Erlang CBOR encoder/decoder

Requirements

• Erlang 17.0 or higher: cbor-erlang uses erlang maps to represent map, which introduced from 17.0.

Highlights

• Optimized: no sub-binary is created while decoding, hot code path is built with single jump table built by pattern matching.
• Tested: contains all testcases on standard documents.
• Correct floating-point implemntation: can decode floating points that are not supported by language: half-presition floating points, NaN, (+-)Infinity.

Decoder mappings

Genral

• Both byte string and UTF-8 string is mapped to binary.
• All numbers(integers, bignums, floating points) are mapped to number.
• Lists are mapped to plain lists, maps are mapped to maps.

Tags

• Bignums are decoded to numbers.
• Date/time string (tag 0) is mapped to {timetext, Binary}
• Epoch-based date/time (tag 1) is mapped to {timeepoch, Number}
• Other tags are mapped to {tag, TagId, Value}

Encoder mappings

Encoder encodes erlang terms to CBOR-encoded iolist. iolist_to_binary could be used if you need single binary output.

• Numbers (fixed-size numbers, big numbers, floating-point numbers) are supported.
• All binaries are encoded to byte strings.
• Atoms true, false, null, undefined are suported.
• All lists/maps are encoded to break-terminating format.

TODO

• cbor-erlang does not distingiush byte string and UTF-8 string while decoding. Some invalid CBOR tagged values could be decoded without problem, for example date/time string (tag 0) with byte string data items. The bug should be fixed later.
• No error reporting on encoding/decoding: no detailed error information is provided to caller, for example why encoding/decoding is failed or which byte decoder failed to decode.
• Decimal fraction (tag 3) and bigfloat (tag 4) is not supported.
• Add faster encoder which generates non-canonical CBOR data.