It's definitely not compliant. I just read the code, there's no support for escapes in strings, except for skipping over escaped double quotes to find the string end. This applies to any string, so it also won't correctly match keys that contain escapes. (hence the small number of allocations)

Edited to add: this has its uses, if you know that's all you need. Long ago I wrote a similar xml parser for a project because we had control over the xml being produced, and the garbage overhead from element names was killing us with Xerces; I didn't bother implementing things like DTD support, because YAGNI.

Anyone knows good JSON specs to run against?