Variables can be used to store intermediate results during flow execution. A valid variable name starts with $ followed by a letter a…z or A…Z or an underscore _. More letters, underscores, the numbers 0…9 or hyphens - may follow.

Predefined Variables

The following predefined variables exist:

• ​$body: client request body

• $env: environment variables​

• ​$request: client request information

• $server: server information

• ​$upstream: upstream response information

• ​$error: Contains information regarding the most recent error, but is initially empty.

Try the following flow with

$ curl --data hello localhost:8080 | jq

<flow>
  <template>
  {
    "$request": {
      {{with $request }}{{: * }}{{end}}
    },
    "$body": {{ $body }},
    "$upstream": {{ $upstream }},
    "$server": {
      {{: $server/* }}
    },
    "$env": {
      {{: $env/* }}
    }
  }
  </template>
</flow>

The actions request and requests set the $upstream variable, too:

<flow>
  <requests>
  {
    "ok": {
      "url": "https://httpbin.org/status/200"
    },
    "failure": {
      "url": "https://httpbin.org/status/500"
    }
  }
  </requests>
​
  <template>
  {
    "$upstream": {
      {{with $upstream }}{{: * }}{{end}}
    }
  }
  </template>
</flow>

{
  "$upstream": {
    "ok": {
      "url": "https://httpbin.org/status/200",
      "status": 200,
      "cacheHit": false,
      "headers": { … }
    },
    "failure": {
      …
    }
  }
}

Defining and Accessing Variables

Global variables are usually defined as the output of eval actions. The variable name is defined in the out="…" attribute. It must begin with $ followed by a letter and then more letters or numbers.

<flow>
  <!-- $x = 1 -->
  <eval out="$x">1</eval>
  <!-- $x = $x + 5 -->
  <eval out="$x">$x + 5</eval>
  <!-- $answer = $x * 7 -->
  <eval out="$answer">$x * 7</eval>
</flow>

For structured JSON variables, you can use the template action:

<flow>
  <template out="$cfg">
  {
    "stage": "prod",
    "mock": {{ boolean($request/get/mock) }},
    "answer": {{ $answer }}
  }
  </template>
</flow>

Variables may be copied with eval, too:

<flow>
  <eval out="$client_request">$request</eval>
  …
</flow>

Local Variables

​Templates may define local variables with {{$… := …}}. Those variables are undefined outside the template they're defined in:

<flow>
  <template>{{$answer:= 42 }}</template>
  <!-- null -->
  <template>{{ $answer }}</template>
</flow>

📎 If a variable containing binary content is processed in a template or xslt action, its content will probably end up being truncated, garbled or both.

Undefined Variables

Attempting to access a variable that has not been set previously will yield an empty node-set. The empty node-set will be evaluated to false in conditions and produces the string null in placeholders:

<flow>
  <template>
    {{if $undefined }}
      will never be reached
    {{else}}
      <!-- null -->
      {{ $undefined }}
    {{end}}
  </template>
</flow>

$body

The $body variable contains the request body.

If the request body is JSON (Content-Type: application/json) $body contains the parsed JSON. You can access its properties with XPath expressions with a json segment before the top-level properties. E.g.

{
  "foo": 1,
  "bar": {
    "baz": true
  }
}

The value for foo can be accessed by $body/json/foo, the value for baz by $body/json/bar/baz.

In other cases the content is stored in $body as a string and cannot be accessed by XPath.

$request

The $request variable contains information about the incoming client request, such as the URL, the request header fields and possibly the query component or cookies, if any were sent.

Example:

<request>
  <method>POST</method>
  <purpose>main</purpose>
  <debug/>
  <host>localhost</host>
  <port>12345</port>
  <id>XeeSVJ5AFt8VyXYagp3lvgAAACc</id>
  <url>http://localhost:12345/api/proxy/foo?a=b&amp;c=d</url>
  <path>/api/proxy/foo</path>
  <query>a=b&amp;c=d</query>
  <headers>
    <host>localhost:12345</host>
    <user-agent>curl/7.64.0</user-agent>
    <accept>*/*</accept>
    <cookie>NAME1=VALUE1; NAME2=VALUE2</cookie>
    <content-type>application/x-www-form-urlencoded</content-type>
    <foo>asdf</foo>
    <bar>qwer</bar>
    <content-length>17</content-length>
  </headers>
  <get>
    <a>b</a>
    <c>d</c>
  </get>
  <cookies>
    <NAME1>VALUE1</NAME1>
    <NAME2>VALUE2</NAME2>
  </cookies>
  <endpoint>/api/proxy</endpoint>
</request>

As HTTP request headers are defined to be case-insensitive, their names are lower-cased for convenient access even if the client has sent the field name with upper-case letters, e.g.:

$request/headers/user-agent

If a client URL path is matched by a wildcard path, $request/endpoint is the path part preceding the part matched by /**. Otherwise, $request/endpoint is the same as $request/path.

…
basePath: /api
…
paths:
  /**:
    …
  /foo/**:
    …
  /foo/qux:
    …
  /foo/{p1}:
    …

$upstream

The $upstream variable contains information about upstream responses. The properties for each upstream response are stored with the request ID (id property or content attribute).

• url - The request URL (string)

• status - The response status code (integer)

• cacheHit - true if the response was served from a cache (see use-http-cache or force-cache-ttl request options), false otherwise

• headers - The response headers, each with a lower-cased field name

Example:

<upstream>
  <request1>
    <url>https://httpbin.org/status/200</url>
    <status number="">200</status>
    <cacheHit>false</cacheHit>
    <headers>
      <date>Thu, 27 Aug 2020 14:12:33 GMT</date>
      <content-type>text/html; charset=utf-8</content-type>
      <connection>keep-alive</connection>
      <server>gunicorn/19.9.0</server>
      <access-control-allow-origin>*</access-control-allow-origin>
      <access-control-allow-credentials>true</access-control-allow-credentials>
    </headers>
  </request1>
  <request2>
    <url>https://httpbin.org/status/500</url>
    <status number="">500</status>
    <cacheHit>false</cacheHit>
    <headers>
      <date>Thu, 27 Aug 2020 14:12:33 GMT</date>
      <content-type>text/html; charset=utf-8</content-type>
      <connection>keep-alive</connection>
      <server>gunicorn/19.9.0</server>
      <access-control-allow-origin>*</access-control-allow-origin>
      <access-control-allow-credentials>true</access-control-allow-credentials>
    </headers>
  </request2>
</upstream>

To check, for example, if the status code of the response with the ID myRequest is successful you can use the following XPath expression:

$upstream/myRequest/status = 200

$error

Client request and response validation, upstream connection and request and response validation errors, and those triggered by the error action will store information about the error in $error. While initially empty, $error will have the following properties containing information about the most recent error, unless it is triggered by the error action:

• status - the HTTP status that is used by default for responses if the error was passed to the client (type: number)

• code - an error code (type: number)

• message - a single line of text describing the error (type: string)

• info - detailed information about the error (type: array of string)

• requestID - the requestID as it should appear in the logs (type: string)

Example:

{
  "status": 400,
  "code": 3204,
  "message": "Input Validation Failed",
  "info": [
    "Path /api/empty-body/ not found."
  ],
  "requestID": "[email protected]"
}

If set by the error action, $error contains the JSON template result of the action's element body.

See also

• ​Using Env Vars (cookbook)