boost::ptree find? or how to access deep arrays? C++ - json

Question

I have been trying for too much time now, to access a json_reader ptree from the boost library.
I have a json file that is capsulated really often: (pseudo-json:)
"Foo": {
"nameofFoo:"foofoo"
"Bar": [{
"BarFoo":
{ BarFooDeep: {
BarFooDeepDeep: {
"BarFooValue1": 123
"BarFooValue2" : 456
}
}
}
"FooBar": [ {
"FooBarDeep" :[ {
FooBarDeepDeep:[ {
FooBarValue1: "ineedthis"
FooBarValue2: "andthis"
} ]
FooBarDeepDeep1:[ {
FooBarValue1: "ineedthis"
FooBarValue2: "andthis"
} ]
"FooBarDeep" :[ {
FooBarDeepDeep2:[ {
FooBarValue1: "ineedthis"
FooBarValue2: "andthis"
} ]
FooBarDeepDeep3:[ {
FooBarValue1: "ineedthis"
FooBarValue2: "andthis"
} ]
and so on .... won t complete this now...
Now I only need to get FooBarValue1 and FooBarValue2 of all FooBar.
I know ptree puts arrays together with empty childs ("")
I can access all the members by itereration over all the childs recursively.
But is there not a better way to access special values?
how does ptree find works? i always get compiler errors ...
ptree jsonPT;
read_json( JSON_PATH, jsonPT);
ptree::const_iterator myIT = jsonPT.find("FooBarValue1");
double mlat = boost::lexical_cast<int>(myIT->second.data());
error: conversion from
‘boost::property_tree::basic_ptree,
std::basic_string >::assoc_iterator’ to non-scalar type
‘boost::property_tree::basic_ptree,
std::basic_string >::const_iterator’ requested
ptree::const_iterator myIT = jsonPT.find("FooBarValue1");
Can anyone give me a useful hint how to get access to this ptree?!?

Answer 1

As hinted in the linked answer I commented (Boost.PropertyTree subpath processing), you could write your own "selector" query, so you could write stuff like:
read_json("input.txt", pt);
std::ostream_iterator<std::string> out(std::cout, ", ");
std::cout << "\nSpecific children but in arrays: ";
enumerate_path(pt, "Foo.Bar..FooBar..FooBarDeep1..FooBarDeepDeep6..FooBarValue2", out);
std::cout << "\nSingle wildcard: ";
enumerate_path(pt, "Foo.Bar..FooBar..FooBarDeep1..*..FooBarValue2", out);
std::cout << "\nTwo wildcards: ";
enumerate_path(pt, "Foo.Bar..FooBar..*..*..FooBarValue2", out);
The enumerate_path function need not be too complicated and takes any output iterator (so you can back_inserter(some_vector) just as well):
template <typename Tree, typename Out, typename T = std::string>
Out enumerate_path(Tree const& pt, typename Tree::path_type path, Out out) {
if (path.empty())
return out;
if (path.single()) {
*out++ = pt.template get<T>(path);
} else {
auto head = path.reduce();
for (auto& child : pt) {
if (head == "*" || child.first == head) {
out = enumerate_path(child.second, path, out);
}
}
}
return out;
}
As simple working demo prints:
Specific children but in arrays: andthis6,
Single wildcard: andthis6, andthis7, andthis8, andthis9,
Two wildcards: andthis1, andthis2, andthis3, andthis4, andthis6, andthis7, andthis8, andthis9,
That is with the following input.txt:
{
"Foo": {
"nameofFoo": "foofoo",
"Bar": [{
"BarFoo": {
"BarFooDeep": {
"BarFooDeepDeep": {
"BarFooValue1": 123,
"BarFooValue2": 456
}
}
},
"FooBar": [{
"FooBarDeep0": [{
"FooBarDeepDeep1": [{
"FooBarValue1": "ineedthis1",
"FooBarValue2": "andthis1"
}],
"FooBarDeepDeep2": [{
"FooBarValue1": "ineedthis2",
"FooBarValue2": "andthis2"
}]
},
{
"FooBarDeepDeep3": [{
"FooBarValue1": "ineedthis3",
"FooBarValue2": "andthis3"
}],
"FooBarDeepDeep4": [{
"FooBarValue1": "ineedthis4",
"FooBarValue2": "andthis4"
}]
}],
"FooBarDeep1": [{
"FooBarDeepDeep6": [{
"FooBarValue1": "ineedthis6",
"FooBarValue2": "andthis6"
}],
"FooBarDeepDeep7": [{
"FooBarValue1": "ineedthis7",
"FooBarValue2": "andthis7"
}]
},
{
"FooBarDeepDeep8": [{
"FooBarValue1": "ineedthis8",
"FooBarValue2": "andthis8"
}],
"FooBarDeepDeep9": [{
"FooBarValue1": "ineedthis9",
"FooBarValue2": "andthis9"
}]
}]
}]
}]
}
}
Live On Coliru
Full Listing
Live On Coliru
#include <boost/property_tree/ptree.hpp>
#include <boost/property_tree/json_parser.hpp>
#include <iostream>
template <typename Tree, typename Out, typename T = std::string>
Out enumerate_path(Tree const& pt, typename Tree::path_type path, Out out) {
if (path.empty())
return out;
if (path.single()) {
*out++ = pt.template get<T>(path);
} else {
auto head = path.reduce();
for (auto& child : pt) {
if (head == "*" || child.first == head) {
out = enumerate_path(child.second, path, out);
}
}
}
return out;
}
int main() {
std::ostream_iterator<std::string> out(std::cout, ", ");
using namespace boost::property_tree;
ptree pt;
read_json("input.txt", pt);
std::cout << "\nSpecific children but in arrays: ";
enumerate_path(pt, "Foo.Bar..FooBar..FooBarDeep1..FooBarDeepDeep6..FooBarValue2", out);
std::cout << "\nSingle wildcard: ";
enumerate_path(pt, "Foo.Bar..FooBar..FooBarDeep1..*..FooBarValue2", out);
std::cout << "\nTwo wildcards: ";
enumerate_path(pt, "Foo.Bar..FooBar..*..*..FooBarValue2", out);
}

Answer 2

find() is for retrieving a child node by key; it doesn't search the whole ptree. It returns an assoc_iterator (or const_assoc_iterator), which you can convert to an iterator via the to_iterator() method on the parent:
ptree::const_assoc_iterator assoc = jsonPT.find("FooBarValue1");
ptree::const_iterator myIT = jsonPT.to_iterator(assoc);
To search the ptree, you'll need to iterate it recursively:
struct Searcher {
struct Path { std::string const& key; Path const* prev; };
void operator()(ptree const& node, Path const* path = nullptr) const {
auto it = node.find("FooBarValue1");
if (it == node.not_found()) {
for (auto const& child : node) { // depth-first search
Path next{child.first, path};
(*this)(child.second, &next);
}
} else { // found "FooBarValue1"
double mlat = boost::lexical_cast<int>(myIT->second.data());
// ...
std::cout << "Mlat: " << mlat << std::endl;
std::cout << "Path (reversed): ";
for (Path const* p = path; p != nullptr; p = p->prev)
std::cout << p << ".";
std::cout << std::endl;
}
}
};
Searcher{}(jsonPT);
Alternatives for writing the recursive traversal would be a C++14 generic lambda, or in C++11 a type-erased concrete lambda using std::function.