The goal of blob_tree is to bring the heirarchical tree like structure of JSON to the world of binary data. Put the header wherever is convenient for your project. First declare a blob_tree.

#include "blob_tree.h"
#include <assert.h>
using namespace std;
int main(int argc, char* argv[])
{
    dicroce::blob_tree bt;

Next, prepare a blob to insert.

    vector<uint8_t> b1 = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};

Create a node called "blobs" and store a blob ("b1") under it. Note here that you never had to specify a schema. You just start using the [] operator and storing blobs in named locations.

    bt["blobs"]["b1"] = make_pair(b1.size(), &b1[0]); // data is inserted as a pair of pointer and size variables

You can also create arrays simply by using a number as the key type to the blob_tree [] operator.

    vector<uint8_t> b3 = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9};
    vector<uint8_t> b4 = {9, 8, 7, 6, 5, 4, 3, 2, 1, 0};
    bt["blobs"]["container"][0] = make_pair(b3.size(), &b3[0]);
    bt["blobs"]["container"][1] = make_pair(b4.size(), &b4[0]);

You can get the number of elements in an array simply by calling .size().

    bt["blobs"]["container"].size()

Finally when you are done youu jusut call the serialize() static method and pass in your blob_tree. serialize() returns a vector of bytes suitable for writing to a file, or perhaps sending across a network. Note: serialize() can also be passed a version number that will be stored in the blob_tree header.

    vector<uint8_t> serialized = dicroce::blob_tree::serialize(bt);

When reading a blob tree you simply call deserialize and pass in a vector of bytes.

    auto bt2 = dicroce::blob_tree::deserialize(&serialized[0], serialized.size());

You can then use the blob_tree to access fields with the .get() method. The return value of .get() is the pair<> you previously inserted.

    auto val = bt2["blobs"]["b1"].get();