The `O` format causes reference counts to increase, but in an error they
are not released. Callers to unpack functions that use the `O` format
should use pointers pre-initialized to NULL so they can safely release
the reference on error.
Also corrected typo which said this was like `O` (itself).
Fixes#135
The JSON_EMBED encoding flag causes the opening and closing characters
of the top-level array ('[', ']') or object ('{', '}') to be omitted
during encoding. This feature makes it possible to concatenate multiple
arrays or objects in the stream output. It also makes it possible to
perform outputs of partial composes.
One such example of a partial compose is when outputting a JWE object.
The output is a JSON object. But it has one top-level attribute
("ciphertext") that can grow out of proportion with the rest of the
metadata. With the JSON_EMBED flag, the other metadata can be composed
ahead of time and dumped during the beginning of output, where the
"ciphertext" and "tag" attributes can be streamed out in chunks. Thus,
the header material can be composed with Jansson and the ciphertext
itself can be composed manually.
This function encodes the json_t object to a pre-allocated buffer.
It compliments the already existing json_loadb() function and is
useful for parsing JSON-RPC (among other protocols) when sent over
datagram sockets.
Signed-off-by: Nathaniel McCallum <npmccallum@redhat.com>
The object_to_key_iter() example is now formatted like
json_object_foreach() and json_object_foreach(). The edited descriptions were
redundant, and the removed headers didn't add useful information.
This is particularly useful in modular situations where the allocation
functions are either unknown or private. For instance, in such cases,
the caller of json_dumps() has no way to free the returned buffer.
