[edit] [comment] [remove] |2006-03-17| e1 # XML and JSON play nice together

More and more web service providers seem to be interested in offering JSON APIs beneath their XML APIs. One considerable advantage of using a JSON API is its ability to provide cross-domain requests while bypassing the restrictive same-domain-policy of the XmlHttpRequest object.

On the client-side JSON comes with a native language compliant data structure. With that it performs much better than corresponding DOM calls required for XML processing. Finally transforming JSON structures to presentational data can be easily achieved with tools like JSONT.

So there might be a wish to convert an existing XML document to a JSON structure, simultaneously requiring

  • preserving structure
  • preserving order
  • no loss of information

In an ideal world the resulting JSON structure can be converted back to its original XML document easily.

It seems to be worthwhile, to discuss some common patterns as the foundation of a potentially bidirectional conversion process between XML and JSON.

A similar discussion can be found over at BadgerFish and Yahoo!, without the reversibility aspect though.


[edit] [comment] [remove] |2006-04-17| e3 # A pragmatic approach

A single structured XML element might come in seven flavors:

  1. empty element.
  2. element with pure text content.
  3. empty element with attributes.
  4. element with pure text content and attributes.
  5. element containing elements with different names.
  6. element containing elements with identical names.
  7. element containing elements and contiguous text.

The following table shows the corresponding conversion patterns between XML and JSON.

"e": null
"e": "text"
<e name="value" />
"e":{"@name": "value"}
<e name="value">text</e>
"e": {
  "@name": "value",
  "#text": "text"
"e": { 
  "a": "text",
  "b": "text"
"e": {
  "a": ["text", "text"]
"e": {
  "#text": "text",
  "a": "text"

Please note, that all patterns are considered to describe structured elements, despite the fact, that the element of pattern 7 is commonly understood as a semistructured element.

A pragmatic approach to convert a XML document to a JSON structure and vice versa can be based on the seven patterns above. It always assumes a normalized XML document for input and doesn't take into consideration:

  • XML declaration
  • processing instructions
  • explicit handling of namespace declarations
  • XML comments

[edit] [comment] [remove] |2006-04-18| e4 # Preserving order

JSON is built on two internal structures:

  • A collection of name/value pairs with unique names (associative array).
  • An ordered list of values (array).

An attempt to map a structured XML element


to the following JSON object

"e": {
  "a": "some",
  "b": "textual",
  "a": "content"

will yield an invalid result, since the name "a" is not unique in the associative array. So we need to collect all elements of identical names in an array each. Using the patterns 5 and 6 above it results in

"e": {
  "a": [ "some", "content" ],
  "b": "textual"

But now we have a structure, that doesn't preserve element order. This might be tolerable or not, depending on whether above XML element order matters or not.

So as a rule of thumb we can phrase

A structured XML element can be converted to a reversible JSON structure, if

  • all subelement names occur exactly once, or …
  • subelements with identical names are in sequence.


A structured XML element can be converted to an irreversible but semantically equivalent JSON structure, if

  • multiple homonymous subelements occur not in sequence and …
  • element order doesn't matter.

If none of these two rules match, there is no pragmatic way to convert XML to JSON using the patterns above. Here come SVG and SMIL documents to mind, which implicitly rely on element order.


[edit] [comment] [remove] |2006-03-19| e2 # Semi-Structured XML

XML documents can contain semi–structured elements, which are elements with mixed content of text and child elements, usually seen in documentation markup. If the textual content is contiguous, as in

  some textual

we can apply pattern 7 and yield for this special case

"e": {
  "#text": "some textual",
  "a":  "content",

But how do we convert textual content mixed up with elements, as in


It obviously doesn't make sense in most cases, to collect all text nodes in an array,

"e": {
  "#text": ["some", "content"],
  "a": "textual"

which does not preserve order nor semantics.

So it seems to be the best solution in accordance with a pragmatic approach to treat mixed semi–structured content in JSON the same way as XML treats CDATA sections – as unknown markup.

"e": "some <a>textual</a> content"

As another rule we state, that XML elements with

  • mixed content of text and element nodes and
  • CDATA sections

are converted to a reversible JSON string containing the complete XML markup according to pattern 2 or 4.


[edit] [comment] [remove] |2006-05-05| e6 # Examples

Now let's have a look at two examples with the insight we gained up to here. Microformats are well suited, as they are an open standard and short enough for a brief discussion.

XOXO as a simple XHTML based outline format is one of several microformats. The slightly modified sample from the Draft Specification reads:

<ol class="xoxo">
  <li>Subject 1
        <li>subpoint a</li>
        <li>subpoint b</li>
  <li><span>Subject 2</span>
    <ol compact="compact">
        <li>subpoint c</li>
        <li>subpoint d</li>

Now we apply the patterns above to convert this XML document fragment to a JSON structure.

  1. The outer list with two list items is converted using pattern 6.
  2. The first list item contains a single textual content "Subject 1" and an inner list element. So it can be treated according to pattern 7.
  3. The first inner list is converted with pattern 6 again.
  4. Pattern 5 is applied to the second item of the outer list.
  5. The second inner list is converted using a combination of pattern 3 and 6.

Here is the resulting JSON structure, which is reversible without loss of information.

"ol": {
  "li": [ 
      "#text": "Subject 1",
      "ol": {
        "li": ["subpoint a", "subpoint b"]
      "span": "Subject 2",
      "ol": {
        "@compact": "compact",
        "li": ["subpoint c", "subpoint d"]

hCalender is another microformat based on the iCalender standard. We just ignore the fact, that the iCalender format could be more easily converted to JSON and look at a hCalender event example, which is also slightly modified so, that it is a structured rather than a mixed semi-structured document fragment.

<span class="vevent">
  <a class="url" href="http://www.web2con.com/">
    <span class="summary">Web 2.0 Conference</span>
    <abbr class="dtstart" title="2005-10-05">October 5</abbr>
    <abbr class="dtend" title="2005-10-08">7</abbr>
    <span class="location">Argent Hotel, San Francisco, CA</span>

Here patterns 2, 3, 4, 5 and 6 are used to generate the following JSON structure:

"span": {
  "a": {
    "@class": "url",
    "@href": "http://www.web2con.com/",
    "span": [
      { "@class": "summery", "#text": "Web 2.0 Conference" },
      { "@class": "location", "#text": "Argent Hotel, San Francisco, CA" }
    "abbr": [
      { "@class": "dtstart", "title": "2005-10-05", "#text": "October 5" },
      { "@class": "dtend", "title": "2005-10-08", "#text": "7" }

This example demonstrates a conversion which does not preserve the original element order. Even if this may not change semantics here, we generally can

  1. state, that a conversion isn't sufficiently possible.
  2. tolerate the result, if order doesn't matter.
  3. try to make our XML document more JSON-friendly.

In many cases the last point may be not acceptable, at least when the XML document is based on existing standards. But in other cases it may be worth the effort to consider some subtle XML changes, which can make XML and JSON play nicely together. Changing the <abbr> elements to <span> elements in the hCalender example would be such an improvement.


[edit] [comment] [remove] |2006-04-24| e5 # Conclusion

XML is a document centric format, while JSON is a format for structured data. This fundamental difference may be irrelevant, as XML is also capable of describing structured data. If XML is used to describe highly structured documents, these may play very well together with JSON.

Problems may arise, if XML documents

  • implicitly rely on element order.
  • contain lot of semi-structured data.

As a proof of concept I have implemented two Javascript functions

  • xml2json
  • json2xml

based on the six patterns above, which can be used for

  • client side conversion of
    • a parsed XML document via DOM to a JSON structure.
    • a JSON structure to a (textual) XML document.
  • implementing converters in other server side languages.

Future XML document design might be influenced by these ore similar patterns, in order to get the best of the two – XML and JSON – worlds.