As the name implies, the servlet interceptor is both a servlet and an interceptor. It is created when an application calls api:create-server[] or api:create-servlet[].
When created, the servlet interceptor creates the global interceptor queue, which includes a number of interceptors private to the servlet interceptor, and additional interceptors provided by the application; importantly, the final interceptor will perform routing, which will add route-matched interceptors to the queue.
The servlet interceptor is a chain provider that invokes interceptors when an HTTP request arrives.
The servlet interceptor performs the following tasks:
Sets up the context-map.adoc and request-map.adoc
Executes the interceptor queue (:enter, then :leave)
Catches any exceptions that aren’t handled by error-handling.adoc within the interceptors
Writes the final response, using the response-map.adoc
Part of this work is by adding new keys to the default context map; beyond that, additional interceptors are prepended:
The stylobate interceptor handles otherwise uncaught unhandled exceptions that occur during interceptor chain execution.
It also works with the ring-response
interceptor when dealing
with async response bodies.
Responsible for converting the :response key in the context, when present, into a response delivered to the client; this may include async streaming.
There are a couple of special cases to note.
Interceptors may rewrite the interceptor queue by calling functions api:enqueue[ns=io.pedestal.interceptor.chain] and api:terminate[ns=io.pedestal.interceptor.chain].
Interceptors may also change the interceptor queue by directly changing the values of the :io.pedestal.interceptor.chain/stack and :io.pedestal.interceptor.chain/queue in the context-map.adoc. This should be avoided when possible, as manipulating the stack and queue improperly may have unintended consequences. |
The queue contains the interceptors to be executed next. The first item in the queue will be the next interceptor considered. When an interceptor’s :enter function is executed, it is removed from the queue and added to the stack.
The stack contains the previously executed interceptors that need their :leave functions called. The top of the stack will be considered first.
Before invoking the :enter functions, the servlet interceptor sets up a terminator predicate on the context. This predicate is passed the context map and returns a truthy value if execution should transition from :enter to :leave.
The default predicate checks to see if the :response context key is present and, if so, is a valid Ring response map; that is, does it have a :status key of type integer, and a :headers key containing a map?
The end result is that in some cases, interceptors (including handlers) will not get executed for a particular request. For example, an application that has a concept of authentication may include an interceptor that checks for valid credentials in the request, and adds a :response map to the context when the credentials are missing or invalid. That will cause the response to immediately be sent to the client, bypassing any later interceptors.
In addition, it is possible to have more than one termination predicate; if any predicate returns a truthy value, then the :enter execution will terminate. New predicates can be appended using api:terminate-when[ns=io.pedestal.interceptor.chain].
In very rare cases, different interceptors may provide different parts of response. Ultimately, the response must be a valid Ring response, but different interceptors may provide different parts of the response (the :body or certain values in the :headers map).
If your application requires this, be very careful about early termination.
Put the initial value into the response map in the last :enter function, then use the :leave functions to refine that value.
Put the initial value into the context under a different key and refine that value in either :enter or :leave functions. When the response is complete, transfer the response map to the :response key.
Can you improve this documentation?Edit on GitHub
cljdoc is a website building & hosting documentation for Clojure/Script libraries
× close