|
1
|
package eu.dnetlib.data.collector.plugins;
|
|
2
|
|
|
3
|
import java.io.*;
|
|
4
|
import java.net.URL;
|
|
5
|
import java.util.Iterator;
|
|
6
|
import java.util.Set;
|
|
7
|
|
|
8
|
import com.google.common.collect.Iterators;
|
|
9
|
import eu.dnetlib.data.collector.plugin.AbstractCollectorPlugin;
|
|
10
|
import eu.dnetlib.data.collector.plugins.oai.engine.XmlCleaner;
|
|
11
|
import eu.dnetlib.data.collector.rmi.CollectorServiceException;
|
|
12
|
import eu.dnetlib.data.collector.rmi.InterfaceDescriptor;
|
|
13
|
import org.apache.commons.csv.CSVFormat;
|
|
14
|
import org.apache.commons.csv.CSVParser;
|
|
15
|
import org.apache.commons.io.input.BOMInputStream;
|
|
16
|
import org.apache.commons.lang3.StringUtils;
|
|
17
|
import org.apache.commons.logging.Log;
|
|
18
|
import org.apache.commons.logging.LogFactory;
|
|
19
|
import org.dom4j.Document;
|
|
20
|
import org.dom4j.DocumentHelper;
|
|
21
|
import org.dom4j.Element;
|
|
22
|
|
|
23
|
/**
|
|
24
|
* The Class HttpCSVCollectorPlugin.
|
|
25
|
*/
|
|
26
|
public class HttpCSVCollectorPlugin extends AbstractCollectorPlugin {
|
|
27
|
|
|
28
|
private static final Log log = LogFactory.getLog(HttpCSVCollectorPlugin.class);
|
|
29
|
|
|
30
|
public static final String UTF8_BOM = "\uFEFF";
|
|
31
|
|
|
32
|
/**
|
|
33
|
* The Class HTTPCSVIterator.
|
|
34
|
*/
|
|
35
|
class HTTPCSVIterator implements Iterable<String> {
|
|
36
|
|
|
37
|
/** The descriptor. */
|
|
38
|
private InterfaceDescriptor descriptor;
|
|
39
|
|
|
40
|
/**
|
|
41
|
* Instantiates a new HTTPCSV iterator.
|
|
42
|
*
|
|
43
|
* @param descriptor
|
|
44
|
* the descriptor
|
|
45
|
*/
|
|
46
|
public HTTPCSVIterator(final InterfaceDescriptor descriptor) {
|
|
47
|
this.descriptor = descriptor;
|
|
48
|
}
|
|
49
|
|
|
50
|
/**
|
|
51
|
* Iterator.
|
|
52
|
*
|
|
53
|
* @return the iterator
|
|
54
|
*/
|
|
55
|
@SuppressWarnings("resource")
|
|
56
|
@Override
|
|
57
|
public Iterator<String> iterator() {
|
|
58
|
|
|
59
|
try {
|
|
60
|
final String separator = descriptor.getParams().get("separator");
|
|
61
|
final String identifier = descriptor.getParams().get("identifier");
|
|
62
|
final String quote = descriptor.getParams().get("quote");
|
|
63
|
final URL url = new URL(descriptor.getBaseUrl());
|
|
64
|
long nLines = 0;
|
|
65
|
|
|
66
|
// FIX
|
|
67
|
// This code should skip the lines with invalid quotes
|
|
68
|
final File tempFile = File.createTempFile("csv-", ".tmp");
|
|
69
|
try (InputStream is = url.openConnection().getInputStream();
|
|
70
|
BOMInputStream bomIs = new BOMInputStream(is);
|
|
71
|
BufferedReader reader = new BufferedReader(new InputStreamReader(bomIs));
|
|
72
|
FileWriter fw = new FileWriter(tempFile)) {
|
|
73
|
|
|
74
|
String line;
|
|
75
|
while ((line = reader.readLine()) != null) {
|
|
76
|
if (StringUtils.isBlank(quote) || (quote.charAt(0) != '"') || verifyQuotes(line, separator.charAt(0))) {
|
|
77
|
fw.write(line);
|
|
78
|
fw.write("\n");
|
|
79
|
nLines++;
|
|
80
|
}
|
|
81
|
}
|
|
82
|
}
|
|
83
|
// END FIX
|
|
84
|
|
|
85
|
final CSVFormat format = CSVFormat.EXCEL
|
|
86
|
.withHeader()
|
|
87
|
.withDelimiter(separator.equals("\\t") || StringUtils.isBlank(separator) ? '\t' : separator.charAt(0))
|
|
88
|
.withQuote(StringUtils.isBlank(quote) ? null : quote.charAt(0))
|
|
89
|
.withTrim();
|
|
90
|
|
|
91
|
final CSVParser parser = new CSVParser(new FileReader(tempFile), format);
|
|
92
|
final Set<String> headers = parser.getHeaderMap().keySet();
|
|
93
|
|
|
94
|
final long nRecords = nLines - 1;
|
|
95
|
|
|
96
|
return Iterators.transform(parser.iterator(), input -> {
|
|
97
|
try {
|
|
98
|
final Document document = DocumentHelper.createDocument();
|
|
99
|
final Element root = document.addElement("csvRecord");
|
|
100
|
for (final String key : headers) {
|
|
101
|
final Element row = root.addElement("column");
|
|
102
|
final String value = XmlCleaner.cleanAllEntities(input.get(key));
|
|
103
|
if (value!= null) row.addAttribute("name", key).addText(value);
|
|
104
|
if (key.equals(identifier)) {
|
|
105
|
row.addAttribute("isID", "true");
|
|
106
|
}
|
|
107
|
}
|
|
108
|
|
|
109
|
return document.asXML();
|
|
110
|
} finally {
|
|
111
|
log.debug(tempFile.getAbsolutePath());
|
|
112
|
if (parser.getRecordNumber() == nRecords) {
|
|
113
|
log.debug("DELETING " + tempFile.getAbsolutePath());
|
|
114
|
tempFile.delete();
|
|
115
|
}
|
|
116
|
}
|
|
117
|
});
|
|
118
|
} catch (final Exception e) {
|
|
119
|
log.error("Error iterating csv lines", e);
|
|
120
|
return null;
|
|
121
|
}
|
|
122
|
}
|
|
123
|
|
|
124
|
}
|
|
125
|
|
|
126
|
/*
|
|
127
|
* (non-Javadoc)
|
|
128
|
*
|
|
129
|
* @see eu.dnetlib.data.collector.plugin.CollectorPlugin#collect(eu.dnetlib.data.collector.rmi.InterfaceDescriptor, java.lang.String,
|
|
130
|
* java.lang.String)
|
|
131
|
*/
|
|
132
|
@Override
|
|
133
|
public Iterable<String> collect(final InterfaceDescriptor descriptor, final String fromDate, final String untilDate) throws CollectorServiceException {
|
|
134
|
|
|
135
|
return new HTTPCSVIterator(descriptor);
|
|
136
|
}
|
|
137
|
|
|
138
|
public boolean verifyQuotes(final String line, final char separator) {
|
|
139
|
final char[] cs = line.trim().toCharArray();
|
|
140
|
boolean inField = false;
|
|
141
|
boolean skipNext = false;
|
|
142
|
for (int i = 0; i < cs.length; i++) {
|
|
143
|
if (skipNext) {
|
|
144
|
skipNext = false;
|
|
145
|
} else if (inField) {
|
|
146
|
if ((cs[i] == '\"') && ((i == (cs.length - 1)) || (cs[i + 1] == separator))) {
|
|
147
|
inField = false;
|
|
148
|
} else if ((cs[i] == '\"') && (i < (cs.length - 1))) {
|
|
149
|
if ((cs[i + 1] == '\"')) {
|
|
150
|
skipNext = true;
|
|
151
|
} else {
|
|
152
|
log.warn("Skipped invalid line: " + line);
|
|
153
|
return false;
|
|
154
|
}
|
|
155
|
}
|
|
156
|
} else {
|
|
157
|
if ((cs[i] == '\"') && ((i == 0) || (cs[i - 1] == separator))) {
|
|
158
|
inField = true;
|
|
159
|
}
|
|
160
|
}
|
|
161
|
}
|
|
162
|
|
|
163
|
if (inField) {
|
|
164
|
log.warn("Skipped invalid line: " + line);
|
|
165
|
return false;
|
|
166
|
}
|
|
167
|
|
|
168
|
return true;
|
|
169
|
}
|
|
170
|
|
|
171
|
}
|