left join on non-unique key values

[manuel2a]

Greetings! We're evaluating DataFrame for including in our wasm analytics component. Nice!

One of our requirements is to join DataFrames (like pandas merge or join)
It seems to me that in DataFrame, join includes only the first matched key and not subsequent matches of the same key.

• my code below
  • Expected: three records: Bob/Pinto; Mary/Tesla; Mary/Jeep
  • Actual: two records: Bob/Pinto; Mary/Tesla; Jeep is returned unmatched 🤔
• in pandas (see last example)

What am I missing?

---

#include <iostream>
#include <vector>

#include <DataFrame/DataFrame.h>

using namespace std;
using namespace hmdf;

using UIDF = StdDataFrame<unsigned int>;

int main( int ia, char** sa ){
  vector<unsigned int> personIndex = { 0, 1 };
  const vector<int> personKey   = { 0, 1 };
  const vector<string> personName = { "Bob", "Mary" };

  vector<unsigned int> carIndex = { 0, 1, 2 };
  const vector<int> carOwnerKey   = { 0, 1, 1 };
  const vector<string> carName = { "Pinto", "Tesla", "Jeep" };

  UIDF People;
  UIDF Cars;

  People.load_index( move( personIndex ));
  People.load_column( "personName", personName );
  People.load_column( "key", personKey );

  Cars.load_index( move( carIndex ));
  Cars.load_column( "carName", carName );
  Cars.load_column( "key", carOwnerKey );

  People  // returns "Jeep", but unmatched (see below)
    .join_by_column<decltype( Cars ), int, string, int>( Cars, "key", join_policy::right_join )
    .write<ostream, string, int, unsigned int>( cout, io_format::csv2 );

  return 0;
}

0,0,0,0,Bob,Pinto
1,1,1,1,Mary,Tesla
2,0,2,1,,Jeep

[hosseinmoein]

Thanks for reporting this. This bug was introduced recently.
It is fixed in master now.

[manuel2a]

Works! Thank you @hosseinmoein

0,0,0,0,Bob,Pinto
1,1,1,1,Mary,Tesla
2,1,2,1,Mary,Jeep

[hosseinmoein]

@manuel2a
If you are comparing DataFrame with Pandas or other data frames. I appreciate it, if you tabulate the results here too.
I always wanted to do that, but never had the time and resources to do it.