Revision 53662
Added by Claudio Atzori almost 6 years ago
| modules/dnet-collector-plugins/trunk/src/test/java/eu/dnetlib/data/collector/plugins/schemaorg/SchemaOrgMainKaggle.java | ||
|---|---|---|
| 1 |
package eu.dnetlib.data.collector.plugins.schemaorg; |
|
| 2 |
|
|
| 3 |
import eu.dnetlib.data.collector.rmi.InterfaceDescriptor; |
|
| 4 |
import org.apache.commons.io.FileUtils; |
|
| 5 |
import org.apache.commons.lang3.StringUtils; |
|
| 6 |
import org.apache.commons.logging.Log; |
|
| 7 |
import org.apache.commons.logging.LogFactory; |
|
| 8 |
import org.apache.log4j.ConsoleAppender; |
|
| 9 |
import org.apache.log4j.Level; |
|
| 10 |
import org.apache.log4j.Logger; |
|
| 11 |
import org.apache.log4j.PatternLayout; |
|
| 12 |
import org.dom4j.Document; |
|
| 13 |
import org.dom4j.io.SAXReader; |
|
| 14 |
|
|
| 15 |
import java.io.File; |
|
| 16 |
import java.io.FileWriter; |
|
| 17 |
import java.io.StringReader; |
|
| 18 |
import java.nio.charset.StandardCharsets; |
|
| 19 |
import java.util.HashMap; |
|
| 20 |
import java.util.concurrent.TimeUnit; |
|
| 21 |
|
|
| 22 |
public class SchemaOrgMainKaggle {
|
|
| 23 |
|
|
| 24 |
private static final Log log = LogFactory.getLog(SchemaOrgMainKaggle.class); |
|
| 25 |
|
|
| 26 |
public static void main(String[] args) throws Exception {
|
|
| 27 |
|
|
| 28 |
ConsoleAppender console = new ConsoleAppender(); |
|
| 29 |
console.setLayout(new PatternLayout("%d [%p|%c|%C{1}] %m%n"));
|
|
| 30 |
console.setThreshold(Level.DEBUG); |
|
| 31 |
console.activateOptions(); |
|
| 32 |
Logger.getLogger("eu.dnetlib.data.collector.plugins").addAppender(console);
|
|
| 33 |
|
|
| 34 |
HashMap<String,String> params = new HashMap<>(); |
|
| 35 |
params.put("consumerBlockPolling", Boolean.toString(true));
|
|
| 36 |
params.put("consumerBlockPollingTimeout", "2");
|
|
| 37 |
params.put("consumerBlockPollingTimeoutUnit", TimeUnit.MINUTES.toString());
|
|
| 38 |
params.put("endpointCharset", StandardCharsets.UTF_8.name());
|
|
| 39 |
params.put("updatedDateFormat", "YYYY-MM-DD");
|
|
| 40 |
params.put("createdDateFormat", "YYYY-MM-DD");
|
|
| 41 |
params.put("publicationDateFormat", "YYYY-MM-DD");
|
|
| 42 |
params.put("contributorFallbackType", DatasetDocument.Contributor.ContributorType.Other.toString());
|
|
| 43 |
params.put("identifierFallbackType", "Handle");
|
|
| 44 |
params.put("identifierFallbackURL", Boolean.toString(true));
|
|
| 45 |
params.put("identifierMappingARK", "ark, ARK");
|
|
| 46 |
params.put("identifierMappingDOI", "doi, DOI");
|
|
| 47 |
params.put("identifierMappingHandle", "Handle, HANDLE");
|
|
| 48 |
params.put("identifierMappingPURL", "purl, PURL");
|
|
| 49 |
params.put("identifierMappingURN", "urn, URN");
|
|
| 50 |
params.put("identifierMappingURL", "url, URL");
|
|
| 51 |
|
|
| 52 |
params.put("repositoryAccessType", "httpapi-kaggle");
|
|
| 53 |
|
|
| 54 |
params.put("httpapi-kaggle_queueSize", "100");
|
|
| 55 |
params.put("httpapi-kaggle_APICharset", StandardCharsets.UTF_8.name());
|
|
| 56 |
params.put("httpapi-kaggle_queryUrl", "https://www.kaggle.com/datasets_v2.json?sortBy=updated&group=public&page={PAGE}&pageSize=20&size=sizeAll&filetype=fileTypeAll&license=licenseAll");
|
|
| 57 |
params.put("httpapi-kaggle_queryPagePlaceholder", "{PAGE}");
|
|
| 58 |
params.put("httpapi-kaggle_responsePropertyTotalDataset", "totalDatasetListItems");
|
|
| 59 |
params.put("httpapi-kaggle_responsePropertyDatasetList", "datasetListItems");
|
|
| 60 |
params.put("httpapi-kaggle_responsePropertyDatasetUrl", "datasetUrl");
|
|
| 61 |
params.put("httpapi-kaggle_responseBaseDatasetUrl", "https://www.kaggle.com");
|
|
| 62 |
|
|
| 63 |
InterfaceDescriptor descriptor = new InterfaceDescriptor(); |
|
| 64 |
descriptor.setId("schema.org - kaggle");
|
|
| 65 |
descriptor.setBaseUrl("https://www.kaggle.com");
|
|
| 66 |
|
|
| 67 |
descriptor.setParams(params); |
|
| 68 |
|
|
| 69 |
SchemaOrgPlugin schemaOrgPlugin = new SchemaOrgPlugin(); |
|
| 70 |
|
|
| 71 |
Iterable<String> iterable = schemaOrgPlugin.collect(descriptor, null, null); |
|
| 72 |
|
|
| 73 |
String outDir = params.get("repositoryAccessType");
|
|
| 74 |
|
|
| 75 |
log.info("saving content in " + outDir);
|
|
| 76 |
|
|
| 77 |
File directory = new File(outDir); |
|
| 78 |
if (directory.exists()) {
|
|
| 79 |
log.info(directory.getAbsolutePath() + " exists, cleaning up"); |
|
| 80 |
FileUtils.deleteDirectory(directory); |
|
| 81 |
} |
|
| 82 |
FileUtils.forceMkdir(directory); |
|
| 83 |
|
|
| 84 |
int skipped = 0; |
|
| 85 |
for(String item : iterable) {
|
|
| 86 |
|
|
| 87 |
final Document doc = new SAXReader().read(new StringReader(item)); |
|
| 88 |
|
|
| 89 |
String id = doc.valueOf("/*[local-name() = 'dataset']/*[local-name() = 'identifier']/text()");
|
|
| 90 |
if (StringUtils.isNotBlank(id)) {
|
|
| 91 |
log.info(item); |
|
| 92 |
String fileName = outDir + "/" + id; |
|
| 93 |
FileWriter w = new FileWriter(fileName); |
|
| 94 |
w.write(item); |
|
| 95 |
w.close(); |
|
| 96 |
log.info("wrote " + fileName);
|
|
| 97 |
} else {
|
|
| 98 |
skipped++; |
|
| 99 |
} |
|
| 100 |
if (skipped % 100 == 0) {
|
|
| 101 |
log.info("skipped so far " + skipped);
|
|
| 102 |
} |
|
| 103 |
} |
|
| 104 |
|
|
| 105 |
log.info("Done! skipped " + skipped);
|
|
| 106 |
} |
|
| 107 |
|
|
| 108 |
} |
|
| modules/dnet-collector-plugins/trunk/src/test/java/eu/dnetlib/data/collector/plugins/schemaorg/SchemaOrgMainReactome.java | ||
|---|---|---|
| 1 |
package eu.dnetlib.data.collector.plugins.schemaorg; |
|
| 2 |
|
|
| 3 |
import eu.dnetlib.data.collector.plugins.schemaorg.sitemapindex.SitemapFileIterator; |
|
| 4 |
import eu.dnetlib.data.collector.rmi.InterfaceDescriptor; |
|
| 5 |
import org.apache.commons.io.FileUtils; |
|
| 6 |
import org.apache.commons.lang3.StringUtils; |
|
| 7 |
import org.apache.commons.logging.Log; |
|
| 8 |
import org.apache.commons.logging.LogFactory; |
|
| 9 |
import org.apache.log4j.ConsoleAppender; |
|
| 10 |
import org.apache.log4j.Level; |
|
| 11 |
import org.apache.log4j.Logger; |
|
| 12 |
import org.apache.log4j.PatternLayout; |
|
| 13 |
import org.dom4j.Document; |
|
| 14 |
import org.dom4j.io.SAXReader; |
|
| 15 |
|
|
| 16 |
import java.io.File; |
|
| 17 |
import java.io.FileWriter; |
|
| 18 |
import java.io.StringReader; |
|
| 19 |
import java.nio.charset.StandardCharsets; |
|
| 20 |
import java.util.HashMap; |
|
| 21 |
import java.util.concurrent.TimeUnit; |
|
| 22 |
|
|
| 23 |
public class SchemaOrgMainReactome {
|
|
| 24 |
|
|
| 25 |
private static final Log log = LogFactory.getLog(SchemaOrgMainReactome.class); |
|
| 26 |
|
|
| 27 |
public static void main(String[] args) throws Exception {
|
|
| 28 |
|
|
| 29 |
ConsoleAppender console = new ConsoleAppender(); |
|
| 30 |
console.setLayout(new PatternLayout("%d [%p|%c|%C{1}] %m%n"));
|
|
| 31 |
console.setThreshold(Level.DEBUG); |
|
| 32 |
console.activateOptions(); |
|
| 33 |
Logger.getLogger("eu.dnetlib.data.collector.plugins").addAppender(console);
|
|
| 34 |
|
|
| 35 |
HashMap<String,String> params = new HashMap<>(); |
|
| 36 |
params.put("consumerBlockPolling", Boolean.toString(true));
|
|
| 37 |
params.put("consumerBlockPollingTimeout", "2");
|
|
| 38 |
params.put("consumerBlockPollingTimeoutUnit", TimeUnit.MINUTES.toString());
|
|
| 39 |
params.put("endpointCharset", StandardCharsets.UTF_8.name());
|
|
| 40 |
params.put("updatedDateFormat", "YYYY-MM-DD");
|
|
| 41 |
params.put("createdDateFormat", "YYYY-MM-DD");
|
|
| 42 |
params.put("publicationDateFormat", "YYYY-MM-DD");
|
|
| 43 |
params.put("contributorFallbackType", DatasetDocument.Contributor.ContributorType.Other.toString());
|
|
| 44 |
params.put("identifierFallbackType", "Handle");
|
|
| 45 |
params.put("identifierFallbackURL", Boolean.toString(true));
|
|
| 46 |
params.put("identifierMappingARK", "ark, ARK");
|
|
| 47 |
params.put("identifierMappingDOI", "doi, DOI");
|
|
| 48 |
params.put("identifierMappingHandle", "Handle, HANDLE");
|
|
| 49 |
params.put("identifierMappingPURL", "purl, PURL");
|
|
| 50 |
params.put("identifierMappingURN", "urn, URN");
|
|
| 51 |
params.put("identifierMappingURL", "url, URL");
|
|
| 52 |
|
|
| 53 |
params.put("repositoryAccessType", "sitemapindex");
|
|
| 54 |
params.put("sitemap_queueSize", "100");
|
|
| 55 |
params.put("sitemap_IndexCharset", StandardCharsets.UTF_8.name());
|
|
| 56 |
params.put("sitemap_FileCharset", StandardCharsets.UTF_8.name());
|
|
| 57 |
params.put("sitemap_FileSchema", SitemapFileIterator.Options.SitemapSchemaType.Text.toString());
|
|
| 58 |
params.put("sitemap_FileType", SitemapFileIterator.Options.SitemapFileType.GZ.toString());
|
|
| 59 |
|
|
| 60 |
|
|
| 61 |
|
|
| 62 |
InterfaceDescriptor descriptor = new InterfaceDescriptor(); |
|
| 63 |
descriptor.setId("schema.org - reactome");
|
|
| 64 |
descriptor.setBaseUrl("https://reactome.org/sitemapindex.xml");
|
|
| 65 |
|
|
| 66 |
descriptor.setParams(params); |
|
| 67 |
|
|
| 68 |
SchemaOrgPlugin schemaOrgPlugin = new SchemaOrgPlugin(); |
|
| 69 |
|
|
| 70 |
Iterable<String> iterable = schemaOrgPlugin.collect(descriptor, null, null); |
|
| 71 |
|
|
| 72 |
String outDir = params.get("repositoryAccessType");
|
|
| 73 |
|
|
| 74 |
log.info("saving content in " + outDir);
|
|
| 75 |
|
|
| 76 |
File directory = new File(outDir); |
|
| 77 |
if (directory.exists()) {
|
|
| 78 |
log.info(directory.getAbsolutePath() + " exists, cleaning up"); |
|
| 79 |
FileUtils.deleteDirectory(directory); |
|
| 80 |
} |
|
| 81 |
FileUtils.forceMkdir(directory); |
|
| 82 |
|
|
| 83 |
int skipped = 0; |
|
| 84 |
for(String item : iterable) {
|
|
| 85 |
|
|
| 86 |
final Document doc = new SAXReader().read(new StringReader(item)); |
|
| 87 |
|
|
| 88 |
String id = doc.valueOf("/*[local-name() = 'dataset']/*[local-name() = 'identifier']/text()");
|
|
| 89 |
if (StringUtils.isNotBlank(id)) {
|
|
| 90 |
log.info(item); |
|
| 91 |
String fileName = outDir + "/" + id; |
|
| 92 |
FileWriter w = new FileWriter(fileName); |
|
| 93 |
w.write(item); |
|
| 94 |
w.close(); |
|
| 95 |
log.info("wrote " + fileName);
|
|
| 96 |
} else {
|
|
| 97 |
skipped++; |
|
| 98 |
} |
|
| 99 |
if (skipped % 100 == 0) {
|
|
| 100 |
log.info("skipped so far " + skipped);
|
|
| 101 |
} |
|
| 102 |
} |
|
| 103 |
|
|
| 104 |
log.info("Done! skipped " + skipped);
|
|
| 105 |
} |
|
| 106 |
|
|
| 107 |
} |
|
| modules/dnet-collector-plugins/trunk/pom.xml | ||
|---|---|---|
| 11 | 11 |
<scm> |
| 12 | 12 |
<developerConnection>scm:svn:https://svn.driver.research-infrastructures.eu/driver/dnet45/modules/dnet-collector-plugins/trunk</developerConnection> |
| 13 | 13 |
</scm> |
| 14 |
|
|
| 15 |
<build> |
|
| 16 |
<plugins> |
|
| 17 |
<plugin> |
|
| 18 |
<artifactId>maven-assembly-plugin</artifactId> |
|
| 19 |
<configuration> |
|
| 20 |
<archive> |
|
| 21 |
<manifest> |
|
| 22 |
<mainClass>eu.dnetlib.data.collector.plugins.schemaorg.SchemaOrgMainReactome</mainClass> |
|
| 23 |
</manifest> |
|
| 24 |
</archive> |
|
| 25 |
<descriptorRefs> |
|
| 26 |
<descriptorRef>jar-with-dependencies</descriptorRef> |
|
| 27 |
</descriptorRefs> |
|
| 28 |
</configuration> |
|
| 29 |
</plugin> |
|
| 30 |
</plugins> |
|
| 31 |
</build> |
|
| 32 |
|
|
| 14 | 33 |
<dependencies> |
| 15 | 34 |
<dependency> |
| 16 | 35 |
<groupId>eu.dnetlib</groupId> |
Also available in: Unified diff
added main classes to verify the content collected from Kaggle and Reactome