jsmn (pronounced like ‘jasmine’) is a minimalistic JSON parser in C. It can be easily integrated into the resource-limited projects or embedded systems.

You can find more information about JSON format at json.org.

Library sources are available at https://github.com/zserge/jsmn.


Usually JSON parsers convert JSON string to internal object representation. But if you are using C it becomes tricky as there is no hash tables, no reflection etc. That’s why most JSON parsers written in C try to reinvent the wheel, and either invent custom JSON-like objects, custom hash maps, or use callbacks like SAX parsers do.

jsmn is missing all that functionality, but instead is designed to be robust (it should work fine even with erroneous data), fast (it parses data on the fly and is re-entrant), portable (no superfluous dependencies or non-standard C extensions). And of course, simplicity is a key feature.



jsmn splits JSON string into tokens. Let’s consider a JSON string:

'{ "name" : "Jack", "age" : 27 }'

jsmn will split it into the following tokens:

The key moment is that jsmn tokens do not hold any data, but just point to the token boundaries in JSON string instead. In the example above jsmn creates tokens like:

Every jsmn token has a type which is one of the following:

Besides start/end positions, jsmn tokens for complex types (like arrays or objects) also contain a number of child items, so you can easily follow object hierarchy.

This approach provides enough information for parsing any JSON data and makes it possible to use zero-copy techniques.


To clone the repository you should have mercurial installed. Just run:

$ hg clone http://bitbucket.org/zserge/jsmn jsmn

Repository layout is simple: jsmn.c and jsmn.h are library files; demo.c is an example of how to use jsmn (it is also used in unit tests); test.sh is a test script. You will also find README, LICENSE and Makefile files inside.

To build the library, run make. It is also recommended to run make test. Let me know, if some tests fail.

If build was successful, you should get a libjsmn.a library. The header file you should include is called "jsmn.h".


Token types are described by jsmntype_t:

typedef enum {
    JSMN_OBJECT = 1,
    JSMN_ARRAY = 2,
} jsmntype_t;

Note: Unlike JSON data types, primitive tokens are not divided into numbers, booleans and null, because one can easily tell the type using the first character:

Token is an object of jsmntok_t type:

typedef struct {
    jsmntype_t type; /* Token type */
    int start;       /* Token start position */
    int end;         /* Token end position */
    int size;        /* Number of child (nested) tokens */
} jsmntok_t;

Note: string tokens point to the first character after the opening quote and the previous symbol before final quote. This was made to simplify string extraction from JSON data.

All job is done by jsmn_parser object. You can initialize a new parser using:

jsmn_parser parser;


This will initialize (or reset) the parser.

Later, you can use jsmn_parse() function to process JSON string with the parser:

jsmntok_t tokens[256];
const char *js;
int r;

js = ...;
r = jsmn_parse(&parser, js, strlen(js), tokens, 256);

A non-negative return value of jsmn_parse is the number of tokens actually used by the parser.

Passing NULL instead of the tokens array would not store parsing results, but instead the function will return the value of tokens needed to parse the given string. This can be useful if you don’t know yet how many tokens to allocate.

If something goes wrong, you will get a negative error returned by jsmn_parse(). Return value will be one of these:

If you get JSON_ERROR_NOMEM, you should re-allocate more tokens and call jsmn_parse once again. If you read JSON data from the stream, you can periodically call jsmn_parse and check if return value is not JSON_ERROR_PART - this will indicate the end of the JSON data.

jsmn stores the offsets inside parser structure, not pointers. It means you can use realloc() to get more tokens, or reallocated your js string when more data arrives.

Non-strict mode

By default jsmn is working in a non-strict mode. It allows you to use it for other useful formats that JSON. In non-strict mode jsmn accepts:

It means the following text will be valid for jsmn:

server: example.com
post: 80
message: "hello world"

Looks like a config file, right? And this is how you can use jsmn to parse JavaScript code:

    server: "example.com",
    post: 80,
    message: "hello world"

To switch to strict mode you should define JSMN_STRICT preprocessor variable.

In the benchmark jsmn have shown pretty good results for small objects (~4KB). But for large objects it was terribly slow. This can be fixed by storing links to the parent nodes. You will loose about 4 bytes per token, but speed will be much higher (by the way, after this little hack jsmn seems to be the fastest parser in that benchmark).

To enable parent links you should define JSMN_PARENT_LINKS before compiling jsmn.

There are several posts about jsmn around the web:

Other info

This software is distributed under MIT license so feel free to integrate it in your commercial products.