C++ Join two pipe divided files on key fields

问题:

i am currently trying to create a C++ function to join two pipe divided files with over 10.000.000 records on one or two key fields.

The fiels look like

    P2347|John Doe|C1234
    P7634|Peter Parker|D2344
    P522|Toni Stark|T288

and

    P2347|Bruce Wayne|C1234
    P1111|Captain America|D534
    P522|Terminator|T288

To join on field 1 and 3, the expected output should show:

    P2347|C1234|John Doe|Bruce Wayne
    P522|T288|Toni Stark|Terminator

What I currently thinking about is using a set/array/vector to read in the files and create something like:

    P2347|C1234>>John Doe
    P522|T288>>Toni Stark

and

    P2347|C1234>>Bruce Wayne
    P522|T288>>Terminator

And then use the slip the first part as the key and match that against the second set/vector/array.

What I currently have is: Read in the first file and match the second file line by line against the set. It takes the whole line and matches it:

    #include iostream>
    #include fstream>
    #include string>
    #include set>
    #include ctime>
    using namespace std;

    int main()
    {

        clock_t startTime = clock();

        ifstream inf("test.txt");
        set lines;
        string line;
        for (unsigned int i=1; std::getline(inf,line); ++i)
            lines.insert(line);

        ifstream inf2("test2.txt");

        clock_t midTime = clock();

        ofstream outputFile("output.txt");  
        while (getline(inf2, line))
        {
            if (lines.find(line) != lines.end())
                outputFile > a;
        return 0;

}

I am very happy for any suggestion. I am also happy to change the whole concept if there is any better (faster) way. Speed is critical as there might be even more than 10 million records.

EDIT: Another idea would be to take a map and have the key being the key - but this might be a little slower. Any suggestions?

Thanks a lot for any help!

回答1:

I tried multiple ways to get this task completed, none of it was efficient so far:

Read everything into a set and parse the key fields into a format: keys >> values simulating an array type set. Parsing took a long time, but memory usage stays relatively low. Not fully developed code:

        #include \
        #include \
        #include \
        #include \
        #include \
        #include \
        #include \
        std::vector &split(const std::string &s, char delim, std::vector &elems) {
        std::stringstream ss(s);
        std::string item;
        while (std::getline(ss, item, delim)) {
            elems.push_back(item);
        }
        return elems;
    }


    std::vector split(const std::string &s, char delim) {
        std::vector elems;
        split(s, delim, elems);
        return elems;
    }

    std::string getSelectedRecords(std::string record, int position){

        std::string values;
        std::vector tokens = split(record, ' ');




        //get position in vector
        for(auto& s: tokens)
            //pick last one or depending on number, not developed
            values = s;

        return values;
    }

    int main()
    {

        clock_t startTime = clock();

        std::ifstream secondaryFile("C:/Users/Batman/Desktop/test/secondary.txt");
        std::set secondarySet;
        std::string record;

        for (unsigned int i=1; std::getline(secondaryFile,record); ++i){
            std::string keys = getSelectedRecords(record, 2);
            std::string values = getSelectedRecords(record, 1);
            secondarySet.insert(keys + ">>>" + values);
        }

        clock_t midTime = clock();

        std::ifstream primaryFile("C:/Users/Batman/Desktop/test/primary.txt");
        std::ofstream outputFile("C:/Users/Batman/Desktop/test/output.txt");

        while (getline(primaryFile, record))
        {
            //rewrite find() function to go through set and find all keys (first part until >> ) and output values
            std::string keys = getSelectedRecords(record, 2);

            if (secondarySet.find(keys) != secondarySet.end())
                outputFile > a;
        return 0;
        }

Instead of pipe divided it currently uses space divided, but that should not be a problem. Reading the data is very quick, but parsing it takes an awful lot of time

The other option was taking a multimap. Similar concept with key fields pointing to values, but this one is very low and memory intensive.

    #include \
    #include \
    #include \
    #include \
    #include \
    #include \
    #include \

    int main()
    {


    std::clock_t startTime = clock();

    std::ifstream inf("C:/Users/Batman/Desktop/test/test.txt");
    typedef std::multimap Map;
    Map map;

    std::string line;

    for (unsigned int i=1; std::getline(inf,line); ++i){
        //load tokens into vector
        std::istringstream buffer(line);
        std::istream_iterator beg(buffer), end;
        std::vector tokens(beg, end);
        //get keys
        for(auto& s: tokens)
            //std::cout >>" second;
            outputFile > a;
    return 0;
    }

Further thoughts are: Splitting the pipe divided files into different files with one column each right when importing the data. With that I will not have to parse anything but can read in each column individually.

EDIT: optimized the first example with a recursive split function. Still >30 seconds for 100.000 records. Would like to see that faster plus the actual find() function is still missing.

Any thoughts? Thanks!