2020-11-23 09:49:22 +01:00
|
|
|
|
|
|
|
package eu.dnetlib.doiboost.orcid;
|
|
|
|
|
|
|
|
import static eu.dnetlib.dhp.common.SparkSessionSupport.runWithSparkSession;
|
|
|
|
|
|
|
|
import java.text.SimpleDateFormat;
|
|
|
|
import java.util.Date;
|
|
|
|
import java.util.Optional;
|
|
|
|
|
|
|
|
import org.apache.commons.io.IOUtils;
|
2021-04-01 17:07:49 +02:00
|
|
|
import org.apache.commons.lang3.StringUtils;
|
2020-11-23 09:49:22 +01:00
|
|
|
import org.apache.hadoop.io.Text;
|
|
|
|
import org.apache.hadoop.mapreduce.lib.output.SequenceFileOutputFormat;
|
|
|
|
import org.apache.http.client.methods.CloseableHttpResponse;
|
|
|
|
import org.apache.http.client.methods.HttpGet;
|
|
|
|
import org.apache.http.impl.client.CloseableHttpClient;
|
|
|
|
import org.apache.http.impl.client.HttpClients;
|
|
|
|
import org.apache.spark.SparkConf;
|
|
|
|
import org.apache.spark.api.java.JavaPairRDD;
|
|
|
|
import org.apache.spark.api.java.JavaSparkContext;
|
|
|
|
import org.apache.spark.api.java.function.Function;
|
|
|
|
import org.apache.spark.util.LongAccumulator;
|
|
|
|
import org.slf4j.Logger;
|
|
|
|
import org.slf4j.LoggerFactory;
|
|
|
|
|
|
|
|
import eu.dnetlib.dhp.application.ArgumentApplicationParser;
|
|
|
|
import eu.dnetlib.doiboost.orcid.model.DownloadedRecordData;
|
2021-04-01 17:07:49 +02:00
|
|
|
import eu.dnetlib.doiboost.orcid.util.HDFSUtil;
|
2020-11-23 09:49:22 +01:00
|
|
|
import scala.Tuple2;
|
|
|
|
|
|
|
|
public class SparkDownloadOrcidAuthors {
|
|
|
|
|
|
|
|
static Logger logger = LoggerFactory.getLogger(SparkDownloadOrcidAuthors.class);
|
|
|
|
static final String DATE_FORMAT = "yyyy-MM-dd HH:mm:ss";
|
|
|
|
|
2020-12-23 16:59:52 +01:00
|
|
|
public static void main(String[] args) throws Exception {
|
2020-11-23 09:49:22 +01:00
|
|
|
|
|
|
|
final ArgumentApplicationParser parser = new ArgumentApplicationParser(
|
|
|
|
IOUtils
|
|
|
|
.toString(
|
|
|
|
SparkDownloadOrcidAuthors.class
|
|
|
|
.getResourceAsStream(
|
|
|
|
"/eu/dnetlib/dhp/doiboost/download_orcid_data.json")));
|
|
|
|
parser.parseArgument(args);
|
|
|
|
Boolean isSparkSessionManaged = Optional
|
|
|
|
.ofNullable(parser.get("isSparkSessionManaged"))
|
|
|
|
.map(Boolean::valueOf)
|
|
|
|
.orElse(Boolean.TRUE);
|
|
|
|
logger.info("isSparkSessionManaged: {}", isSparkSessionManaged);
|
|
|
|
final String workingPath = parser.get("workingPath");
|
2020-12-23 16:59:52 +01:00
|
|
|
logger.info("workingPath: {}", workingPath);
|
2020-11-23 09:49:22 +01:00
|
|
|
final String outputPath = parser.get("outputPath");
|
2020-12-23 16:59:52 +01:00
|
|
|
logger.info("outputPath: {}", outputPath);
|
2020-11-23 09:49:22 +01:00
|
|
|
final String token = parser.get("token");
|
|
|
|
final String lambdaFileName = parser.get("lambdaFileName");
|
2020-12-23 16:59:52 +01:00
|
|
|
logger.info("lambdaFileName: {}", lambdaFileName);
|
2021-04-01 17:07:49 +02:00
|
|
|
final String hdfsServerUri = parser.get("hdfsServerUri");
|
2020-11-23 09:49:22 +01:00
|
|
|
|
|
|
|
SparkConf conf = new SparkConf();
|
|
|
|
runWithSparkSession(
|
|
|
|
conf,
|
|
|
|
isSparkSessionManaged,
|
|
|
|
spark -> {
|
2021-04-01 17:07:49 +02:00
|
|
|
String lastUpdate = HDFSUtil.readFromTextFile(hdfsServerUri, workingPath, "last_update.txt");
|
|
|
|
logger.info("lastUpdate: {}", lastUpdate);
|
|
|
|
if (StringUtils.isBlank(lastUpdate)) {
|
|
|
|
throw new RuntimeException("last update info not found");
|
|
|
|
}
|
2020-11-23 09:49:22 +01:00
|
|
|
JavaSparkContext sc = JavaSparkContext.fromSparkContext(spark.sparkContext());
|
|
|
|
|
2020-11-23 16:01:23 +01:00
|
|
|
LongAccumulator parsedRecordsAcc = spark.sparkContext().longAccumulator("parsed_records");
|
|
|
|
LongAccumulator modifiedRecordsAcc = spark.sparkContext().longAccumulator("to_download_records");
|
|
|
|
LongAccumulator downloadedRecordsAcc = spark.sparkContext().longAccumulator("downloaded_records");
|
|
|
|
LongAccumulator errorHTTP403Acc = spark.sparkContext().longAccumulator("error_HTTP_403");
|
2021-04-01 17:07:49 +02:00
|
|
|
LongAccumulator errorHTTP404Acc = spark.sparkContext().longAccumulator("error_HTTP_404");
|
2020-11-23 16:01:23 +01:00
|
|
|
LongAccumulator errorHTTP409Acc = spark.sparkContext().longAccumulator("error_HTTP_409");
|
|
|
|
LongAccumulator errorHTTP503Acc = spark.sparkContext().longAccumulator("error_HTTP_503");
|
|
|
|
LongAccumulator errorHTTP525Acc = spark.sparkContext().longAccumulator("error_HTTP_525");
|
|
|
|
LongAccumulator errorHTTPGenericAcc = spark.sparkContext().longAccumulator("error_HTTP_Generic");
|
2020-11-23 09:49:22 +01:00
|
|
|
|
|
|
|
logger.info("Retrieving data from lamda sequence file");
|
|
|
|
JavaPairRDD<Text, Text> lamdaFileRDD = sc
|
|
|
|
.sequenceFile(workingPath + lambdaFileName, Text.class, Text.class);
|
2021-04-01 17:07:49 +02:00
|
|
|
final long lamdaFileRDDCount = lamdaFileRDD.count();
|
|
|
|
logger.info("Data retrieved: " + lamdaFileRDDCount);
|
2020-11-23 09:49:22 +01:00
|
|
|
|
|
|
|
Function<Tuple2<Text, Text>, Boolean> isModifiedAfterFilter = data -> {
|
|
|
|
String orcidId = data._1().toString();
|
|
|
|
String lastModifiedDate = data._2().toString();
|
|
|
|
parsedRecordsAcc.add(1);
|
2021-04-01 17:07:49 +02:00
|
|
|
if (isModified(orcidId, lastModifiedDate, lastUpdate)) {
|
2020-11-23 09:49:22 +01:00
|
|
|
modifiedRecordsAcc.add(1);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
};
|
|
|
|
|
|
|
|
Function<Tuple2<Text, Text>, Tuple2<String, String>> downloadRecordFunction = data -> {
|
|
|
|
String orcidId = data._1().toString();
|
|
|
|
String lastModifiedDate = data._2().toString();
|
|
|
|
final DownloadedRecordData downloaded = new DownloadedRecordData();
|
|
|
|
downloaded.setOrcidId(orcidId);
|
|
|
|
downloaded.setLastModifiedDate(lastModifiedDate);
|
2021-04-01 17:07:49 +02:00
|
|
|
CloseableHttpClient client = HttpClients.createDefault();
|
|
|
|
HttpGet httpGet = new HttpGet("https://api.orcid.org/v3.0/" + orcidId + "/record");
|
|
|
|
httpGet.addHeader("Accept", "application/vnd.orcid+xml");
|
|
|
|
httpGet.addHeader("Authorization", String.format("Bearer %s", token));
|
|
|
|
long startReq = System.currentTimeMillis();
|
|
|
|
CloseableHttpResponse response = client.execute(httpGet);
|
|
|
|
long endReq = System.currentTimeMillis();
|
|
|
|
long reqTime = endReq - startReq;
|
|
|
|
if (reqTime < 1000) {
|
|
|
|
Thread.sleep(1000 - reqTime);
|
|
|
|
}
|
|
|
|
int statusCode = response.getStatusLine().getStatusCode();
|
|
|
|
downloaded.setStatusCode(statusCode);
|
|
|
|
if (statusCode != 200) {
|
|
|
|
switch (statusCode) {
|
|
|
|
case 403:
|
|
|
|
errorHTTP403Acc.add(1);
|
|
|
|
case 404:
|
|
|
|
errorHTTP404Acc.add(1);
|
|
|
|
case 409:
|
|
|
|
errorHTTP409Acc.add(1);
|
|
|
|
case 503:
|
|
|
|
errorHTTP503Acc.add(1);
|
|
|
|
case 525:
|
|
|
|
errorHTTP525Acc.add(1);
|
|
|
|
default:
|
|
|
|
errorHTTPGenericAcc.add(1);
|
2020-11-23 09:49:22 +01:00
|
|
|
}
|
|
|
|
return downloaded.toTuple2();
|
|
|
|
}
|
2021-04-01 17:07:49 +02:00
|
|
|
downloadedRecordsAcc.add(1);
|
|
|
|
downloaded
|
|
|
|
.setCompressedData(
|
|
|
|
ArgumentApplicationParser
|
|
|
|
.compressArgument(IOUtils.toString(response.getEntity().getContent())));
|
|
|
|
client.close();
|
2020-11-23 09:49:22 +01:00
|
|
|
return downloaded.toTuple2();
|
|
|
|
};
|
|
|
|
|
|
|
|
sc.hadoopConfiguration().set("mapreduce.output.fileoutputformat.compress", "true");
|
|
|
|
|
|
|
|
logger.info("Start execution ...");
|
2020-11-23 16:01:23 +01:00
|
|
|
JavaPairRDD<Text, Text> authorsModifiedRDD = lamdaFileRDD.filter(isModifiedAfterFilter);
|
2021-04-01 17:07:49 +02:00
|
|
|
long authorsModifiedCount = authorsModifiedRDD.count();
|
|
|
|
logger.info("Authors modified count: " + authorsModifiedCount);
|
|
|
|
|
2020-11-23 09:49:22 +01:00
|
|
|
logger.info("Start downloading ...");
|
|
|
|
authorsModifiedRDD
|
2021-04-01 17:07:49 +02:00
|
|
|
.repartition(100)
|
2020-11-23 09:49:22 +01:00
|
|
|
.map(downloadRecordFunction)
|
|
|
|
.mapToPair(t -> new Tuple2(new Text(t._1()), new Text(t._2())))
|
|
|
|
.saveAsNewAPIHadoopFile(
|
|
|
|
workingPath.concat(outputPath),
|
|
|
|
Text.class,
|
|
|
|
Text.class,
|
|
|
|
SequenceFileOutputFormat.class,
|
|
|
|
sc.hadoopConfiguration());
|
2021-04-01 17:07:49 +02:00
|
|
|
|
2020-11-23 09:49:22 +01:00
|
|
|
logger.info("parsedRecordsAcc: " + parsedRecordsAcc.value().toString());
|
|
|
|
logger.info("modifiedRecordsAcc: " + modifiedRecordsAcc.value().toString());
|
|
|
|
logger.info("downloadedRecordsAcc: " + downloadedRecordsAcc.value().toString());
|
2020-11-23 16:01:23 +01:00
|
|
|
logger.info("errorHTTP403Acc: " + errorHTTP403Acc.value().toString());
|
2021-04-01 17:07:49 +02:00
|
|
|
logger.info("errorHTTP404Acc: " + errorHTTP404Acc.value().toString());
|
2020-11-23 16:01:23 +01:00
|
|
|
logger.info("errorHTTP409Acc: " + errorHTTP409Acc.value().toString());
|
|
|
|
logger.info("errorHTTP503Acc: " + errorHTTP503Acc.value().toString());
|
|
|
|
logger.info("errorHTTP525Acc: " + errorHTTP525Acc.value().toString());
|
|
|
|
logger.info("errorHTTPGenericAcc: " + errorHTTPGenericAcc.value().toString());
|
2020-11-23 09:49:22 +01:00
|
|
|
});
|
|
|
|
|
|
|
|
}
|
|
|
|
|
2021-04-01 17:07:49 +02:00
|
|
|
public static boolean isModified(String orcidId, String modifiedDate, String lastUpdate) {
|
2020-12-23 16:59:52 +01:00
|
|
|
Date modifiedDateDt;
|
|
|
|
Date lastUpdateDt;
|
2021-04-01 17:07:49 +02:00
|
|
|
String lastUpdateRedux = "";
|
2020-11-23 09:49:22 +01:00
|
|
|
try {
|
2021-04-01 17:07:49 +02:00
|
|
|
if (modifiedDate.equals("last_modified")) {
|
|
|
|
return false;
|
|
|
|
}
|
2020-11-23 09:49:22 +01:00
|
|
|
if (modifiedDate.length() != 19) {
|
|
|
|
modifiedDate = modifiedDate.substring(0, 19);
|
|
|
|
}
|
2021-04-01 17:07:49 +02:00
|
|
|
if (lastUpdate.length() != 19) {
|
|
|
|
lastUpdateRedux = lastUpdate.substring(0, 19);
|
|
|
|
} else {
|
|
|
|
lastUpdateRedux = lastUpdate;
|
|
|
|
}
|
2020-11-23 09:49:22 +01:00
|
|
|
modifiedDateDt = new SimpleDateFormat(DATE_FORMAT).parse(modifiedDate);
|
2021-04-01 17:07:49 +02:00
|
|
|
lastUpdateDt = new SimpleDateFormat(DATE_FORMAT).parse(lastUpdateRedux);
|
2020-11-23 09:49:22 +01:00
|
|
|
} catch (Exception e) {
|
2021-04-01 17:07:49 +02:00
|
|
|
throw new RuntimeException("[" + orcidId + "] modifiedDate <" + modifiedDate + "> lastUpdate <" + lastUpdate
|
|
|
|
+ "> Parsing date: " + e.getMessage());
|
2020-11-23 09:49:22 +01:00
|
|
|
}
|
|
|
|
return modifiedDateDt.after(lastUpdateDt);
|
|
|
|
}
|
|
|
|
}
|