an article about using forRoot and forChild in Angular; I'm not particularly versed on the subject, but perhaps it can be helpful!