dnet-core/dnet-data-services/src/main/java/eu/dnetlib/data/collector/plugins/schemaorg/SchemaOrgPlugin.java

154 lines
12 KiB
Java

package eu.dnetlib.data.collector.plugins.schemaorg;
import eu.dnetlib.data.collector.plugin.AbstractCollectorPlugin;
import eu.dnetlib.data.collector.plugins.schemaorg.httpapi.kaggle.KaggleRepositoryIterable;
import eu.dnetlib.data.collector.plugins.schemaorg.sitemapindex.SitemapFileIterator;
import eu.dnetlib.data.collector.plugins.schemaorg.sitemapindex.SitemapIndexIterator;
import eu.dnetlib.data.collector.plugins.schemaorg.sitemapindex.SitemapIndexRepositoryIterable;
import eu.dnetlib.data.collector.rmi.CollectorServiceException;
import eu.dnetlib.data.collector.rmi.InterfaceDescriptor;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import java.net.MalformedURLException;
import java.net.URL;
import java.nio.charset.StandardCharsets;
import java.util.concurrent.TimeUnit;
public class SchemaOrgPlugin extends AbstractCollectorPlugin {
private static final Log log = LogFactory.getLog(SchemaOrgPlugin.class);
public String hello(){
return "hello";
}
@Override
public Iterable<String> collect(final InterfaceDescriptor interfaceDescriptor, final String fromDate, final String untilDate) throws CollectorServiceException {
try {
RepositoryIterable repository = null;
String repositoryAccessType = Utils.getAsString(interfaceDescriptor.getParams(), "repositoryAccessType", null);
switch(repositoryAccessType) {
case "sitemapindex": {
SitemapIndexRepositoryIterable.Options repositoryOptions = this.compileSitemapIndexRepositoryOptions(interfaceDescriptor);
SitemapIndexRepositoryIterable repositoryIterable = new SitemapIndexRepositoryIterable(repositoryOptions);
repositoryIterable.bootstrap();
repository = repositoryIterable;
break;
}
case "httpapi-kaggle": {
KaggleRepositoryIterable.Options repositoryOptions = this.compileKaggleRepositoryOptions(interfaceDescriptor);
KaggleRepositoryIterable repositoryIterable = new KaggleRepositoryIterable(repositoryOptions);
repositoryIterable.bootstrap();
repository = repositoryIterable;
break;
}
default:
throw new CollectorServiceException(String.format("unrecognized repository access type ", repositoryAccessType));
}
SchemaOrgIterable.Options schemaOrgOptions = this.compileSchemaOrgOptions(interfaceDescriptor);
SchemaOrgIterable iterable = new SchemaOrgIterable(schemaOrgOptions, repository);
return iterable;
} catch (Exception e) {
throw new CollectorServiceException("Could not create iterator", e);
}
}
private KaggleRepositoryIterable.Options compileKaggleRepositoryOptions(InterfaceDescriptor interfaceDescriptor) throws MalformedURLException {
KaggleRepositoryIterable.Options kaggleRepositoryOptions = new KaggleRepositoryIterable.Options();
kaggleRepositoryOptions.setQueueSize(Utils.getAsInt(interfaceDescriptor.getParams(), "httpapi-kaggle_queueSize", 100));
kaggleRepositoryOptions.setPutTimeout(Utils.getAsLong(interfaceDescriptor.getParams(), "httpapi-kaggle_producerBlockPollingTimeout", 20));
kaggleRepositoryOptions.setPutTimeoutUnit(Utils.getAsEnum(interfaceDescriptor.getParams(), "httpapi-kaggle_producerBlockPollingTimeoutUnit", TimeUnit.MINUTES, TimeUnit.class));
kaggleRepositoryOptions.setCharset(Utils.getAsCharset(interfaceDescriptor.getParams(), "httpapi-kaggle_APICharset", StandardCharsets.UTF_8));
kaggleRepositoryOptions.setQueryUrl(Utils.getAsString(interfaceDescriptor.getParams(), "httpapi-kaggle_queryUrl", null));
kaggleRepositoryOptions.setQueryPagePlaceholder(Utils.getAsString(interfaceDescriptor.getParams(), "httpapi-kaggle_queryPagePlaceholder", "{PAGE}"));
kaggleRepositoryOptions.setResponsePropertyTotalDataset(Utils.getAsString(interfaceDescriptor.getParams(), "httpapi-kaggle_responsePropertyTotalDataset", "totalDatasetListItems"));
kaggleRepositoryOptions.setResponsePropertyDatasetList(Utils.getAsString(interfaceDescriptor.getParams(), "httpapi-kaggle_responsePropertyDatasetList", "datasetListItems"));
kaggleRepositoryOptions.setResponsePropertyDatasetUrl(Utils.getAsString(interfaceDescriptor.getParams(), "httpapi-kaggle_responsePropertyDatasetUrl", "datasetUrl"));
kaggleRepositoryOptions.setResponseBaseDatasetUrl(Utils.getAsString(interfaceDescriptor.getParams(), "httpapi-kaggle_responseBaseDatasetUrl", interfaceDescriptor.getBaseUrl()));
kaggleRepositoryOptions.setRepositoryQueueIteratorOptions(this.compileRepositoryQueueOptions(interfaceDescriptor));
return kaggleRepositoryOptions;
}
private SitemapIndexIterator.Options compileSitemapIndexOptions(InterfaceDescriptor interfaceDescriptor) throws MalformedURLException {
SitemapIndexIterator.Options sitemapIndexIteratorOptions = new SitemapIndexIterator.Options();
sitemapIndexIteratorOptions.setCharset(Utils.getAsCharset(interfaceDescriptor.getParams(), "sitemap_IndexCharset", StandardCharsets.UTF_8));
sitemapIndexIteratorOptions.setIndexUrl(new URL(interfaceDescriptor.getBaseUrl()));
return sitemapIndexIteratorOptions;
}
private SitemapFileIterator.Options compileSitemapFileOptions(InterfaceDescriptor interfaceDescriptor) throws MalformedURLException {
SitemapFileIterator.Options sitemapFileIteratorOptions = new SitemapFileIterator.Options();
sitemapFileIteratorOptions.setCharset(Utils.getAsCharset(interfaceDescriptor.getParams(), "sitemap_FileCharset", StandardCharsets.UTF_8));
sitemapFileIteratorOptions.setSchemaType(Utils.getAsEnum(interfaceDescriptor.getParams(), "sitemap_FileSchema", SitemapFileIterator.Options.SitemapSchemaType.Xml, SitemapFileIterator.Options.SitemapSchemaType.class));
sitemapFileIteratorOptions.setFileType(Utils.getAsEnum(interfaceDescriptor.getParams(), "sitemap_FileType", SitemapFileIterator.Options.SitemapFileType.Text, SitemapFileIterator.Options.SitemapFileType.class));
return sitemapFileIteratorOptions;
}
private RepositoryQueueIterator.Options compileRepositoryQueueOptions(InterfaceDescriptor interfaceDescriptor) throws MalformedURLException {
RepositoryQueueIterator.Options repositoryQueueIteratorOptions = new RepositoryQueueIterator.Options();
repositoryQueueIteratorOptions.setBlockPolling(Utils.getAsBoolean(interfaceDescriptor.getParams(), "consumerBlockPolling", true));
repositoryQueueIteratorOptions.setPollTimeout(Utils.getAsLong(interfaceDescriptor.getParams(), "consumerBlockPollingTimeout", 2));
repositoryQueueIteratorOptions.setPollTimeoutUnit(Utils.getAsEnum(interfaceDescriptor.getParams(), "consumerBlockPollingTimeoutUnit", TimeUnit.MINUTES, TimeUnit.class));
return repositoryQueueIteratorOptions;
}
private SitemapIndexRepositoryIterable.Options compileSitemapIndexRepositoryOptions(InterfaceDescriptor interfaceDescriptor) throws MalformedURLException {
SitemapIndexRepositoryIterable.Options sitemapIndexRepositoryIterableOptions = new SitemapIndexRepositoryIterable.Options();
sitemapIndexRepositoryIterableOptions.setQueueSize(Utils.getAsInt(interfaceDescriptor.getParams(), "sitemap_queueSize", 100));
sitemapIndexRepositoryIterableOptions.setPutTimeout(Utils.getAsLong(interfaceDescriptor.getParams(), "sitemap_producerBlockPollingTimeout", 20));
sitemapIndexRepositoryIterableOptions.setPutTimeoutUnit(Utils.getAsEnum(interfaceDescriptor.getParams(), "sitemap_producerBlockPollingTimeoutUnit", TimeUnit.MINUTES, TimeUnit.class));
sitemapIndexRepositoryIterableOptions.setRepositoryQueueIteratorOptions(this.compileRepositoryQueueOptions(interfaceDescriptor));
sitemapIndexRepositoryIterableOptions.setSitemapFileIteratorOptions(this.compileSitemapFileOptions(interfaceDescriptor));
sitemapIndexRepositoryIterableOptions.setSitemapIndexIteratorOptions(this.compileSitemapIndexOptions(interfaceDescriptor));
return sitemapIndexRepositoryIterableOptions;
}
private EndpointAccessIterator.Options compileEndpointAccessOptions(InterfaceDescriptor interfaceDescriptor) throws MalformedURLException {
EndpointAccessIterator.Options endpointAccessIteratorOptions = new EndpointAccessIterator.Options();
endpointAccessIteratorOptions.setCharset(Utils.getAsCharset(interfaceDescriptor.getParams(), "endpointCharset", StandardCharsets.UTF_8));
return endpointAccessIteratorOptions;
}
private DatasetMappingIterator.Options compileDatasetMappingOptions(InterfaceDescriptor interfaceDescriptor) throws MalformedURLException {
DatasetMappingIterator.Options datasetMappingIteratorOptions = new DatasetMappingIterator.Options();
DatasetMappingIterator.Options.UpdatedDateOptions datasetMappingIteratorUpdatedDateOptions = new DatasetMappingIterator.Options.UpdatedDateOptions();
datasetMappingIteratorUpdatedDateOptions.format =Utils.getAsString(interfaceDescriptor.getParams(), "updatedDateFormat", "YYYY-MM-DD");
datasetMappingIteratorOptions.setUpdatedDateOptions(datasetMappingIteratorUpdatedDateOptions);
DatasetMappingIterator.Options.CreatedDateOptions datasetMappingIteratorCreatedDateOptions = new DatasetMappingIterator.Options.CreatedDateOptions();
datasetMappingIteratorCreatedDateOptions.format =Utils.getAsString(interfaceDescriptor.getParams(), "createdDateFormat", "YYYY-MM-DD");
datasetMappingIteratorOptions.setCreatedDateOptions(datasetMappingIteratorCreatedDateOptions);
DatasetMappingIterator.Options.PublicationDateOptions datasetMappingIteratorPublicationDateOptions = new DatasetMappingIterator.Options.PublicationDateOptions();
datasetMappingIteratorPublicationDateOptions.format =Utils.getAsString(interfaceDescriptor.getParams(), "publicationDateFormat", "YYYY-MM-DD");
datasetMappingIteratorOptions.setPublicationDateOptions(datasetMappingIteratorPublicationDateOptions);
DatasetMappingIterator.Options.ContributorOptions datasetMappingIteratorContributorOptions = new DatasetMappingIterator.Options.ContributorOptions();
datasetMappingIteratorContributorOptions.fallbackType =Utils.getAsEnum(interfaceDescriptor.getParams(), "contributorFallbackType",DatasetDocument.Contributor.ContributorType.Other, DatasetDocument.Contributor.ContributorType.class);
datasetMappingIteratorOptions.setContributorOptions(datasetMappingIteratorContributorOptions);
DatasetMappingIterator.Options.IdentifierOptions datasetMappingIteratorIdentifierOptions = new DatasetMappingIterator.Options.IdentifierOptions();
datasetMappingIteratorIdentifierOptions.fallbackType = Utils.getAsEnum(interfaceDescriptor.getParams(), "identifierFallbackType", null, DatasetDocument.Identifier.IdentifierType.class);
datasetMappingIteratorIdentifierOptions.fallbackURL = Utils.getAsBoolean(interfaceDescriptor.getParams(), "identifierFallbackURL", true);
datasetMappingIteratorIdentifierOptions.mappingARK = Utils.getAsStringCsv(interfaceDescriptor.getParams(), "identifierMappingARK", null);
datasetMappingIteratorIdentifierOptions.mappingDOI = Utils.getAsStringCsv(interfaceDescriptor.getParams(), "identifierMappingDOI", null);
datasetMappingIteratorIdentifierOptions.mappingHandle = Utils.getAsStringCsv(interfaceDescriptor.getParams(), "identifierMappingHandle", null);
datasetMappingIteratorIdentifierOptions.mappingPURL = Utils.getAsStringCsv(interfaceDescriptor.getParams(), "identifierMappingPURL", null);
datasetMappingIteratorIdentifierOptions.mappingURL = Utils.getAsStringCsv(interfaceDescriptor.getParams(), "identifierMappingURL", null);
datasetMappingIteratorIdentifierOptions.mappingURN = Utils.getAsStringCsv(interfaceDescriptor.getParams(), "identifierMappingURN", null);
datasetMappingIteratorOptions.setIdentifierOptions(datasetMappingIteratorIdentifierOptions);
return datasetMappingIteratorOptions;
}
private SchemaOrgIterable.Options compileSchemaOrgOptions(InterfaceDescriptor interfaceDescriptor) throws MalformedURLException {
SchemaOrgIterable.Options schemaOrgIterableOptions = new SchemaOrgIterable.Options();
schemaOrgIterableOptions.setDatasetMappingOptions(this.compileDatasetMappingOptions(interfaceDescriptor));
schemaOrgIterableOptions.setEndpointAccessOptions(this.compileEndpointAccessOptions(interfaceDescriptor));
return schemaOrgIterableOptions;
}
}