Parsing and printing XML

Representation of XML trees

For example,

<p:a xmlns:p="<http://example.com/p>">
  Text 1
  <p:b att1="one" att2="two">
    <q:a xmlns:q="<http://example.com/q>" p:att3="three">
    Text 2
    </q:a>
  </p:b>
  Text 3
</p:a>

is represented as (ignoring whitespace):

xml.elementNode({
  name: "p:a",
  space: "<http://example.com/p>",
  prefix: "p",
  local: "a",
  props: {},
  children: [
    xml.textNode("Text 1"),
    xml.elementNode({
      name: "p:b",
      space: "<http://example.com/p>",
      prefix: "p",
      local: "b",
      props: {
        "att1": { name:"att1", space:"", prefix:"", local:"att1", value:"one" },
        "att2": { name:"att1", space:"", prefix:"", local:"att2", value:"two" }
      },
      children: [
        xml.elementNode({
          name: "q:a",
          space: "<http://example.com/p>",
          prefix: "q",
          local: "a",
          props: {
            "p:att3": { name:"p:att3", space:"<http://example.com/p>", prefix:"p", local:"att3", value:"three" }
          }.
          children: [ xml.textNode("Text 2") ]
        })
      ]
    }),
    xml.textNode("Text 3")
  ]
})

Notes:

• whitespace inside elements is, of course, significant, and here just omitted for better readability
• we do neither represent processing instructions, directives, nor comments
• the xmlns attributes are not represented as such in the tree. They are removed by the parser, and the namespace information is only represented by the space fields. The formatter adds xmlns attributes back.
• only name and space are authoritative. prefix and local are just entered by the parser for convenience, and are directly taken from name. The formatter only looks at name and space.
• if space is the empty string, the element or attribute is outside any namespace
• the parser can normalize the prefixes if a namespace map is given. If this map e.g. maps http://example.com/p to the prefix abc, this prefix will appear in the tree even if the XML text uses p
• the formatter understands also simplified properties like props: { att1: "one" }. In this case, however, no namespace can be used.