/*!
* \brief l_dnaUnionByAset()
*
* \param[in] da1, da2
* \return dad with the union of the set of numbers, or NULL on error
*
* <pre>
* Notes:
* (1) See sarrayUnionByAset() for the approach.
* (2) Here, the key in building the sorted tree is the number itself.
* (3) Operations using an underlying tree are O(nlogn), which is
* typically less efficient than hashing, which is O(n).
* </pre>
*/
L_DNA *
l_dnaUnionByAset(L_DNA *da1,
L_DNA *da2)
{
L_DNA *da3, *dad;
PROCNAME("l_dnaUnionByAset");
if (!da1)
return (L_DNA *)ERROR_PTR("da1 not defined", procName, NULL);
if (!da2)
return (L_DNA *)ERROR_PTR("da2 not defined", procName, NULL);
/* Join */
da3 = l_dnaCopy(da1);
l_dnaJoin(da3, da2, 0, -1);
/* Eliminate duplicates */
dad = l_dnaRemoveDupsByAset(da3);
l_dnaDestroy(&da3);
return dad;
}
l_int32 main(int argc,
char **argv)
{
L_ASET *set;
L_DNA *da1, *da2, *da3, *da4, *da5, *da6, *da7, *da8, *dav, *dac;
L_DNAHASH *dahash;
NUMA *nav, *nac;
PTA *pta1, *pta2, *pta3;
SARRAY *sa1, *sa2, *sa3, *sa4;
lept_mkdir("lept/hash");
#if 1
/* Test string hashing with aset */
fprintf(stderr, "Set results with string hashing:\n");
sa1 = BuildShortStrings(3, 0);
sa2 = BuildShortStrings(3, 1);
fprintf(stderr, " size with unique strings: %d\n", sarrayGetCount(sa1));
fprintf(stderr, " size with dups: %d\n", sarrayGetCount(sa2));
startTimer();
set = l_asetCreateFromSarray(sa2);
fprintf(stderr, " time to make set: %5.3f sec\n", stopTimer());
fprintf(stderr, " size of set without dups: %d\n", l_asetSize(set));
l_asetDestroy(&set);
startTimer();
sa3 = sarrayRemoveDupsByAset(sa2);
fprintf(stderr, " time to remove dups: %5.3f sec\n", stopTimer());
fprintf(stderr, " size without dups = %d\n", sarrayGetCount(sa3));
startTimer();
sa4 = sarrayIntersectionByAset(sa1, sa2);
fprintf(stderr, " time to intersect: %5.3f sec\n", stopTimer());
fprintf(stderr, " intersection size = %d\n", sarrayGetCount(sa4));
sarrayDestroy(&sa3);
sarrayDestroy(&sa4);
/* Test sarray set operations with dna hash.
* We use the same hash function as is used with aset. */
fprintf(stderr, "\nDna hash results for sarray:\n");
fprintf(stderr, " size with unique strings: %d\n", sarrayGetCount(sa1));
fprintf(stderr, " size with dups: %d\n", sarrayGetCount(sa2));
startTimer();
dahash = l_dnaHashCreateFromSarray(sa2);
fprintf(stderr, " time to make hashmap: %5.3f sec\n", stopTimer());
fprintf(stderr, " entries in hashmap with dups: %d\n",
l_dnaHashGetTotalCount(dahash));
l_dnaHashDestroy(&dahash);
startTimer();
sarrayRemoveDupsByHash(sa2, &sa3, NULL);
fprintf(stderr, " time to remove dups: %5.3f sec\n", stopTimer());
fprintf(stderr, " size without dups = %d\n", sarrayGetCount(sa3));
startTimer();
sa4 = sarrayIntersectionByHash(sa1, sa2);
fprintf(stderr, " time to intersect: %5.3f sec\n", stopTimer());
fprintf(stderr, " intersection size = %d\n", sarrayGetCount(sa4));
sarrayDestroy(&sa3);
sarrayDestroy(&sa4);
sarrayDestroy(&sa1);
sarrayDestroy(&sa2);
#endif
#if 1
/* Test point hashing with aset.
* Enter all points within a 1500 x 1500 image in pta1, and include
* 450,000 duplicates in pta2. With this pt hashing function,
* there are no hash collisions among any of the 400 million pixel
* locations in a 20000 x 20000 image. */
pta1 = BuildPointSet(1500, 1500, 0);
pta2 = BuildPointSet(1500, 1500, 1);
fprintf(stderr, "\nSet results for pta:\n");
fprintf(stderr, " pta1 size with unique points: %d\n", ptaGetCount(pta1));
fprintf(stderr, " pta2 size with dups: %d\n", ptaGetCount(pta2));
startTimer();
pta3 = ptaRemoveDupsByAset(pta2);
fprintf(stderr, " Time to remove dups: %5.3f sec\n", stopTimer());
fprintf(stderr, " size without dups = %d\n", ptaGetCount(pta3));
ptaDestroy(&pta3);
startTimer();
pta3 = ptaIntersectionByAset(pta1, pta2);
fprintf(stderr, " Time to intersect: %5.3f sec\n", stopTimer());
fprintf(stderr, " intersection size = %d\n", ptaGetCount(pta3));
ptaDestroy(&pta1);
ptaDestroy(&pta2);
ptaDestroy(&pta3);
#endif
#if 1
/* Test pta set operations with dna hash, using the same pt hashing
* function. Although there are no collisions in 20K x 20K images,
* the dna hash implementation works properly even if there are some. */
pta1 = BuildPointSet(1500, 1500, 0);
pta2 = BuildPointSet(1500, 1500, 1);
fprintf(stderr, "\nDna hash results for pta:\n");
fprintf(stderr, " pta1 size with unique points: %d\n", ptaGetCount(pta1));
fprintf(stderr, " pta2 size with dups: %d\n", ptaGetCount(pta2));
startTimer();
ptaRemoveDupsByHash(pta2, &pta3, NULL);
fprintf(stderr, " Time to remove dups: %5.3f sec\n", stopTimer());
fprintf(stderr, " size without dups = %d\n", ptaGetCount(pta3));
ptaDestroy(&pta3);
startTimer();
pta3 = ptaIntersectionByHash(pta1, pta2);
fprintf(stderr, " Time to intersect: %5.3f sec\n", stopTimer());
fprintf(stderr, " intersection size = %d\n", ptaGetCount(pta3));
ptaDestroy(&pta1);
ptaDestroy(&pta2);
ptaDestroy(&pta3);
#endif
/* Test dna set and histo operations using dna hash */
#if 1
fprintf(stderr, "\nDna hash results for dna:\n");
da1 = l_dnaMakeSequence(0.0, 0.125, 8000);
da2 = l_dnaMakeSequence(300.0, 0.125, 8000);
da3 = l_dnaMakeSequence(600.0, 0.125, 8000);
da4 = l_dnaMakeSequence(900.0, 0.125, 8000);
da5 = l_dnaMakeSequence(1200.0, 0.125, 8000);
l_dnaJoin(da1, da2, 0, -1);
l_dnaJoin(da1, da3, 0, -1);
l_dnaJoin(da1, da4, 0, -1);
l_dnaJoin(da1, da5, 0, -1);
l_dnaRemoveDupsByHash(da1, &da6, &dahash);
l_dnaHashDestroy(&dahash);
fprintf(stderr, " dna size with dups = %d\n", l_dnaGetCount(da1));
fprintf(stderr, " dna size of unique numbers = %d\n", l_dnaGetCount(da6));
l_dnaMakeHistoByHash(da1, &dahash, &dav, &dac);
nav = l_dnaConvertToNuma(dav);
nac = l_dnaConvertToNuma(dac);
fprintf(stderr, " dna number of histo points = %d\n", l_dnaGetCount(dac));
gplotSimpleXY1(nav, nac, GPLOT_IMPULSES, GPLOT_PNG,
"/tmp/lept/hash/histo", "Histo");
da7 = l_dnaIntersectionByHash(da2, da3);
fprintf(stderr, " dna number of points: da2 = %d, da3 = %d\n",
l_dnaGetCount(da2), l_dnaGetCount(da3));
fprintf(stderr, " dna number of da2/da3 intersection points = %d\n",
l_dnaGetCount(da7));
l_fileDisplay("/tmp/lept/hash/histo.png", 700, 100, 1.0);
l_dnaDestroy(&da1);
l_dnaDestroy(&da2);
l_dnaDestroy(&da3);
l_dnaDestroy(&da4);
l_dnaDestroy(&da5);
l_dnaDestroy(&da6);
l_dnaDestroy(&da7);
l_dnaDestroy(&dac);
l_dnaDestroy(&dav);
l_dnaHashDestroy(&dahash);
numaDestroy(&nav);
numaDestroy(&nac);
#endif
#if 1
da1 = l_dnaMakeSequence(0, 3, 10000);
da2 = l_dnaMakeSequence(0, 5, 10000);
da3 = l_dnaMakeSequence(0, 7, 10000);
l_dnaJoin(da1, da2, 0, -1);
l_dnaJoin(da1, da3, 0, -1);
fprintf(stderr, "\nDna results using set:\n");
fprintf(stderr, " da1 count: %d\n", l_dnaGetCount(da1));
set = l_asetCreateFromDna(da1);
fprintf(stderr, " da1 set size: %d\n\n", l_asetSize(set));
l_asetDestroy(&set);
da4 = l_dnaUnionByAset(da2, da3);
fprintf(stderr, " da4 count: %d\n", l_dnaGetCount(da4));
set = l_asetCreateFromDna(da4);
fprintf(stderr, " da4 set size: %d\n\n", l_asetSize(set));
l_asetDestroy(&set);
da5 = l_dnaIntersectionByAset(da1, da2);
fprintf(stderr, " da5 count: %d\n", l_dnaGetCount(da5));
set = l_asetCreateFromDna(da5);
fprintf(stderr, " da5 set size: %d\n\n", l_asetSize(set));
l_asetDestroy(&set);
da6 = l_dnaMakeSequence(100000, 11, 5000);
l_dnaJoin(da6, da1, 0, -1);
fprintf(stderr, " da6 count: %d\n", l_dnaGetCount(da6));
set = l_asetCreateFromDna(da6);
fprintf(stderr, " da6 set size: %d\n\n", l_asetSize(set));
l_asetDestroy(&set);
da7 = l_dnaIntersectionByAset(da6, da3);
fprintf(stderr, " da7 count: %d\n", l_dnaGetCount(da7));
set = l_asetCreateFromDna(da7);
fprintf(stderr, " da7 set size: %d\n\n", l_asetSize(set));
l_asetDestroy(&set);
da8 = l_dnaRemoveDupsByAset(da1);
fprintf(stderr, " da8 count: %d\n\n", l_dnaGetCount(da8));
l_dnaDestroy(&da1);
l_dnaDestroy(&da2);
l_dnaDestroy(&da3);
l_dnaDestroy(&da4);
l_dnaDestroy(&da5);
l_dnaDestroy(&da6);
l_dnaDestroy(&da7);
l_dnaDestroy(&da8);
#endif
return 0;
}
