view UNIT_TEST_LSH.cpp @ 548:e18843dc0aea

Implement a rudimentary API for audioDB::liszt The API is rudimentary because we've dropped support for the incremental retrieval of tracks and their number of vectors (at the API level; the SOAP and command-line support is still there -- no changes should be visible). This is potentially bad for the large-scale databases, of course; one million tracks will take of the order of 16MB of RAM, more if I'm unlucky about how std::string.c_str() is implemented. Both this liszt operation and querying (and sampling, forthcoming...) would benefit from a `cursor-like' interface to retrieval results: for an API like that, instead of getting a struct with the data there, you get a cookie with which you can ask the database for successive results. This would be neat for all sorts of reasons. In the meantime, at least this change fixes SOAP memory leaks related to liszt. Make liszt.o part of LIBOBJS rather than ordinary OBJS, so that the liszt functionality is actually compiled into the library. Add a test for this library functionality; also modify the command-line test file to run the SOAP server on its own port.
author mas01cr
date Wed, 11 Feb 2009 12:38:03 +0000
parents 071a108580a4
children
line wrap: on
line source
// UNIT_TEST_LSH.cpp

#include <vector>
#include "lshlib.h"
#include "reporter.h"

#define LSH_IN_CORE


#define N_POINT_BITS 14
#define POINT_BIT_MASK 0x00003FFF

// Callback method for LSH point retrieval
void add_point(void* reporter, Uns32T pointID, Uns32T qpos, float dist)
{
  ReporterBase* pr = (ReporterBase*)reporter;
  pr->add_point(pointID>>N_POINT_BITS, qpos, pointID&POINT_BIT_MASK, dist);
}

int main(int argc, char* argv[]){

  int nT = 100; // num tracks 
  int nP = 1000;  // num points-per-track
  float w = 4.0;// LSH bucket width
  int k = 10;
  int m = 2;
  int d = 10;
  int N = 100000;
  int C = 200;

  float radius = 0.001;
  char FILENAME[] = "foo.lsh";

  assert(nP>=nT);

  int fid = open(FILENAME,O_RDONLY);
  LSH* lsh;
  bool serialized = false;
  Uns32T trackBase = 0;

  if(fid< 0){ // Make a new serial LSH file
    lsh = new LSH(w,k,m,d,N,C,radius);
    assert(lsh);
    cout << "NEW LSH:" << endl;
    }
  else{
    close(fid); // Load LSH structures from disk
    lsh = new LSH(FILENAME); 
    assert(lsh);
    cout << "MERGE WITH EXISTING LSH:" << FILENAME << endl;
    serialized=true;
    trackBase = (lsh->get_maxp()>>N_POINT_BITS)+1; // Our encoding of tracks and points
  }  
  cout << "k:" << lsh->get_numFuns() << " ";
  cout << "L:" << lsh->get_numTables() << " ";
  cout << "d:" << lsh->get_dataDim() << " ";
  cout << "N:" << lsh->get_numRows() << " ";
  cout << "C:" << lsh->get_numCols() << " ";
  cout << "R:" << lsh->get_radius() << " ";
  cout << "p:" << lsh->get_maxp() << endl;
  cout.flush();

  cout << endl << "Constructing " << nT << " tracks with " << nP << " vectors of dimension " << d << endl;
  cout.flush();
  // Construct sets of database vectors, use one point from each set for testing
  vector< vector<float> > vv = vector< vector<float> >(nP); // track vectors
  vector< vector<float> > qq = vector< vector<float> >(nP);// query vectors
  for(int i=0; i< nP ; i++){
    vv[i]=vector<float>(d);  // allocate vector
    qq[i]=vector<float>(d);  // allocate vector
  }
  
  for(int k = 0 ; k < nT ; k ++){
    cout << "[" << k << "]";
    cout.flush();
    for(int i = 0 ; i< nP ; i++)
      for(int j=0; j< d ; j++)
	vv[i][j] =   genrand_real2() / radius; // MT_19937 random numbers
    lsh->insert_point_set(vv, (trackBase+k)<<N_POINT_BITS);
    qq[k] = vv[k]; // One identity query per set of database vectors
  }
  cout << endl;
  cout.flush();

  cout << "Writing serialized LSH tables..." << endl;
  // TEST SERIALIZED LSH RETRIEVAL
  lsh->serialize(FILENAME);

  // TEST LSH RETRIEVAL IN CORE
  printf("\n********** In-core LSH retrieval from %d track%c **********\n", 
	 (lsh->get_maxp()>>N_POINT_BITS)+1,(lsh->get_maxp()>>N_POINT_BITS)>0?'s':' ');
  fflush(stdout);  
  for(int i = 0; i < nT ; i++ ){
    trackSequenceQueryRadNNReporter* pr = new trackSequenceQueryRadNNReporter(nP,nT,(lsh->get_maxp()>>N_POINT_BITS)+1); 
    lsh->retrieve_point(qq[i], i, &add_point, (void*)pr); // LSH point retrieval from core
    printf("query vector %d] t1:%u t2:%0X\n", i, lsh->get_t1(), lsh->get_t2());
    fflush(stdout);
    pr->report(0,0);
    delete pr;
  }
  delete lsh;

  cout << "Loading Serialized LSH functions from disk ..." << endl;
  cout.flush();
  lsh = new LSH(FILENAME);
  assert(lsh);  
  //  lsh->serial_dump_tables(FILENAME);
  printf("\n********** Serialized LSH retrieval from %d track%c **********\n", (lsh->get_maxp()>>N_POINT_BITS)+1,(lsh->get_maxp()>>N_POINT_BITS)>1?'s':' ');
  fflush(stdout);  
  for(int i= 0; i < nT ; i++ ){
    trackSequenceQueryRadNNReporter* pr = new trackSequenceQueryRadNNReporter(nP,nT,(lsh->get_maxp()>>N_POINT_BITS)+1); 
    lsh->serial_retrieve_point(FILENAME, qq[i], i, &add_point, (void*) pr); // LSH serialized point retrieval method  
    printf("query vector %d] t1:%u t2:%0X\n", i, lsh->get_t1(), lsh->get_t2());
    fflush(stdout);
    pr->report(0,0);
    delete pr;
  }  
  delete lsh;

#ifdef LSH_IN_CORE  
  cout << "Loading Serialized LSH functions and tables from disk ..." << endl;
  cout.flush();
  // Unserialize entire lsh tree to core
  lsh = new LSH(FILENAME,1);
  
  // TEST UNSERIALIZED LSH RETRIEVAL IN CORE
  printf("\n********** Unserialized LSH in-core retrieval from %d track%c **********\n", (lsh->get_maxp()>>N_POINT_BITS)+1,(lsh->get_maxp()>>N_POINT_BITS)>1?'s':' ');
  fflush(stdout);
  for(int i = 0; i < nT ; i++ ){
    trackSequenceQueryRadNNReporter* pr = new trackSequenceQueryRadNNReporter(nP,nT,(lsh->get_maxp()>>N_POINT_BITS)+1); 
    lsh->retrieve_point(qq[i], i, &add_point, (void*) pr); // LSH point retrieval from core
    printf("query vector %d] t1:%u t2:%0X\n", i, lsh->get_t1(), lsh->get_t2());
    fflush(stdout);
    pr->report(0,0);
    delete pr;
  }  
  delete lsh;
#endif

}