A Spring Cloud Stream binder for a Servlet container using Spring MVC.
Usage
Add this jar to the classpath of a Spring Cloud Stream application as the only binder implementation. Endpoints are exposed at
| Method | Path | Description |
|---|---|---|
| GET | /stream/{channel}/{route} |
If the channel is an @Output returns a list of the payloads of all messages sent to the channel with a routing key equal to {route}. |
| GET | /stream/{channel}/{route}/{body} |
If the channel is an @Input sends the last segment of the path (the {body}) as a payload to the {route}. If the channel is not an input, it's an error (404). |
| POST | /stream/{channel}/{route} |
Accepts a single value or a list of payloads and sends them to the @Input called {channel} with routing key header equal to {route}. |
The result of the POST depends on whether an @Output is linked to the @Input. By default a link is made if the user has @EnableBinding with an interface having precisely one @Output and one @Input (e.g. using Processor from Spring Cloud Stream). In the case that there is no linked @Output, the return value from the POST is a 202 (Accepted) and a mirror of the input. If an @Output is linked, then the contents of the output channel are returned with a 200 status (OK).
The routing key is sent via a message header named stream_routekey. It is a plain string, and can be multiple URI segments (i.e. contain "/"). It will show up in HTTP response headers if present. Message headers can be sent in a POST request using HTTP headers (as well as the special case of the route key being part of the path).
Both the channel and route are optional, but can be used to disambiguate if necessary. So for example:
| Method | Path | Description |
|---|---|---|
| GET | /stream |
Defaults the channel name to output (or to the name of the @Output if there is only one). Messages sent with no routing key are delivered. |
| GET | /stream/{channel} |
Uses and explicit channel name but an empty routing key. If the channel is not a registered output, then it will be interpreted as a route (see below). |
| GET | /stream/{route}/{body} |
Equivalent to GET /stream/input/{route}/{body} if there is a unique input channel. |
| GET | /stream/{route} |
If the channel is missing (i.e. the first path segment is not a registered output or input), then it will be interpreted as a route. Equivalent to GET /stream/output/{route} if there is a unique output channel. |
| POST | /stream/{route} |
As long as the first segment of the route is not an input channel name, this defaults the channel to input (or to the name of the @Input if it is unique). |
| POST | /stream |
Defaults the channel to input (or to the name of the @Input if it is unique). The routing key is empty. |
Note that with a GET, if the channel is not a registered output, then it will be interpreted as a route. So if there is a default input channel, then the path will be transformed into {route}/{body} (agin with route optional, if there is only one path segment) and sent to the input channel.
The result of a GET is a moving time window by default (the last 10 seconds of data are buffered). Clients can request an infinite stream of data using GET with Accept: text/event-stream (or a compatible media type).
Configuration properties (in addition to the ones provided by Spring Cloud Stream for bindings and channel names, etc.):
| Key | Default | Description |
|---|---|---|
spring.cloud.stream.binder.servlet.prefix |
stream |
The prefix for the URL paths |
spring.cloud.stream.binder.servlet.buffer-timeout-seconds |
10 | The buffer size in seconds to store messages from the output channels. |
spring.cloud.stream.binder.servlet.receive-timeout-millis |
100 | The timeout for send and receive if POST has a linked output channel. Only relevant if the message processing is asynchronous. |