Traverse multi-index table through non-primary value and delete the record

Question

Below is my smart contract:

#include <eosio/eosio.hpp>
#include <eosio/print.hpp>
#include <eosio/system.hpp>
#include <string>

using namespace eosio;
using std::string;

class [[eosio::contract("document")]] document : public eosio::contract {
        public:
                using contract::contract;
                document(name receiver, name code, datastream<const char*> ds):contract(receiver, code, ds) {}

        [[eosio::action]]
        void erasedoc(name key) {
                            document_index documents( get_self(), get_first_receiver().value);
                            bool found = false;
                            for(auto itrer = documents.begin(); itrer != documents.end() && found!=true; itrer++) {
                                  if(itrer->key == key) {
                                          found = true;
                                          documents.erase(itrer);
                                  }
                            }
                            check(found == false, "Document does not exist");
         }
         private:
                struct [[eosio::table]] documenttable {
                        uint64_t doc_id;
                        name key;
                        std::string doc_name;
                        std::string doc_description;

                        uint64_t primary_key() const { return doc_id;}
                };
                typedef eosio::multi_index<"documents"_n, documenttable> document_index;
};

doc_id is the primary key. But I want to delete the record using key. As, key is not the primary key I cannot use auto iterator = documents.find(key.value);. So, I implemented this logic. But this returns errors. Error:

Error 3160005: The table operation is not allowed
Error Details:
dereference of deleted object
pending console output:

Please help. How could I delete the record using key?

Answer 1

A secondary index sounds like a reasonable solution, depending on the amount and type of data stored but this should solve your problem as well.

... 
if(itrer->key == key) {
    found = true;
    itrer = documents.erase(itrer);
}
...

Iterators are invalidated when erased which is why the error is thrown, but erase returns the next iterator so setting it should work.