Handling malformed JSON in PHP

Viewed 7410

I'm trying to write a php script that handles data from a webservice that delivers "json" as a string. The problem is the string isn't really json; it's javascript. Specifically, the keys are not quoted, although the variables are. Example (the actual data is much longer and more complicated):

{desc:'User defined payload'}

As described by the php manual, json_decode() correctly fails to interpret this string.

My question is, how can I successfully interpret a string like this in php?

The only solution I can think of is to write some regular expressions that fix the syntax, but then I'd have two problems.

EDIT

Hadvig's suggestion of using the Services_JSON pear module worked, and looks like a general solution. Once I had the module installed, my code looked like this:

require_once 'PEAR.php';
require_once 'Services/JSON.php';

$Services_JSON = new Services_JSON();
$data = $Services_JSON->decode($malformed_json);

Unfortunately, this is SLOW. To interpret the whole string (~400,000 chars) took > 36 seconds! Using a regular expression to fix the quotes and then using json_decode took ~0.04 seconds. Here's what I used:

// fix single quotes
$s = str_replace("'", '"', $malformed_json);

// fix unquoted keys
$valid_json = preg_replace('/([{\[,])\s*([a-zA-Z0-9_]+?):/', '$1"$2":', $s);

$data = json_decode($valid_json);

Of course, this will break if the data contains any quotes, brackets, or commas.

5 Answers

Using regexp is a no-go. JSON grammar cannot be correctly parsed using regexp. You will open yourself to a ton of future bugs.

I recommend using some kind of a YAML parser. YAML is backwards-compatible with JSON and allows unquoted literals at the same time.

Symfony YAML component worked great for me.

And remember that there will be a performance penalty in comparison to json_decode because it's implemented natively.

Related