How to erase child from boost tree if it is double registered? - c++

Question

Could you please help me with the following?
I populate this finalTree = treeA + treeB
However, the problem is that some elements of treeB have the same name with some of treeA. As a result I might have double registries for some children.
ie.
<category>
<fruit type="banana">
<characteristic>
<point v="0"/>
</characteristic>
</fruit>
<fruit type="orange">
<characteristic>
<point v="1"/>
</characteristic>
</fruit>
<fruit type="banana">
<characteristic>
<point v="2"/>
</characteristic>
</fruit>
<fruit type="fig">
<characteristic>
<point v="3"/>
</characteristic>
</fruit>
</category>
What I want to achieve is to delete the first entry of banana and keep the last.
So far I do:
boost::property_tree::ptree & node = informationTree.add("information.fruitTypes", "");
node.add("<xmlattr>.type", fruit);
node.add_child("characteristic", char);
The problem is that I don't know how to remove it, as I don't know whether the double entry will be banana or something else the next time. Should I copy populate tree? What do you suggest please?

Answer 1

If you're just building the tree, you can just use put_* instead of add_* and it would overwrite an element if it already exists by that name.
If you have a tree and want to remove the duplicates at a certain subtree, you have to do it manually, e.g.:
Live On Coliru
#include <boost/property_tree/xml_parser.hpp>
#include <iostream>
#include <map>
using boost::property_tree::ptree;
template <typename KeyF>
ptree nodup(ptree const& pt, KeyF key_accessor) {
ptree filtered;
std::map<std::string, std::reference_wrapper<ptree> > seen;
for (auto& entry : pt) {
auto key = key_accessor(entry);
auto previous = seen.find(key);
if (seen.end() == previous)
seen.emplace(key, filtered.add_child(entry.first, entry.second));
else
previous->second.get() = entry.second; // overwrite
}
return filtered;
}
int main() {
ptree pt;
{
std::istringstream iss( "<category><fruit type=\"banana\"><characteristic><point v=\"0\"/></characteristic></fruit><fruit type=\"orange\"><characteristic><point v=\"1\"/></characteristic></fruit><fruit type=\"banana\"><characteristic><point v=\"2\"/></characteristic></fruit><fruit type=\"fig\"><characteristic><point v=\"3\"/></characteristic></fruit></category>");
read_xml(iss, pt);
}
write_xml(std::cout, pt, boost::property_tree::xml_writer_make_settings<std::string>(' ', 4, "utf-8"));
auto& subtree = pt.get_child("category");
subtree = nodup(subtree, [](ptree::value_type const& item) { return item.second.get("<xmlattr>.type", ""); });
write_xml(std::cout, pt, boost::property_tree::xml_writer_make_settings<std::string>(' ', 4, "utf-8"));
}
If you are a bit more performance concerned, you can iterate backwards and avoid some overwrite actions:
Live On Coliru
template <typename KeyF>
ptree nodup(ptree const& pt, KeyF key_accessor) {
ptree filtered;
std::map<std::string, std::reference_wrapper<ptree> > seen;
for (auto entry = pt.rbegin(), last = pt.rend(); entry != last; ++entry) {
auto key = key_accessor(*entry);
auto previous = seen.find(key);
if (seen.end() == previous)
seen.emplace(key, filtered.add_child(entry->first, entry->second));
}
return filtered;
}
However, keep in mind this potentially alters the order in which the fruits appear in the subtree.