Proxy middleware for Deno Oak HTTP servers.
import { proxy } from "https://deno.land/x/oak_http_proxy@2.3.0/mod.ts";
import { Application } from "https://deno.land/x/oak@v12.6.2/mod.ts";
const app = new Application();
app.use(proxy("https://github.com/oakserver/oak"));
await app.listen({ port: 3000 });
This is a Deno module available to import direct from this repo and via the Deno Registry.
Before importing, download and install Deno.
You can then import oak-http-proxy straight into your project:
import { proxy } from "https://deno.land/x/oak_http_proxy@2.3.0/mod.ts";
oak-http-proxy is also available on nest.land, a package registry for Deno on the Blockchain.
import { proxy } from "https://x.nest.land/oak-http-proxy@2.3.0/mod.ts";
The url argument that can be a string, URL or a function that returns a string or URL. This is used as the url to proxy requests to. Query string parameters and hashes are transferred from incoming request urls onto the proxy url.
router.get("/string", proxy("http://google.com"));
router.get("/url", proxy(new URL("http://google.com")));
router.get(
"/function",
proxy((ctx) => new URL("http://google.com"))
);
Note: Unmatched path segments of the incoming request url are not transferred to the outbound proxy URL. For dynamic proxy urls use the function form.
Proxy requests and user responses are piped/streamed/chunked by default.
If you define a response modifier (srcResDecorator
, srcResHeaderDecorator
),
or need to inspect the response before continuing (filterRes
), streaming is
disabled, and the request and response are buffered. This can cause performance
issues with large payloads.
You can also provide several options which allow you to filter, customize and decorate proxied requests and responses.
app.use(proxy("http://google.com", proxyOptions));
The filterReq
option can be used to limit what requests are proxied.
Return false to continue to execute the proxy; return true to skip the proxy for this request.
app.use(
"/proxy",
proxy("www.google.com", {
filterReq: (req, res) => {
return req.method === "GET";
},
})
);
Promise form:
app.use(
proxy("localhost:12346", {
filterReq: (req, res) => {
return new Promise((resolve) => {
resolve(req.method === "GET");
});
},
})
);
Decorate the inbound response object from the proxied request.
router.get(
"/proxy",
proxy("www.example.com", {
srcResDecorator: (req, res, proxyRes, proxyResData) => {
data = JSON.parse(new TextDecoder().decode(proxyResData));
data.newProperty = "exciting data";
return JSON.stringify(data);
},
})
);
app.use(
proxy("httpbin.org", {
srcResDecorator: (req, res, proxyRes, proxyResData) => {
return new Promise((resolve) => {
proxyResData.message = "Hello Deno!";
setTimeout(() => {
resolve(proxyResData);
}, 200);
});
},
})
);
When your proxied service returns 304 Not Modified this step will be skipped, since there should be no body to decorate.
The intent is that this be used to modify the proxy response data only.
Note: The other arguments are passed by reference, so you can currently exploit this to modify either response's headers, for instance, but this is not a reliable interface.
Defaults to true
.
When true, the url
argument will be parsed on first request, and memoized for subsequent requests.
When false
, url
argument will be parsed on each request.
For example:
function coinToss() {
return Math.random() > 0.5;
}
function getUrl() {
return coinToss() ? "http://yahoo.com" : "http://google.com";
}
app.use(
proxy(getUrl, {
memoizeUrl: false,
})
);
In this example, when memoizeUrl: false
, the coinToss occurs on each request, and each request could get either value.
Conversely, When memoizeUrl: true
, the coinToss would occur on the first request, and all additional requests would return the value resolved on the first request.
Decorate the inbound response headers from the proxied request.
router.get(
"/proxy",
proxy("www.google.com", {
srcResHeaderDecorator(headers, req, res, proxyReq, proxyRes) {
return headers;
},
})
);
Allows you to inspect the proxy response, and decide if you want to continue processing (via oak-http-proxy) or continue onto the next middleware.
router.get(
"/proxy",
proxy("www.google.com", {
filterRes(proxyRes) {
return proxyRes.status === 404;
},
})
);
By default, oak-http-proxy
will throw any errors except ECONNRESET
and ECONTIMEDOUT
via ctx.throw(err)
, so that your application can handle or react to them, or just drop through to your default error handling.
If you would like to modify this behavior, you can provide your own proxyErrorHandler
.
// Example of skipping all error handling.
app.use(
proxy("localhost:12346", {
proxyErrorHandler(err, ctx, next) {
ctx.throw(err);
},
})
);
// Example of rolling your own error handler
app.use(
proxy("localhost:12346", {
proxyErrorHandler(err, ctx, next) {
switch (err && err.code) {
case "ECONNRESET": {
ctx.response.status = 405;
return;
}
case "ECONNREFUSED": {
ctx.response.status = 200;
return;
}
default: {
ctx.throw(err);
}
}
},
})
);
Decorate the outbound proxied request url.
The returned url is used for the fetch
method internally.
router.get(
"/proxy",
proxy("www.google.com", {
proxyReqUrlDecorator(url, req) {
url.pathname = "/";
return url;
},
})
);
You can also use Promises:
router.get(
"/proxy",
proxy("localhost:3000", {
proxyReqOptDecorator(url, req) {
return new Promise((resolve, reject) => {
if (url.pathname === "/login") {
url.port = 8080;
}
resolve(url);
});
},
})
);
Generally it is advised to use the function form for the passed URL argument as this provides the full context object whereas the proxyReqOptDecorator
passes only the context.request
object.
Potential use cases for proxyReqOptDecorator
include:
Decorate the outbound proxied request initialization options.
This configuration will be used within the fetch
method internally to make the request to the provided url.
router.get(
"/proxy",
proxy("www.google.com", {
proxyReqInitDecorator(proxyReqOpts, srcReq) {
// you can update headers
proxyReqOpts.headers.set("Content-Type", "text/html");
// you can change the method
proxyReqOpts.method = "GET";
return proxyReqOpts;
},
})
);
You can also use Promises:
router.get(
"/proxy",
proxy("www.google.com", {
proxyReqOptDecorator(proxyReqOpts, srcReq) {
return new Promise((resolve, reject) => {
proxyReqOpts.headers.set("Content-Type", "text/html");
resolve(proxyReqOpts);
});
},
})
);
Normally, your proxy request will be made on the same protocol as the url
parameter. If you'd like to force the proxy request to be https, use this option.
app.use(
"/proxy",
proxy("http://www.google.com", {
secure: true,
})
);
Note: if the proxy is passed a url without a protocol then HTTP will be used by default unless overridden by this option.
You can copy the host HTTP header to the proxied Oak server using the preserveHostHeader
option.
router.get(
"/proxy",
proxy("www.google.com", {
preserveHostHeader: true,
})
);
The parseReqBody
option allows you to control whether the request body should be parsed and sent with the proxied request. If set to false
then an incoming request body will not be sent with the proxied request.
Configure whether the proxied request body should be sent as a UInt8Array buffer.
Ignored if parseReqBody
is set to false
.
router.get(
"/proxy",
proxy("www.google.com", {
reqAsBuffer: true,
})
);
The request body encoding to use. Currently only "utf-8" is supported.
Ignored if parseReqBody
is set to false
.
router.get(
"/post",
proxy("httpbin.org", {
reqBodyEncoding: "utf-8",
})
);
The request body size limit to use.
Ignored if reqBodyLimit
is set to Infinity
.
router.get(
"/post",
proxy("httpbin.org", {
reqBodyLimit: 10_485_760, // 10MB
})
);
Configure a timeout in ms for the outbound proxied request.
If not provided the request will never time out.
Timed-out requests will respond with 504 status code and a X-Timeout-Reason header.
router.get(
"/",
proxy("httpbin.org", {
timeout: 2000, // in milliseconds, two seconds
})
);
oak-http-proxy is licensed under the MIT License.
Generated using TypeDoc