XML to JSON

The three styles compared

Input:

<book id="1" lang="en">
  <title>1984</title>
  <author>Orwell</author>
</book>

Compact (drops attributes):

{
  "book": {
    "title": "1984",
    "author": "Orwell"
  }
}

Badgerfish (attributes under @, text under $):

{
  "book": {
    "@id": "1",
    "@lang": "en",
    "title": { "$": "1984" },
    "author": { "$": "Orwell" }
  }
}

Literal / JsonML-ish (full fidelity):

{
  "book": {
    "@attributes": { "id": "1", "lang": "en" },
    "title": "1984",
    "author": "Orwell"
  }
}

Forced arrays

XML has no way to say “this is a list of one”. A single-element sequence looks identical to a scalar:

<items>
  <item>A</item>
</items>

That can serialise as either {"item": "A"} or {"item": ["A"]}. The converter lets you mark element names (by XPath) as always-arrays, so your downstream JSON consumer gets consistent types.

Mixed content

Mixed content, text interleaved with child elements, is the one place where any XML-to-JSON conversion loses information:

<p>Hello <em>world</em>!</p>

Most styles either drop text or collapse it. The literal style preserves order via an array:

{
  "p": {
    "@content": ["Hello ", { "em": "world" }, "!"]
  }
}

Type inference

By default values are strings. Toggle “parse numeric/boolean” and the converter will cast 42, 3.14, true, false to the matching JSON types. Be careful: an XML attribute id="007" becomes 7 with casting on, which you probably do not want.

When to pick which

• Compact: for human-readable JSON where attributes are noise.
• Badgerfish: when you need attributes and text at the same depth as children.
• Literal: when you must round-trip back to XML without loss.