CBOR: read object from byte array and get the offset to the end of the read object in array

[martinpaljak]

I have a byte array containing CBOR object and additional data after it. I'd like to read that object with Jackson tools and get the offset in the array where this object ends and other data starts. Is that exposed somehow or possible without going into too internals?

[martinpaljak]

Answering my own question:

CBORFactory cf = new CBORFactory();
cf.setCodec(new CBORMapper());
CBORParser parser = cf.createParser(cbor);
System.out.println("start: "  + parser.getCurrentLocation().getByteOffset());
JsonNode object = parser.readValueAsTree();
System.out.println("stop: "  + parser.getCurrentLocation().getByteOffset());

[cowtowncoder]

Right, there TWO locations: one is the START of the currently available token: the other is current location by parser, typically AFTER the token but not necessarily, in case of things like lazy decoding.

The reliable way would be to get the START location of the token AFTER one you are interested in.

The current location would work in case you ensure that you access whatever data token has (getText() or similar).

[martinpaljak]

I'm using @Deprecated public void writeStartArray(int elementsToWrite) to generate an array of objects that needs to use the concise format with pre-defined length (just as I do for an object/map of fixed size with CBORGenerator.writeStartObject(numElements)). Should I be using some other method?

[cowtowncoder]

Yes, there are the non-deprecated variants like writeStartObject/Array(Object forValue, int size). If called from context other than databind (there is no Java Value for which Object/Array is being written to represent) just pass null.
Deprecated methods will remain for all of 2.x, removed from 3.0.

[martinpaljak]

Thanks for the hint. I was confused about what should I pass there, as I create the array contents manually. Will update to new variant with null!

[martinpaljak]

Hijacking the old thread for a simple question: I don't see any flags for CBORMapper / CBORGenerator.Feature that would enable the "compact" encoding of CBOR maps/arrays (together with length, rather than undefined length and break markers). This is required for FIDO/WebAuthn and thought it might be useful to ask before I write my own small simple "CBOR.compact()" function.

I get around it by manually constructing messages (like described in previous comments) that require this by by the spec, but would be nice to be able to cast any map into a compliant compact representation with object mapper.

https://fidoalliance.org/specs/fido-v2.1-ps-20210615/fido-client-to-authenticator-protocol-v2.1-ps-20210615.html#message-encoding

[cowtowncoder]

True, there are no such settings, partly since most if not all of the work would be from base ObjectMapper and not CBORMapper itself.

Cases are bit different however: for Array-encoded values, I would have thought most serializers would already call length-prefixed variant; for cases where number of elements is known. It would be worth filing a separate issue to note cases where this is not true; it is possible jackson-databind changes were needed.

Objects are a bit more difficult, unfortunately, due to filtering at databind level (for POJOs and Maps), which in many cases is done dynamically as part of serialization itself (that is, decision made on each element when iterating and outputting element; not as a separate step before writing elements).
It'd be ok to file an issue for Object case separately as well, of course, in case someone could come up with a good solution -- perhaps a SerializationFeature for jackson-databind that would either use length-prefix when possible (no filtering applied), or speculatively for all cases (but possibly failing if filtering resulted in size discrepancy).