The JSON document to convert is the ui.json file that we already encountered in the Parsing JSON messages with JsonSlurper recipe.

The conversion is based on groovy.xml.MarkupBuilder, which allows full control on the final output but makes the code more verbose because we have to explicitly specify all the transformation rules.

The code in step 2 is really a testament to the sheer efficiency of builders (see the Defining data structures as code in Groovy recipe in Chapter 3, Using Groovy Language Features). The variable xml is used to construct the XML element names and attributes that are extracted, via dynamic methods, from the ui variable, created by the MarkupBuilder. More information on MarkupBuilder can be found in the Constructing XML content recipe in Chapter 5, Working with XML in Groovy.

To avoid confusing the Groovy compiler, when we are inside the iterator closures (for example, widget.axes.each { widgetAxis -> ...) we have to use variable names that are different from the XML element names (for example, axis(type: widgetAxis.type ...).

The second approach mentioned in this recipe's introduction is to use a third-party library, such as JSON-lib to perform the transformation with little coding. The following code makes use of the XMLSerializer class to transform the JSON data to XML.

@Grapes([
  @Grab(group='net.sf.json-lib',
    module='json-lib',
    version='2.3',
    classifier='jdk15'),
  @Grab('xom:xom:1.2.5')
])
import net.sf.json.JSONObject
import net.sf.json.xml.XMLSerializer

def object = JSONObject.fromObject(new File('ui.json').text)

println new XMLSerializer().write(object)

Here we trade lines of code for lack of control above the produced XML, that is not quite as readable as the Groovy only example:

<?xml version="1.0" encoding="UTF-8"?>
<o>
  <items class="array">
    <e class="object">
      <animate type="boolean">true</animate>
      <axes class="array">
        <e class="object">
          <fields class="array">
            <e type="string">x</e>
          </fields>
          <position type="string">left</position>
          <title type="string">Time</title>
          <type type="string">Time</type>
        </e>
        ...
      </axes>
      ...
    </e>
    ...
  </items>
</o>

The output can still be tweaked by setting different options of the XmlSerializer class, but you would not be able to fully control the complete structure of the output. The technique outlined here may be established for very complex JSON documents, which would require a non-trivial amount of code to perform the conversion to XML. An additional approach could be to apply XSLT transformation to the result produced by the XmlSerializer, but describing the process goes beyond the scope of this recipe.