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
|
}
|