Main Page » Developer Section » RESTXQ

RESTXQ

This page presents one of the Web Application services. It describes how to use the RESTXQ API of BaseX.

RESTXQ, introduced by Adam Retter, is an API that facilitates the use of XQuery as a server-side processing language for the Web. Inspired by the Java JAX-RS API, it provides a pre-defined set of XQuery 3.0 annotations for mapping HTTP requests to XQuery functions, which in turn generate and return HTTP responses.

BaseX provides various extensions to the specification:

  • Multipart types are supported, including multipart/form-data
  • A %rest:error annotation can be used to catch XQuery errors
  • Servlet errors can be redirected to other RESTXQ pages
  • Helper RESTXQ Functions are available
  • Parameters are implicitly cast to the type of the function argument
  • The Path Annotation can contain regular expressions
  • %input annotations, support for input-specific content-type parameters
  • %rest:single annotation to cancel running RESTXQ functions
  • Quality factors in the Accept header will be evaluated
  • Support for server-side quality factors in the %rest:produces annotation
  • Better support for the OPTIONS and HEAD methods

Introduction

Preliminaries

The RESTXQ service is accessible via http://localhost:8080/.

All RESTXQ Annotations are assigned to the http://exquery.org/ns/restxq namespace, which is statically bound to the rest prefix. A Resource Function is an XQuery function that has been marked up with RESTXQ annotations. When an HTTP request comes in, a resource function will be invoked that matches the constraints indicated by its annotations.

If a RESTXQ URL is requested, the RESTXQPATH module directory and its subdirectories will be traversed, and all XQuery files will be parsed for functions with RESTXQ annotations. Subdirectories that include an .ignore file will be skipped.

To speed up processing, the functions of the existing XQuery modules are automatically cached in main memory:

  • Functions will be invalidated and parsed again if the timestamp of their module changes.
  • File monitoring can be adjusted via the PARSERESTXQ option. In productive environments with a high load, it may be recommendable to change the timeout, or completely disable monitoring.
  • If files are replaced while the web server is running, the RESTXQ module cache should be explicitly invalidated by calling the static root path /.init or by calling the rest:init function.

Examples

A first RESTXQ function is shown below:

module namespace page = 'http://basex.org/examples/web-page';

declare %rest:path("hello/{$who}") %rest:GET function page:hello($who) {
  <response>
    <title>Hello { $who }!</title>
  </response>
};

If the URI http://localhost:8080/hello/World is accessed, the result will be:

<response>
  <title>Hello World!</title>
</response>

The next function demonstrates a POST request:

declare
  %rest:path("/form")
  %rest:POST
  %rest:form-param("message","{$message}", "(no message)")
  %rest:header-param("User-Agent", "{$agent}")
function page:hello-postman(
  $message  as xs:string,
  $agent    as xs:string*
) as element(response) {
  <response type='form'>
    <message>{ $message }</message>
    <user-agent>{ $agent }</user-agent>
  </response>
};

If you post something (e.g. using curl or the embedded form at http://localhost:8080/)...

curl -i -X POST --data "message='CONTENT'" http://localhost:8080/form
...you will receive something similar to the following result: 
HTTP/1.1 200 OK
Content-Type: application/xml; charset=UTF-8
Content-Length: 107
Server: Jetty(11.0.20)
<response type="form">
  <message>'CONTENT'</message>
  <user-agent>curl/7.31.0</user-agent>
</response>

Request

This section shows how annotations are used to handle and process HTTP requests.

Constraints

Constraints restrict the HTTP requests that a resource function may process.

Paths

A resource function must have a single Path Annotation with a single string as argument. The function will be called if a URL matches the path segments and templates of the argument. Path templates contain variables in curly brackets, and map the corresponding segments of the request path to the arguments of the resource function. The first slash in the path is optional.

The following example contains a path annotation with three segments and two templates. One of the function arguments is further specified with a data type, which means that the value for $variable will be cast to an xs:integer before being bound:

declare %rest:path("/a/path/{$with}/some/{$variable}")
  function page:test($with, $variable as xs:integer) { ... };

Variables can be enhanced by regular expressions:

(: Matches all paths with "app" as first, a number as second, and "order" as third segment :)
declare %rest:path("app/{$code=[0-9]+}/order")
  function page:order($code) { ... };

(: Matches all other all paths starting with "app/" :)
declare %rest:path("app/{$path=.+}")
  function page:others($path) { ... };

If multiple path candidates are found for the request, the one with more segments will be preferred.

Content Negotiation

Functions can be restricted to specific Media Types. The default type is */*. Multiple types can either be specified by a single or by multiple annotations.

Consuming Data

A function will only be taken into consideration if the HTTP Content-Type header of the request matches one of the given types:

declare
  %rest:POST("{$body}")
  %rest:path("/xml")
  %rest:consumes("application/xml")
  %rest:consumes("text/xml")
function page:xml($body) { $body };

Producing Data

A function will only be chosen if the HTTP Accept header of the request matches one of the given types:

declare
  %rest:path("/xml")
  %rest:produces("application/xml", "text/xml")
function page:xml() { <xml/> };

Note that the annotations will not affect the type of the actual response: You will need to supply an additional %output:media-type annotation or (if a single function may produce results of different types) generate an apt Custom Response.

Quality Factors

A client can supply quality factors to influence the server-side function selection process. If a client sends the following HTTP header with quality factors…

Accept: */*;q=0.5,text/html;q=1.0
…and if two RESTXQ functions exist for the addressed path with two different annotations for producing data… 
declare function %rest:produces("text/html") ...
...
declare function %rest:produces("*/*") ...
…the first of these function will be chosen, as the quality factor for text/html documents is highest.

As we cannot ensure that the client may supply quality factors, the selection process can also be controlled server-side. The qs parameter can be attached server-side to the Media Type. If multiple functions are left in the selection process, the one with the highest quality factor will be favored:

declare function %rest:produces("application/json;qs=1") ...
...
declare function %rest:produces("*/*;qs=0.5") ...

HTTP Methods

Default Methods

The HTTP method annotations are equivalent to all HTTP request methods except TRACE and CONNECT. Zero or more methods may be used on a function; if none is specified, the function will be invoked for each method.

The following function will be called if GET or POST is used as request method:

declare %rest:GET %rest:POST %rest:path("/post")
  function page:post() { "This was a GET or POST request" };

The POST and PUT annotations may optionally take a string literal in order to map the HTTP request body to a function argument. Once again, the target variable must be embraced by curly brackets:

declare %rest:PUT("{$body}") %rest:path("/put")
  function page:put($body) { "Request body: " || $body };

Custom Methods

Custom HTTP methods can be specified with the %rest:method annotation. An optional body variable can be supplied as second argument:

declare
  %rest:path("binary-size")
  %rest:method("SIZE", "{$body}")
function page:binary-size(
  $body  as xs:base64Binary
) {
  "Request method: " || request:method(),
  "Size of body: " || bin:length($body)
};

If an OPTIONS request is received, and if no function is defined, an automatic response will be generated, which includes an Allow header with all supported methods.

If a HEAD request is received, and if no function is defined, the corresponding GET function will be processed, but the response body will be discarded.

Content Types

The body of a POST or PUT request will be converted to an XQuery item. Conversion can be

controlled by specifying a content type. It can be further influenced by specifying additional content-type parameters:
Content-Type Parameters (;name=value) Type of resulting XQuery item
text/xml, application/xml document-node()
text/* xs:string
application/json JSON Options document-node() or map(*)
text/html HTML Options document-node()
text/comma-separated-values CSV Options document-node() or map(*)
others xs:base64Binary
multipart/* sequence (see next paragraph)

For example, if application/json;lax=yes is specified as content type, the input will be transformed to JSON, and the lax QName conversion rules will be applied, as described for the JSON Functions.

Input options

Conversion options for JSON, CSV and HTML can also be specified via annotations with the input prefix. The following function interprets the input as text with the CP1252 encoding and treats the first line as header:

declare
  %rest:path("/store.csv")
  %rest:POST("{$csv}")
  %input:csv("header=true,encoding=CP1252")
function page:store-csv($csv as document-node()) {
  "Number of rows: " || count($csv/csv/record)
};

Multipart Types

The single parts of a multipart message are represented as a sequence,

and each part is converted to an XQuery item as described in the last paragraph.

A function that is capable of handling multipart types is identical to other RESTXQ functions:

declare
  %rest:path("/multipart")
  %rest:POST("{$data}")
  %rest:consumes("multipart/mixed") (: optional :)
function page:multipart($data as item()*) {
  "Number of items: " || count($data)
};

Parameters

The following annotations can be used to bind request values to function arguments. Values will implicitly be cast to the type of the argument.

Query Parameters

The value of the first parameter, if found in the query component, will be assigned to the variable specified as second parameter. If no value is specified in the HTTP request, all additional parameters will be bound to the variable (if no additional parameter is given, an empty sequence will be bound):

declare
  %rest:path("/params")
  %rest:query-param("id", "{$id}")
  %rest:query-param("add", "{$add}", 42, 43, 44)
function page:params($id as xs:string?, $add as xs:integer+) {
  <result id="{ $id }" sum="{ sum($add) }"/>
};

HTML Form Fields

Form parameters are specified the same way as query parameters:

%rest:form-param("city", "{$city}", "no-city-specified")

The values are the result of HTML forms submitted with the (default) content type application/x-www-form-urlencoded:

<form action="/process" method="POST" enctype="application/x-www-form-urlencoded">
  <input type="text" name="city"/>
  <input type="submit"/>
</form>

File Uploads

Files can be uploaded to the server by using the content type multipart/form-data (the HTML5 multiple attribute enables the upload of multiple files):

<form action="/upload" method="POST" enctype="multipart/form-data">
  <input type="file" name="files" multiple="multiple"/>
  <input type="submit"/>
</form>

The file contents are placed in a map, with the filename serving as key. The following example shows how uploaded files can be stored in a temporary directory:

declare
  %rest:POST
  %rest:path("/upload")
  %rest:form-param("files", "{$files}")
function page:upload($files) {
  for $name    in map:keys($files)
  let $content := $files($name)
  let $path    := file:temp-dir() || $name
  return (
    file:write-binary($path, $content),
    <file name="{ $name }" size="{ file:size($path) }"/>
  )
};

HTTP Headers

Header parameters are specified the same way as query parameters:

%rest:header-param("User-Agent", "{$user-agent}")
%rest:header-param("Referer", "{$referer}", "none")

Cookies

Cookie parameters are specified the same way as query parameters:

%rest:cookie-param("username", "{$user}")
%rest:cookie-param("authentication", "{$auth}", "no_auth")

Query Execution

In many web search scenarios, user input from browser forms is processed and search results are returned. Such operations can be made more interactive by sending a new search request to the server with each key click. However, this may lead to many parallel server-side requests, from which only the result of the last request will be relevant for the client.

With the %rest:single annotation, it can be enforced that only one instance of a function will run at the same time and for the same client. If the same function will be called for the second time, a currently executed query will be stopped, and the HTTP error code 460 will be returned instead:

(: If fast enough, returns the result. Otherwise, if called again, raises 460 :)
declare
  %rest:path("/search")
  %rest:query-param("term", "{$term}")
  %rest:single
function page:search($term as xs:string) {
  <ul>{
    for $result in db:get('large-db')//*[text() = $term]
    return <li>{ $result }</li>
  }</ul>
};

By adding a string value to with the annotation, functions can be bundled together, and a running query can be canceled by calling another one that has the same annotation value. This is shown by another example, in which the first function can be interrupted by the second one. If you call both functions in separate browser tabs, you will note that the first tab will return 460, and the second one will return <xml>stopped</xml>.

declare
  %rest:path("/compute")
  %rest:single("EXPENSIVE")
function local:compute() {
  (1 to 100000000000000)[. = 0]
};

declare
  %rest:path("/stop")
  %rest:single("EXPENSIVE")
function local:stop() {
  <xml>stopped</xml>
};

The following things should be noted:

  • If a query will be canceled, there will be no undesirable side effects. For example, it won’t be possible to abort a query if it is currently updating the database or performing any other I/O operations. As a result, the termination of a running query can take some more time as expected.
  • The currently executed function is bound to the current session. This way, a client will not be able to cancel requests from other clients. As a result, functions can only be stopped if there was at least one previous successful response, in which initial session data was returned to the client.

Response

By default, a successful request is answered with the HTTP status code 200 (OK) and is followed by the given content. An erroneous request leads to an error code and an optional error message (e.g. 404 for “resource not found”).

A Server-Timing HTTP header is attached to each response. It indicates how much time was spent for parsing, compiling, evaluating and serializing the query. The last value will not necessarily reflect the full time for serializing the result, as the header is generated before the result is sent to the client. Server-side serialization can be enforced by annotating a function with the %rest:single annotation.

Custom Response

Custom responses can be generated in XQuery by returning an rest:response element, an http:response child node that matches the syntax of the EXPath HTTP Client Module specification, and optional child nodes that will be serialized as usual. A function that yields a response on an unknown resource may look as follows:

declare %output:method("text") %rest:path("") function page:error404() {
  <rest:response>
    <http:response status="404">
      <http:header name="Content-Language" value="en"/>
      <http:header name="Content-Type" value="text/plain; charset=utf-8"/>
    </http:response>
  </rest:response>,
  "The requested resource is not available."
};

For the time being, it is not possible to create multipart responses.

Forwards and Redirects

Redirects

The server can invite the client (e.g., the web browser) to make a second request to another URL by sending a 302 response:

<rest:response>
  <http:response status="302">
    <http:header name="Location" value="new-location"/>
  </http:response>
</rest:response>

The convenience function web:redirect can be called to create such a response.

In the XQuery context, redirects are especially helpful if Updates are performed. An updating request may send a redirect to a second function that generates a success message, or evaluates an updated database:

declare %updating %rest:path('/app/init') function local:create() {
  db:create('app', <root/>, 'root.xml'),
  db:output(web:redirect('/app/ok'))
};

declare %rest:path('/app/ok') function local:ok() {
  'Stored documents: ' || count(db:get('app'))
};

Forwards

A server-side redirect is called forwarding. It reduces traffic among client and server, and the forwarding will not change the URL seen from the client’s perspective:

<rest:forward>new-location</rest:forward>

The response can also be created with the convenience function web:forward.

Updated:

  • The POST and PUT bodies are now available in forwarded endpoints.
  • A log entry with the status code 204 is output before the forwarding takes place.

Output

The content-type of a response can be influenced by the user via Serialization Parameters. The steps are described in the REST chapter. In RESTXQ, serialization parameters can be specified in the query prolog, via annotations, or within the REST response element:

Query Prolog

In main modules, serialization parameters may be specified in the query prolog. These parameters will then apply to all functions in a module. In the following example, the content type of the response is overwritten with the media-type parameter:

declare option output:media-type 'text/plain';

declare %rest:path("version1") function page:version1() {
  'Keep it simple, stupid'
};

Annotations

Global serialization parameters can be overwritten via %output annotations.

An HTML5 response with a <!DOCTYPE html> declaration can be generated as follows:

declare
  %rest:path("index")
  %output:method("html")
  %output:html-version('5')
function page:index() {
  <html>Welcome</html>
};

The following example serializes XML nodes as JSON, using the JsonML format:

declare
  %rest:path("cities")
  %output:method("json")
  %output:json("format=jsonml")
function page:cities() {
  element cities {
    db:get('factbook')//city/name
  }
};

The next function generates XHTML header, and text/html will be set as content type:

declare
  %rest:path("done")
  %output:method("xhtml")
  %output:omit-xml-declaration("no")
  %output:doctype-public("-//W3C//DTD XHTML 1.0 Transitional//EN")
  %output:doctype-system("http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd")
function page:html() {
  <html xmlns="http://www.w3.org/1999/xhtml">
    <body>done</body>
  </html>
};

Response Element

Serialization parameters can also be specified in a REST reponse element in a query. Serialization parameters will be overwritten:

declare %rest:path("version3") function page:version3() {
  <rest:response>
    <output:serialization-parameters>
      <output:media-type value='text/plain'/>
    </output:serialization-parameters>
  </rest:response>,
  'Not that simple anymore'
};

Error Handling

If an error is raised when RESTXQ code is parsed, compiled or evaluated, an HTTP response with the status code 500 is generated.

By default, all server-side errors will be passed on to the client. This is particularly helpful during the development process. In a productive environment, however, it is advisable not to expose errors to the client. This can be realized via the RESTXQERRORS option. If disabled,

  • XQuery modules that cannot be parsed will be ignored and
  • full error messages and stack traces will be suppressed and not included in the HTTP response.

The full error information can still be looked up in the database logs.

Raise Errors

With web:error, you can abort query evaluation, enforce a premature HTTP response and report errors back to the client:

declare
  %rest:path("/teapot")
function page:teapot() {
  web:error(418, "I'm a pretty teapot")
};

In contrast to the standard fn:error function, a status code can be supplied, and the response body will only contain the specified error message and no stack trace.

Catch XQuery Errors

XQuery runtime errors can be processed via error annotations. Error annotations have one or more arguments, which represent the error codes to be caught. The codes equal the names of the try/catch construct:

Precedence Syntax Example
1 prefix:name, Q{uri}name err:FORG0001, Q{http://www.w3.org/2005/xqt-errors}FORG0001
2 prefix:*, Q{uri}* err:*, Q{http://www.w3.org/2005/xqt-errors}*
3 *:name *:FORG0001
4 * *

All error codes that are specified for a function must have the same precedence. The following rules apply when catching errors:

  • Codes with a higher precedence (smaller number) will be given preference.
  • A global RESTXQ error will be raised if two functions with conflicting codes are found.

Similar to try/catch, the pre-defined variables (code, description, value, module, line-number, column-number, additional) can be bound to variables via error parameter annotations, which are specified the same way as query parameters.

Errors may occur unexpectedly. However, they can also be triggered by a query, as demonstrated by the following example:

declare
  %rest:path("/check/{$user}")
function page:check($user) {
  if ($user = ('jack', 'lisa'))
  then 'User exists'
  else error(xs:QName('err:user'), $user)
};

declare
  %rest:error("err:user")
  %rest:error-param("description", "{$user}")
function page:user-error($user) {
  'User "' || $user || '" is unknown'
};

Catch HTTP Errors

Errors that occur outside RESTXQ can be caught by adding error-page elements with an error code and a target location to the web.xml configuration file (find more details in the Jetty Documentation):

<error-page>
  <error-code>404</error-code>
  <location>/error404</location>
</error-page>

The target location may be another RESTXQ function. The request:attribute function can be used to request details on the caught error:

declare %rest:path("/error404") function page:error404() {
  "URL: " || request:attribute("javax.servlet.error.request_uri") || ", " ||
  "Error message: " || request:attribute("javax.servlet.error.message")
};

User Authentication

If you want to provide restricted access to parts of a web applications, you will need to check permissions before returning a response to the client. The Permissions layer is a nice abstraction for defining permission checks.

Functions

The Request Functions provide access to metadata of the current HTTP request. With the Session Functions, you can set and retrieve server-side session data of the current user, while the Sessions Functions allow you to organize session data all users known to the HTTP server. With the RESTXQ Functions, RESTXQ base URIs can be requested and a WADL description of all services can be generated.

The following example returns the current host name:

import module namespace request = "http://exquery.org/ns/request";

declare %rest:path("/host-name") function page:host() {
  'Remote host name: ' || request:remote-hostname()
};

References

Documentation: Examples:
  • Sample code combining XQuery and JavaScript: Materials and paper from Amanda Galtman, Balisage 2016.
  • DBA: The Database Administration interface, bundled with the full distributions of BaseX.

Changelog

Version 11.0
  • Updated: Forwards: The body POST and PUT bodies are now available in forwarded endpoints.
  • Updated: Forwards: A log entry with the status code 204 will be output.
Version 9.6
  • Updated: Response: Server-Timing HTTP header.
Version 9.5
  • Updated: Raise Errors: Status code 400 changed to 500, omit stack trace.
Version 9.3
  • Updated: Custom Methods: Better support for the OPTIONS and HEAD methods.
  • Updated: Catch XQuery Errors: Suppress stack trace and error code in the HTTP response.
  • Removed: rest:redirect element (web:redirect can be used instead)
Version 9.2
  • Updated: Ignore XQuery modules that cannot be parsed
Version 9.0
  • Added: Support for server-side quality factors in the %rest:produces annotation
  • Updated: Status code 410 was replaced with 460
  • Removed: restxq prefix
Version 8.4
  • Added: %rest:single annotation
Version 8.1
  • Added: support for input-specific content-type parameters
  • Added: %input annotations
Version 8.0Version 7.9
  • Added: %rest:method
  • Updated: XQuery Errors, extended error annotations
Version 7.7
  • Added: Error Handling, File Uploads, Multipart Types
  • Updated: RESTXQ function may now also be specified in main modules (suffix: *.xq).
  • Updated: the RESTXQ prefix has been changed from restxq to rest.
  • Updated: parameters are implicitly cast to the type of the function argument
  • Updated: the RESTXQ root url has been changed to http://localhost:8080/
Version 7.5
  • Added: new XML elements <rest:redirect/> and <rest:forward/>
⚡Generated with XQuery
 
Login | Changelog | TOC | Imprint
RESTXQ