This library is an thin C++ wrapper around the jsmn tokenizer: http://zserge.com/jsmn.html
It's design to be very lightweight, works without any allocation on the heap (no malloc) and supports nested objects.
It has been written with Arduino in mind, but it isn't linked to Arduino libraries so you can use this library on any other C++ project.
- Based on the well-proven jsmn tokenizer
- Supports nested objects
- Works with fixed memory allocation : no
malloc() - Low footprint
- MIT License
char json[] = "{\"Name\":\"Blanchon\",\"Skills\":[\"C\",\"C++\",\"C#\"],\"Age\":32,\"Online\":true}";
JsonParser<32> parser;
JsonHashTable hashTable = parser.parseHashTable(json);
if (!hashTable.success())
{
return;
}
char* name = hashTable.getString("Name");
JsonArray skills = hashTable.getArray("Skills");
int age = hashTable.getLong("Age");
bool online = hashTable.getBool("Online");
Download the library and extract it to:
<your Arduino Sketch folder>/libraries/ArduinoJsonParser
Just add the following line on the top of your .ino file:
#include <JsonParser.h>
To extract data from the JSON string, you need to create a JsonParser, and specify the number of token you allocate for the parser itself:
JsonParser<32> parser;
A token is an element of the JSON object: either a key, a value, an hash-table or an array. As an example the
char json[]on the top of this page contains 12 tokens (don't forget to count 1 for the whole object and 1 more for the array itself).
The more tokens you allocate, the more complex the JSON can be, but also the more memory is occupied. Each token takes 8 bytes, so
sizeof(JsonParser<32>)is 256 bytes which is quite big in an Arduino with only 2KB of RAM. Don't forget that you also have to store the JSON string in RAM and it's probably big.
32 tokens may seem small, but it's very decent for an 8-bit processor, you wouldn't get better results with other JSON libraries.
To use this library, you need to know beforehand what is the type of data contained in the JSON string, which is very likely.
The root object has to be either a hash-table (like {"key":"value"}) or an array (like [1,2]).
The nested objects can be either arrays, booleans, hash-tables, numbers or strings. If you need other type, you can get the string value and parse it yourself.
Consider we have a char json[] containing to the following JSON string:
{
"Name":"Blanchon",
"Skills":[
"C",
"C++",
"C#"],
"Age":32,
"Online":true
}
In this case the root object of the JSON string is a hash-table, so you need to extract a JsonHashTable:
JsonHashTable root = parser.parseHashTable(json);
To check if the parsing was successful, you must check:
if (!root.success())
{
// Parsing fail: could be an invalid JSON, or too many tokens
}
And then extract the member you need:
char* name = hashTable.getString("Name");
JsonArray skills = hashTable.getArray("Skills");
int age = hashTable.getLong("Age");
bool online = hashTable.getBool("Online");
Consider we have a char json[] containing to the following JSON string:
[
[ 1.2, 3.4 ],
[ 5.6, 7.8 ]
]
In this case the root object of the JSON string is an array, so you need to extract a JsonArray:
JsonArray root = parser.parseArray(json);
To check if the parsing was successful, you must check:
if (!root.success())
{
// Parsing fail: could be an invalid JSON, or too many tokens
}
And then extract the content by its index in the array:
JsonArray row0 = root.getArray(0);
double a = row0.getDouble(0);
or simply:
double a = root.getArray(0).getDouble(0);
By design, the library has no way to tell you why JsonParser::parseArray() or JsonParser::parseHashTable() failed.
There are basically two reasons why they may fail:
- the JSON string is invalid
- the JSON string contains more tokens that the parser can store
So, if you are sure the JSON string is correct and you still can't parse it, you should slightly increase the number of token of the parser.
You may go into unpredictable trouble if you allocate more memory than your processor really has. It's a very common issue in embedded development.
To diagnose this, look at every big objects in you code and sum their size to check that they fit in RAM.
For example, don't do this:
char json[1024]; // 1 KB
JsonParser<64> parser; // 512 B
because it may be too big for a processor with only 2 KB: you need free memory to store other variables and the call stack.
That is why an 8-bit processor is not able to parse long and complex JSON strings.
To reduce the memory consumption, JsonArray and JsonHashTable contains pointer to the token that are inside the JsonParser. This can only work if the JsonParser is still in memory.
For example, don't do this:
JsonArray getArray(char* json)
{
JsonParser<16> parser;
return parser.parseArray(parser);
}
because the local variable parser will be removed from memory when the function getArray() returns, and the pointer inside JsonArray will point to an invalid location.
This will probably never be an issue, but you need to be aware of this feature.
When you pass a char[] to JsonParser::parseArray() or JsonParser::parseHashTable(), the content of the string will be altered to add \0 at the end of the tokens.
This is because we want functions like JsonArray::getString() to return a null-terminating string without any memory allocation.
Here are the size of the main classes of the library.
This table is for an 8-bit Arduino, types would be bigger on a 32-bit processor.
| Type | Size in bytes |
|---|---|
| Parser<N> | 8 x N |
| JsonArray | 4 |
| JsonHashTable | 4 |
Theses tables has been created by analyzing the map file generated by AVR-GCC after adding -Wl,-Map,foo.map to the command line.
As you'll see the code size if between 1680 and 3528 bytes, depending on the features you use.
| Function | Size in bytes |
|---|---|
| strcmp(char*,char*) | 18 |
| jsmn_init(jsmn_parser*) | 20 |
| jsmn_parse(jsmn_parser*, char const*, jsmntok_t*, unsigned int) | 960 |
| JsonParser::parse(char*) | 106 |
| JsonObjectBase::getNestedTokenCount(jsmntok_t*) | 84 |
| JsonObjectBase::getStringFromToken(jsmntok_t*) | 68 |
| JsonArray::JsonArray(char*, jsmntok_t*) | 42 |
| JsonArray::getToken(int) | 112 |
| JsonArray::getString(int) | 18 |
| JsonHashTable::JsonHashTable(char*, jsmntok_t*) | 42 |
| JsonHashTable::getToken(char*) | 180 |
| JsonHashTable::getString(char*) | 18 |
| TOTAL | 1680 |
| Function | Size in bytes |
|---|---|
| JsonArray::getArray(int) | 42 |
| JsonArray::getHashTable(int) | 64 |
| JsonHashTable::getArray(char*) | 64 |
| JsonHashTable::getHashTable(char*) | 42 |
| TOTAL | 212 |
| Function | Size in bytes |
|---|---|
| JsonObjectBase::getBoolFromToken(jsmntok_t*) | 82 |
| JsonArray::getBool(int) | 18 |
| JsonHashTable::getBool(char*) | 18 |
| TOTAL | 130 |
| Function | Size in bytes |
|---|---|
| strtod(char*,int) | 704 |
| JsonObjectBase::getDoubleFromToken(jsmntok_t*) | 44 |
| JsonArray::getDouble(int) | 18 |
| JsonHashTable::getDouble(char*) | 18 |
| TOTAL | 796 |
| Function | Size in bytes |
|---|---|
| strtol(char*,char**,int) | 606 |
| JsonObjectBase::getLongFromToken(jsmntok_t*) | 56 |
| JsonArray::getLong(int) | 18 |
| JsonHashTable::getLong(char*) | 18 |
| TOTAL | 710 |