CBOR: Improve recognition of CBOR data format

[clintongormley]

Currently we only check if the first byte of the body is a BYTE_OBJECT_INDEFINITE to determine whether the content is CBOR or not. However, what we should actually do is to check whether the "major type" is an object.

See:

• https://github.com/FasterXML/jackson-dataformat-cbor/blob/master/src/main/java/com/fasterxml/jackson/dataformat/cbor/CBORParser.java#L614
• https://github.com/FasterXML/jackson-dataformat-cbor/blob/master/src/main/java/com/fasterxml/jackson/dataformat/cbor/CBORParser.java#L682

Also, CBOR can be prefixed with a self-identifying tag, 0xd9d9f7, which we should check for as well. Currently Jackson doesn't recognise this tag, but it looks like that will change in the future: https://github.com/FasterXML/jackson-dataformat-cbor/issues/6

[clintongormley]

Jackson 2.4.3 now contains the above fixes. We should upgrade and add the changes mentioned above.

[s1monw]

if we get a fix for this I think it should go into 1.3.6

[s1monw]

do we need to do anything else than upgrading jackson? @pickypg do you have a ETA for this?

[pickypg]

I should have this up for review on Monday.

We need to change XContentFactory.xContentType(...) to support the new header. By default, the new CBORGenerator.Feature.WRITE_TYPE_HEADER feature is false, so just upgrading will do nothing (nothing breaks, but nothing improves).

[pickypg]

Merged

[pickypg]

This was reverted because the JSON tokenizer was acting up in some of the randomized tests. I am looking at the root cause (my change or just incoming changes from 2.4.3).

[johnrfrank]

@clintongormley it'd be great to have this feature. What are the chances this will get into an upcoming release of Elasticsearch?

[nik9000]

@johnrfrank this was merged into 2.0.0. We've since deprecated content detection in favor of providing a content-type header.