I am facing a sorting issue based on a Projected value from an iterator in the MultiIndex container. Below is the complete code. What I have removed here is the call of SellOrderRecord because the function called is same GenerateData but the expected result is different i.e; with top 10 Min Records while in BuyOrderRecord should result in Top 10 Max Records.
I am forcefully iterating to 500 assuming that I will get Max/Min value within that range which is wrong. What I need to fetch is Top 10 unique Price and sum its Qty if Price is repeated.
Ex. Price 10 Qty 5; Price 9 Qty 4; Price 8 Qty 7 .... n
Same is the case when I am trying to fetch top 10 Min Price with Qty.
The result generated with using printf (commented in Generate data) shows even if I have sorted it with std::less One record has Max at the end of Buy record while another holder has Min result at the end. What I think is either both should be min to max or max to min. The complete actual code is shared.
namespace bip = boost::interprocess;
namespace bmi = boost::multi_index;
struct MIOrder_Message /// For Data
{
// STREAM_HEADER Global_Header;
// char Message_Type;
MIOrder_Message(Order_Message _Ord)
: Timestamp(_Ord.Timestamp), Order_Id(_Ord.Order_Id), Token(_Ord.Token),
Order_Type(_Ord.Order_Type), Price(_Ord.Price),
Quantity(_Ord.Quantity) {
// std::cout << " Insert data for Token "<< _Ord.Token<<std::endl;
}
long Timestamp;
double Order_Id;
int Token;
char Order_Type;
int Price;
int Quantity;
};
/// Order_Message and MIOrder_Message are almost identical
typedef bip::allocator<MIOrder_Message,
bip::managed_shared_memory::segment_manager>
shared_struct_allocator;
enum {
ORDERVIEW,
TOKENVIEW,
PRICEVIEW,
TYPEVIEW,
};
typedef bmi::multi_index_container<
MIOrder_Message,
bmi::indexed_by<
bmi::ordered_unique<bmi::tag<struct Order_Id>, BOOST_MULTI_INDEX_MEMBER( MIOrder_Message, double, MIOrder_Message::Order_Id)>,
bmi::ordered_non_unique<bmi::tag<struct Token>, BOOST_MULTI_INDEX_MEMBER( MIOrder_Message, int, MIOrder_Message::Token)>,
bmi::ordered_non_unique<bmi::tag<struct Price>, BOOST_MULTI_INDEX_MEMBER( MIOrder_Message, int, MIOrder_Message::Price), std::less<int>>,
bmi::ordered_non_unique<bmi::tag<struct Order_Type>, BOOST_MULTI_INDEX_MEMBER(MIOrder_Message, char, Order_Type)> >,
shared_struct_allocator
> Order_Set;
typedef bmi::nth_index<Order_Set, ORDERVIEW>::type Order_view;
typedef bmi::nth_index<Order_Set, TOKENVIEW>::type Token_View;
typedef bmi::nth_index<Order_Set, PRICEVIEW>::type Price_View;
typedef std::map<int, int> _PricePoint;
_PricePoint GenerateData(int _TKN, std::pair<Order_Set *, std::size_t> OrderRecord, bool IsReverse = false) {
_PricePoint CurrentPrice;
if (OrderRecord.second > 0) {
Token_View::const_iterator t1 =
OrderRecord.first->get<TOKENVIEW>().find(_TKN);
Price_View::const_iterator it2 = OrderRecord.first->project<PRICEVIEW>(t1);
int icount = 0;
int Price = 0;
while (icount < 500)
/// I am forcefully iterating to 500 assuming that i will
/// get Max value within that range. What i need to fetch is
/// Top 10 unique Price and sum its Qty if Price is repeated.
/// Ex. Price 10 Qty 5 ; Price 9 Qty 4; Price 8 Qty 7 .... n
/// Same is the case when I am trying to fetch top 10 Min Price with Qty.
{
auto val = std::next(it2, icount);
if (val->Token == _TKN) {
// printf(" Bid Data Found for token %d , Price %d , Qty %d , Time %li
// , OrderNumber %16f , icount %d \n
// ",val->Token,val->Price,val->Quantity,val->Timestamp,val->Order_Id,icount);
CurrentPrice[val->Price] += val->Quantity;
Price = val->Price;
}
icount++;
}
std::cout << " Bid Price " << Price << std::endl;
}
return CurrentPrice;
}
int main() {
bip::managed_shared_memory segment(bip::open_or_create, "mySharedMemory", 20ull<<20);
Order_Set::allocator_type alloc(segment.get_segment_manager());
Order_Set * BuyOrderRecord = segment.find_or_construct<Order_Set>("MCASTPORT0BUY")(alloc);
if (BuyOrderRecord->empty()) {
}
while (true) {
int _TKN = 49732;
_PricePoint CurrentPrice = GenerateData(_TKN, std::make_pair(BuyOrderRecord, 1), true);
std::cout << "=========================================================" << std::endl;
sleep(2);
}
return 0;
}
You never check whether the result of std::next(it2, icount) is valid.
Chances are, it isn't. After all it2 is just the element found by _TKN, projected on the ordered price index. If _TKN happens to have the highest price in the Order_Set, then it2 is already the last element at
Price_View::const_iterator it2 = OrderRecord.first->project<PRICEVIEW>(t1);
This means that event just incrementing by 1 would return end() of the price index.
It looks to me like you mistakenly thought "projecting" an iterator would project into a filtered subset, which isn't the case. It just projects onto the other index. Of the full set.
If you indeed want to be able to iterate all records matching a token for the highest price, you will prefer to use a composite key:
typedef bmi::multi_index_container<
MIOrder_Message,
bmi::indexed_by<
bmi::ordered_unique<bmi::tag<struct Order_Id>, BOOST_MULTI_INDEX_MEMBER( MIOrder_Message, double, MIOrder_Message::Order_Id)>,
bmi::ordered_non_unique<bmi::tag<struct Token>, BOOST_MULTI_INDEX_MEMBER( MIOrder_Message, int, MIOrder_Message::Token)>,
bmi::ordered_non_unique<bmi::tag<struct Price>, BOOST_MULTI_INDEX_MEMBER( MIOrder_Message, int, MIOrder_Message::Price), std::less<int>>,
bmi::ordered_non_unique<bmi::tag<struct Order_Type>, BOOST_MULTI_INDEX_MEMBER(MIOrder_Message, char, Order_Type)>
, bmi::ordered_non_unique<bmi::tag<struct Composite>,
bmi::composite_key<MIOrder_Message,
bmi::member<MIOrder_Message, int, &MIOrder_Message::Token>,
bmi::member<MIOrder_Message, int, &MIOrder_Message::Price>
>
>
>,
shared_struct_allocator
> Order_Set;
Ordered composite keys accept partial keys, so you can do:
auto range = OrderRecord.first->get<Composite>().equal_range(boost::make_tuple(_TKN));