In some cases you might experience that a circular dependency is preventing your Umbraco installing from starting up.
An example of this could be a circular dependency on IUmbracoContextFactory
. This would happen if your service interacts with third-party code that also depends on an IUmbracoContextFactory
instance.
In this situation, you can request a lazy version of the dependency so it won't evaluate during boot, only when accessed:
Umbraco has a range of 'Core' Services and Helpers that act as a 'gateway' to Umbraco data and functionality to use when extending or implementing an Umbraco site
Umbraco has a range of 'Core' Services and Helpers that act as a 'gateway' to Umbraco data and functionality to use when extending or implementing an Umbraco site.
The general rule of thumb is that management Services provide access to allow the modification of Umbraco data (and therefore aren't optimised for displaying data). Helpers on the other hand provide access to readonly data with performance of displaying data taken into consideration.
Although there is a management Service named the IContentService
- only use this to modify content - do not use the IContentService
in a View/Template to pull back data to display, this will make requests to the database and be slow - here instead inject the IPublishedContentQueryAccessor
interface and get the IPublishedContentQuery
that operate against a cache of published content items, and are significantly quicker.
The management Services and Helpers are all registered with Umbraco's underlying DI framework. This article aims to show examples of gaining access to utilise these resources in multiple different scenarios. There are subtle differences to be aware of depending on what part of Umbraco is being extended.
This article will also suggest how to follow a similar pattern to encapsulate custom 'site specific' implementation logic, in similar services and helpers, registered with the underlying DI contain. This would be to avoid repetition and promote consistency and readability within an Umbraco site solution.
Inside a view/template or partial view, access is also provided by the DI framework, by using the @inject
keyword.
Inside a custom Controller access is provided to Services via the Services
property (ServiceContext) and the UmbracoHelper
via the Umbraco
property (UmbracoHelper).
Controllers and Views can access an IUmbracoContext
by injecting the IUmbracoContextAccessor
, however this is not always the case 'everywhere in Umbraco', for example common extension points: Components,ContentFinders or Custom C# Classes.
IUmbracoContext, UmbracoHelper, IPublishedContentQuery - are all based on an HttpRequest - their lifetime is controlled by an HttpRequest. So if you are not operating within an actual request, you cannot inject these parameters and if you try to ... Umbraco will report an error on startup.
It's possible to inject management Services that do not rely on the UmbracoContext
into the constructor of a component. This example shows injecting the IMediaService
in a Notification Handler to create a corresponding Media Folder for every 'landing page' that is saved in the Content Section, by subscribing to the 'Content Saved' notification.
See documentation on Composing for further examples and information on Components and Composition.
Trying to inject types that are based on an Http Request such as UmbracoHelper
or IPublishedContentQuery
into classes that are not based on an Http Request will trigger an error. However, there is a technique that allows the querying of the Umbraco Published Content, using the UmbracoContextFactory
and calling EnsureUmbracoContext()
.
In this example, when a page is unpublished, instead of a 404 occurring for the content when the url is requested in the future, we might want to serve a 410 'page gone' status code instead. We handle the Unpublishing notification of the ContentService, access the Published Content Cache, determine it's 'published url' and then store for later use in any 'serving the 410' mechanism.
An IContentFinder could be placed in the ContentFinder ordered collection, right before a 404 is served. This could be done to lookup the incoming request against the stored location of 410 urls, and serve the 410 status request code if a match is found for the previously published item.
Inside a ContentFinder access to the content cache is possible by injecting IUmbracoContextAccessor
into the constructor and provided via the PublishedRequest object:
And inside an IPublishedUrlProvider
injection of IUmbracoContextAccessor
into the constructor is also possible.
It is still possible to inject services into IContentFinder's. IContentFinders are singletons, but the example is showing you do not 'need to' in order to access the Published Content Cache.
When implementing an Umbraco site, it is likely to have to execute similar code that accesses or operates on Umbraco data, in multiple places, perhaps using the core management Services or Umbraco Helpers.
For example; Getting a list of the latest News Articles, or building a link to the site's News Section or Contact Us page. Repeating this kind of logic in multiple places, Views, Partial Views / Controllers etc, is possible, but it's generally considered good practice to consolidate this logic into a single place.
One option is to add 'Extension Methods' to the UmbracoHelper
class or IPublishedContentQuery
interface.
Anywhere there is reference to the UmbracoHelper
or IPublishedContentQuery
and a reference is added to the namespace the extension belongs to, it is possible to call the method by writing _publishedContentQuery.GetNewsSection()
.
Another option, is to make use of the underlying DI framework, and create custom Services and Helpers, that in turn can have the 'core' management Services and Umbraco Helpers injected into them.
This approach enables the grouping together of similar methods within a suitably named service, and promotes the possibility of testing this custom logic outside of Controllers and Views.
Depending on where the custom service will be utilised, we will dictate the best practice approach to accessing the 'Published Content Cache'. If it is 100% guaranteed that the service will only be called from a place with an UmbracoContext, eg a controller or view, then it is safe to inject IPublishedContentQuery
etc for simplicity. However if the custom service is called in a location without UmbracoContext (eg an notification handler) it will fail. Therefore the approach of accessing the Published Content Cache via injecting IUmbracoContextFactory and calling EnsureUmbracoContext()
will provide consistency across any custom services no matter where they are utilised.
In this example, we create a custom service, that's responsible for finding key pages within a site, eg the News Section or the Contact Us page. These methods will commonly be called in different places throughout the site, and it's great to encapsulate the logic to retrieve them in a single place - we'll call this service SiteService
.
Create an interface to define the service:
Create the concrete service class that implements the interface:
Register the custom service with Umbraco's underlying DI container using an IComposer
:
"Transient" services can be injected into "Transient" and below ⤵. (i.e. "Transient" services can be injected anywhere)
"Transient" means that anytime this type is needed a brand new instance of this type will be created.
"Scope" services can be injected into "Request"/"Scope" based lifetimes only
"Scope" means that a single instance of this type will be created for the duration of the current HttpRequest. The instance will be disposed of at the end of the current HttpRequest.
"Singleton" services can be injected into "Singletons" and below ⤵.
"Singleton" means that only a single instance of this type will ever be created for the lifetime of the application.
1 - The service will ONLY be used during a request like in a Controller or View
You can avoid repeating common implementation logic in multiple controllers and views. This is done by consolidating these implementations into a custom service. If you are very familiar with IPublishedContentQuery injecting this into the custom service is straight forward, but the caveat is you can only use this service in a controller/view.
For example, locating the 'special' pages in the site using the familiar syntax of the IPublishedContentQuery
:
2 - The service can be used within or outside of a web request
The second approach can seem 'different' or more complex at first glance, but it is the syntax and method names that are slightly different... it enables the registering of the service in Singleton Scope, and its use outside of controllers and views.
Occasionally, you may face a situation where Umbraco fails to boot, due to a circular dependency on IUmbracoContextFactory
. This can happen if your service interacts with third party code that also depends on an IUmbracoContextFactory
instance (e.g. an Umbraco package).
See the Circular Dependencies article for an example on how to get around this.
Aside: What is the IUmbracoContextAccessor then?
The IUmbracoContextFactory
will obtain an UmbracoContext
by first checking to see if one exists on the current thread using the IUmbracoContextAccessor
. This is a singleton that can be injected anywhere and whose function is to provide access to the current UmbracoContext. On a 'non request' thread the IUmbracoContextAccessor's TryGetUmbracoContext method will return false and the IUmbracoContextFactory will create a new instance of the UmbracoContext.
If you need to know whether the UmbracoContext has been obtained from an existing thread, or whether it has been freshly created, you can 'inject' IUmbracoContextAccessor
yourself. This will check if the UmbracoContext is null using the TryGetUmbracoContext method, indicating whether you are in a 'non request' thread or not. You will still need to inject and use an IUmbracoContextFactory if you subsequently want to obtain an UmbracoContext in a non-request thread.
NB: With the IUmbracoContextAccessor
and IUmbracoContextFactory
you should NEVER have to inject the UmbracoContext itself directly into any of your constructors.
Because we've registered the SiteService with Umbraco's underlying DI framework we can inject the service into our controller's constructor, in the same way as 'core' Services and Helpers.
You can generate this in Visual Studio by using either ctrl + . or alt + enter when your cursor is on the base class:
If strictly following the paradigm of MVC, calling custom Services from Views might feel like an anti-pattern. However there isn't necessarily one single 'best practice' approach to working with Umbraco. A lot depends on circumstance, expertise and pragmatism. Allowing Umbraco to handle the flow of incoming requests to a particular page + template, and writing implementation logic in Views/Templates, is still a very common approach. There are circumstances, where the custom implementation logic shared is very 'View' specific. Custom logic for constructing 'Alternative Text' for images or different crop urls for img srcsets can be neatly handled in a custom Helper/Service without having to create a hijacked MVC route for the request and build a complex ViewModel. Custom Services called from Views, can help separate the concerns, even if the 'plumbing' isn't pure MVC.
To access the service directly from the view you would need to use the Razor @inject
keyword to get a reference to the concrete implementation of the service registered with DI:
Sometimes you might want to request, for example "/sitemap.xml" from your server, but since this has a file extension it will be treated as a client-side request and will not work. You can configure routes to be handled as server-side requests in your startup.cs.
For a single route:
For multiple routes: