2.13 Serving Servlets
The first return value is a procedure that flushes the cache. If its first
optional argument is
#f (the default), all servlet caches are
flushed. Otherwise, only those servlet caches to which
url->path
maps the given URLs are flushed. The second optional argument is a procedure
which is invoked on each cached value before it is flushed, which can be used
to finalize servlet resources. Beware that the default value
void
performs no finalization. In particular, it does not shut down the servlet’s
custodian, instead allowing the servlet’s custodian-managed resources (such
as threads) to persist.
The second return value is a procedure that uses
url->path to resolve the URL to a path, then uses
path->servlet to resolve that path to a servlet, caching the
results in an internal table.
Changed in version 1.3 of package web-server-lib: Added optional argument to first return value for list of URLs.
Changed in version 1.3: Added optional argument to first return value for servlet finalizer procedure.
| (make | | url->servlet | | | | [ | #:responders-servlet-loading responders-servlet-loading | | | | | #:responders-servlet responders-servlet]) | |
|
| → dispatcher/c |
| url->servlet : url->servlet/c |
|
|
This dispatcher runs racket servlets, using url->servlet to
resolve URLs to the underlying servlets. If servlets have errors
loading, then responders-servlet-loading is used. Other errors
are handled with responders-servlet. If a servlet raises calls
next-dispatcher, then the signal is propagated by this
dispatcher.
2.13.1 Setting Up Servlets
This module is used internally to build and load servlets. It may be useful to those who are trying to extend the server.
Creates a version 1 servlet that uses directory as its current directory, a timeout manager with a timeout timeout, and start as the request handler.
Creates a version 2 servlet that uses directory as its current directory, a manager as the continuation manager, and start as the request handler.
Creates a stateless
web-server servlet that uses
directory as its current directory,
stuffer as its stuffer, and
manager as the continuation manager, and
start as the request handler.
The modules that the Web Server needs to share with all servlets.
Constructs a procedure that loads a servlet from the path in a namespace created with make-servlet-namespace,
using a timeout manager with timeouts-default-servlet as the default timeout (if no manager is given.)
2.13.2 Servlet Namespaces
This function creates a function that when called will construct a new
namespace that has all the modules from
to-be-copied-module-specs and
additional-specs, as
well as
racket and
mred, provided they are already
attached to the
(current-namespace) of the call-site.
2.13.2.1 Why this is useful
A different namespace is needed for each servlet, so that if servlet A
and servlet B both use a stateful module C, they will be isolated from
one another. We see the Web Server as an operating system for servlets,
so we inherit the isolation requirement on operating systems.
However, there are some modules which must be shared. If they were not,
then structures cannot be passed from the Web Server to the servlets,
because Racket’s structures are generative.
Since, on occasion, a user will actually wanted servlets A and B to
interact through module C. A custom make-servlet-namespace can
be created, through this procedure, that attaches module C to all
servlet namespaces. Through other means (see Dispatchers)
different sets of servlets can share different sets of modules.
2.13.3 Internal Servlet Representation
Instances of this structure hold the necessary parts of a servlet:
the
custodian responsible for the servlet’s resources,
the
namespace the servlet is executed within,
the
manager responsible for the servlet’s continuations,
the current
directory of the servlet,
and the
handler for all requests to the servlet.