2020-03-20 16:48:14 +01:00
|
|
|
package eu.dnetlib.dhp.dedup;
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2019-12-18 16:28:22 +01:00
|
|
|
import com.google.common.hash.HashFunction;
|
|
|
|
import com.google.common.hash.Hashing;
|
2019-12-11 15:43:24 +01:00
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
|
2019-12-06 13:38:00 +01:00
|
|
|
import org.apache.commons.io.IOUtils;
|
2020-03-23 17:13:30 +01:00
|
|
|
import org.apache.hadoop.conf.Configuration;
|
|
|
|
import org.apache.hadoop.fs.*;
|
2020-03-25 16:45:03 +01:00
|
|
|
import org.junit.jupiter.api.BeforeEach;
|
2019-12-06 13:38:00 +01:00
|
|
|
import org.junit.Ignore;
|
2020-03-25 16:45:03 +01:00
|
|
|
import org.junit.jupiter.api.Test;
|
2019-12-06 13:38:00 +01:00
|
|
|
|
|
|
|
import java.io.IOException;
|
|
|
|
|
|
|
|
public class SparkCreateDedupTest {
|
|
|
|
|
2019-12-10 14:57:16 +01:00
|
|
|
String configuration;
|
2020-03-18 17:41:56 +01:00
|
|
|
String entity = "organization";
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-03-25 16:45:03 +01:00
|
|
|
@BeforeEach
|
2019-12-06 13:38:00 +01:00
|
|
|
public void setUp() throws IOException {
|
2020-03-23 17:13:30 +01:00
|
|
|
// configuration = IOUtils.toString(getClass().getResourceAsStream("/eu/dnetlib/dedup/conf/org.curr.conf.json"));
|
|
|
|
configuration = "";
|
2019-12-10 14:57:16 +01:00
|
|
|
}
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-03-18 14:27:49 +01:00
|
|
|
@Test
|
|
|
|
@Ignore
|
2020-03-24 17:40:58 +01:00
|
|
|
public void createSimRelsTest() throws Exception {
|
2020-03-23 17:13:30 +01:00
|
|
|
SparkCreateSimRels.main(new String[]{
|
2020-03-18 14:27:49 +01:00
|
|
|
"-mt", "local[*]",
|
2020-03-24 17:40:58 +01:00
|
|
|
"-i", "/Users/miconis/dumps",
|
|
|
|
"-o", "/tmp/dedup/rawset_test",
|
2020-03-18 17:41:56 +01:00
|
|
|
"-asi", "dedup-similarity-result-levenstein",
|
|
|
|
"-la", "lookupurl",
|
2020-03-24 17:40:58 +01:00
|
|
|
"-w", "workingPath"
|
2020-03-18 14:27:49 +01:00
|
|
|
});
|
|
|
|
}
|
|
|
|
|
2019-12-10 14:57:16 +01:00
|
|
|
@Test
|
|
|
|
@Ignore
|
|
|
|
public void createCCTest() throws Exception {
|
2019-12-06 13:38:00 +01:00
|
|
|
|
2020-03-23 17:13:30 +01:00
|
|
|
SparkCreateConnectedComponent.main(new String[]{
|
2019-12-06 13:38:00 +01:00
|
|
|
"-mt", "local[*]",
|
2020-01-14 14:01:03 +01:00
|
|
|
"-s", "/Users/miconis/dumps",
|
2019-12-11 16:57:37 +01:00
|
|
|
"-e", entity,
|
2019-12-11 15:43:24 +01:00
|
|
|
"-c", ArgumentApplicationParser.compressArgument(configuration),
|
2019-12-06 13:38:00 +01:00
|
|
|
"-t", "/tmp/dedup",
|
|
|
|
});
|
|
|
|
}
|
|
|
|
|
|
|
|
@Test
|
|
|
|
@Ignore
|
|
|
|
public void dedupRecordTest() throws Exception {
|
2020-03-23 17:13:30 +01:00
|
|
|
SparkCreateDedupRecord.main(new String[]{
|
2019-12-06 13:38:00 +01:00
|
|
|
"-mt", "local[*]",
|
2020-01-14 14:01:03 +01:00
|
|
|
"-s", "/Users/miconis/dumps",
|
2019-12-11 16:57:37 +01:00
|
|
|
"-e", entity,
|
2019-12-11 15:43:24 +01:00
|
|
|
"-c", ArgumentApplicationParser.compressArgument(configuration),
|
|
|
|
"-d", "/tmp/dedup",
|
2019-12-06 13:38:00 +01:00
|
|
|
});
|
|
|
|
}
|
2019-12-11 15:43:24 +01:00
|
|
|
|
|
|
|
@Test
|
2020-03-23 17:13:30 +01:00
|
|
|
@Ignore
|
2020-01-14 14:01:03 +01:00
|
|
|
public void printConfiguration() throws Exception {
|
2019-12-11 15:43:24 +01:00
|
|
|
System.out.println(ArgumentApplicationParser.compressArgument(configuration));
|
|
|
|
}
|
|
|
|
|
2019-12-18 16:28:22 +01:00
|
|
|
@Test
|
2020-03-23 17:13:30 +01:00
|
|
|
@Ignore
|
2019-12-18 16:28:22 +01:00
|
|
|
public void testHashCode() {
|
|
|
|
final String s1 = "20|grid________::6031f94bef015a37783268ec1e75f17f";
|
|
|
|
final String s2 = "20|nsf_________::b12be9edf414df8ee66b4c52a2d8da46";
|
|
|
|
|
|
|
|
final HashFunction hashFunction = Hashing.murmur3_128();
|
|
|
|
|
2020-03-23 17:13:30 +01:00
|
|
|
System.out.println(s1.hashCode());
|
2020-03-20 16:48:14 +01:00
|
|
|
System.out.println(hashFunction.hashString(s1).asLong());
|
2020-03-23 17:13:30 +01:00
|
|
|
System.out.println(s2.hashCode());
|
2020-03-20 16:48:14 +01:00
|
|
|
System.out.println(hashFunction.hashString(s2).asLong());
|
2020-03-20 13:01:56 +01:00
|
|
|
}
|
2019-12-06 13:38:00 +01:00
|
|
|
}
|