Interface CommentOptions

Should numbers and strings be created as boxed instances, which retain their original encoding for round-tripping?

If this is true, saveOriginal is also set to true. Think of this as "saveOriginal + extras". The thought is that most use cases for saveOriginal will want the original encoding of an object or array, and won't care about the original encoding of strings and numbers. Turning this on also has the side-effect of making all CBOR maps decode as JS Map objects, rather than plain Objects, since the string keys will need to be instances of String.

Turn on all options for draft-ietf-cbor-cde-05.

Should bigints that can fit into normal integers be collapsed into normal integers?

In dCBOR, JS numbers between 2^53 and 2^64 get encoded as CBOR integers. When decoding, present them as JS numbers instead of BigInt, losing accuracy.

Create an object from an array of key-value pairs. The default implementation creates a plain JS object if all of the keys are strings, otherwise creates a Map (unless preferMap or boxed is set, in which case a Map is always created).

Turn on all options for draft-mcnally-deterministic-cbor-11.

When producing diagnostic output, should the size of an element always be appended to that element using an underscore when calling diagnose?

If the input is a string, how should it be decoded into a byte stream? Ignored if the input is a Uint8Array.

Do not consider the global tag registry when decoding tags.

Keep NaN payloads by creating an instance of NAN when needed. Ignored if rejectLongLoundNaN is true.

Maximum allowed depth to parse into CBOR structures. This limit is security-relevant for untrusted inputs. May be set to Infinity for trusted inputs, but be careful!

Always generate bigint numbers from CBOR integers (major type 0 or 1).

This would be used in profiles that want to crisply distinguish between float and int types. On the encode side, you might want avoidInts=true and collapseBigInts=true to pair with this. If true, convertUnsafeIntsToFloat is ignored.

Always generate Map instances when decoding, instead of trying to generate object instances when all of the keys are strings.

A slight performance improvement if you don't need plain objects. If you have the boxed option on, this option has no effect, and Maps are always produced.

Pretty-print diagnostic format.

If there are bigint (tag 2/3) in the incoming data, throw an exception.

If there are duplicate keys in a map, should we throw an exception?

Note: this is more compute-intensive than expected at the moment, but that will be fixed eventually.

Reject any floating point numbers.

This might be used in profiles that are not expecting floats to prevent one from being coerced to an integer-looking number without the receiver knowing.

Reject any mt 0/1 numbers.

This might be used in profiles that expect all numbers to be encoded as floating point.

Reject negative integers in the range [CBOR_NEGATIVE_INT_MAX ... STANDARD_NEGATIVE_INT_MAX - 1].

Reject floating point numbers that should have been encoded in shorter form, including having been encoded as an integer.

Reject NaNs that are not encoded as 0x7e00.

This includes non-trivial NaNs without a quiet bit, with a sign bit, or with a payload. It also includes NaNs that are encoded with a float larger than f16.

If negative zero (-0.0) is received, throw an error.

Reject simple values other than true, false, undefined, and null.

Reject any attempt to decode streaming CBOR.

Enforces the Definite-Length-Only (DLO) constraint.

Reject strings that are not normalized with the given normalization form. Don't use this without Unicode expertise.

Reject subnormal floating point numbers.

Reject the undefined simple value.

Usually used with rejectSimple.

For dCBOR, reject "integers" between 2^53 and 2^64 that were encoded as floats.

Reject integers and additional-information lengths that could have been encoded in a smaller encoding.

Usually rejectLongFloats is also desired.

Save the original bytes associated with every object as a property of that object for exact round-tripping. Use getEncoded(obj) to retrieve the associated bytes. If you need the original encoded form of primitive items such as numbers and strings, set boxed: true as well.

If non-null, keys being decoded MUST be in this order. Note that this is a superset of rejectDuplicateKeys, and is slightly more efficient.

If non-null, prefer any tags in the map to ones have have been registered with Tag.registerDecoder.