LIBMJ(3) Library Functions Manual LIBMJ(3)
libmj -- minimalist JSON lightweight data interchange library
mj_create(mj_t *atom, const char *text, ...);
mj_parse(mj_t *atom, const char *text, int *tokfrom, int *tokto, int
mj_append(mj_t *atom, const char *text, ...);
mj_append_field(mj_t *atom, const char *fieldname, const char
mj_deepcopy(mj_t *dest, mj_t *src);
Access to objects and array entries is made using the following
mj_object_find(mj_t *atom, const char *name, const unsigned startpoint,
const unsigned incr);
mj_get_atom(mj_t *atom, ...);
JSON object output functions:
mj_snprint(char *buffer, size_t size, mj_t *atom);
mj_asprint(char **buffer, mj_t *atom);
mj_pretty(mj_t *atom, void *stream, unsigned depth, const char *trailer);
const char *
libmj is a small library interface to allow JSON text to be created and
parsed. JSON is the Java Script Object Notation, a lightweight data-
interchange format, standardised in the ECMA standard. The library name
libmj is derived from a further acronym of ``minimalist JSON''.
The libmj library can be used to create a string in memory which contains
a textual representation of a number of objects, arbitrarily structured.
The library can also be used to reconstruct the structure. Data can thus
be serialised easily and efficiently, and data structures rebuilt to
produce the original structure of the data.
JSON contains basic units called atoms, the two basic atoms being strings
and numbers. Three other useful atomic values are provided, ``null'',
``false'', and ``true''. Atoms can be grouped together as key/value
pairs in an ``object'', and as individual, ordered atoms, in an
To create a new object, the mj_create() is used. It can be deleted using
the mj_delete() function.
Atoms, objects and arrays can be appended to arrays and objects using the
Objects can be printed out by using the mj_snprint() function. The size
of a string of JSON text can be calculated using the mj_string_size()
function. A utility function mj_asprint() is provided which will
allocate space dynamically, using calloc(3), and the JSON serialised text
is copied into it. This memory can later be de-allocated using free(3).
For formatted output to a FILE * stream, the mj_pretty() function is
used. The calling interface gives the ability to indent the output to a
given depth and for the formatted output to be followed by a trailer
string, which is usually NULL for external calls, but can be any valid
string. Output is sent to the stream file stream.
The type argument given to the mj_create(), mj_append(), and
mj_append_field() functions is taken from a list of ``false'' ``true''
``null'' ``number'' ``integer'' ``string'' ``array'' and ``object''
types. An integer differs from a number in that it cannot take on any
floating point values. It is implemented internally using a signed
64-bit integer type. This restriction of values for an integer type may
be removed at a later date.
Within a JSON object, the key values can be iterated over using an
integer index to access the individual JSON objects. The index can also
be found using the mj_object_find() function.
The way objects arrays are implemented in libmj is by using varying-sized
arrays internally. Objects have the field name as the even entry in this
internal array, with the value being the odd entry. Arrays are
implemented as a simple array. Thus, to find an object in an array using
mj_object_find(), a value of 1 should be used as the increment value.
This means that every entry in the internal array will be examined, and
the first match after the starting point will be returned. For objects,
an incremental value of 2 should be used, and an even start value should
String values should be created and appended using two parameters in the
stdarg fields, that of the string itself, and its length in bytes
immediately after the string. A value of -1 may be used if the string
length is not known.
The follow code fragment will make a JSON object out of the string
``Hello <USERNAME>\n'' in the buffer called ``buf'' where ``USERNAME'' is
the name of the user taken from the runtime environment. The encoded
text will be in an allocated buffer called ``s''
(void) memset(&atom, 0x0, sizeof(atom));
cc = snprintf(buf, sizeof(buf), "Hello %s\n", getenv("USER"));
mj_create(&atom, "string", buf, cc);
cc = mj_asprint(&s, &atom, MJ_JSON_ENCODE);
and the following example will take the (binary) text which has been
encoded into JSON and is in the buffer ``buf'', such as in the previous
example, and re-create the original text:
int from, to, tok, cc;
(void) memset(&atom, 0x0, sizeof(atom));
from = to = tok = 0;
mj_parse(&atom, buf, &from, &to, &tok);
cc = mj_asprint(&s, &atom, MJ_HUMAN);
printf("%.*s", cc, s);
The s pointer points to allocated storage with the original NUL-
terminated string in it.
ECMA-262: ECMAScript Language Specification, Ecma International, December
2009, 5th Edition,
The libmj library first appeared in NetBSD 6.0.
Alistair Crooks <agcATNetBSD.org> wrote this implementation and manual
NetBSD 6.1.5 June 22, 2011 NetBSD 6.1.5