NEWS

---

zucbor 0.1.0 (2026-10-21)

• Initial CRAN release.
• cbor_decode(), cbor_decode_seq(), cbor_read() and cbor_read_seq() turn CBOR (RFC 8949) and CBOR sequences (RFC 8742) into ordinary R values. The whole input is checked first -- well-formedness, UTF-8, tag content, duplicate map keys (by value), and depth, size and item limits -- so a length header can never make R allocate for data the input does not hold. Every fault is a classed condition inheriting zucbor_error.
• cbor_encode() and cbor_encode_seq() write RFC 8949 core deterministic encoding: identical R objects give identical bytes on every platform, and decoding then re-encoding deterministic input reproduces it exactly.
• cbor_annotate() prints an annotated hex dump: each head with its offset, bytes and meaning, for reviewing a signed message byte by byte.
• cbor_decode_prefix() decodes the item a raw vector starts with and reports how many bytes it used, for CBOR inside binary framing such as WebAuthn authData. What follows the item is not read.
• RFC 8746 typed arrays decode to integer and double vectors, and multi-dimensional arrays (tags 40 and 1040) to matrices and arrays. cbor_encode(typed_arrays = TRUE) writes numeric vectors as typed arrays and matrices as tag 1040: smaller, and much faster to read and write. Off by default, so existing output is unchanged.
• Data frames encode as an array of one map per row, keyed by column name, and data_frame = TRUE in the decoders reads such an array (SenML, for instance) back as a data frame. max_cells bounds the cells of a decoded frame before it is built.
• cbor_read_seq(each =) reads a sequence of any length in memory bounded by its largest item: each item is checked whole, decoded and passed to a function as soon as its last byte arrives. The limits then apply per item.
• tag_handlers = in the decoders gives meaning to any tag: a function per tag number turns its decoded content into an R value. Handlers run only after the whole input has been checked, and an error in one is zucbor_handler_error. as_cbor() is the encoding half: an S3 generic that cbor_encode() calls for any class it does not know.
• cbor_validate() runs the check alone; cbor_diagnose() shows CBOR in RFC 8949 diagnostic notation, as the RFC's own examples write it.
• cbor_map(), cbor_tag(), cbor_simple() and cbor_bigint() represent the values R has no native type for: maps with non-text keys, tagged items, simple values and integers beyond 2^53.
• Bundles the parser and validator of TinyCBOR 7.0; no system library is needed.