Revision 59094
Added by Alessia Bardi over 4 years ago
| modules/dnet-modular-collector-service/trunk/src/main/java/eu/dnetlib/data/collector/plugins/filesystem/FilesystemCollectorPlugin.java | ||
|---|---|---|
| 3 | 3 |
import eu.dnetlib.data.collector.plugin.AbstractCollectorPlugin; |
| 4 | 4 |
import eu.dnetlib.data.collector.rmi.CollectorServiceException; |
| 5 | 5 |
import eu.dnetlib.data.collector.rmi.InterfaceDescriptor; |
| 6 |
import org.apache.commons.lang3.StringUtils; |
|
| 6 | 7 |
|
| 7 | 8 |
/** |
| 8 | 9 |
* |
| ... | ... | |
| 17 | 18 |
|
| 18 | 19 |
final String baseUrl = interfaceDescriptor.getBaseUrl(); |
| 19 | 20 |
if (baseUrl == null || baseUrl.isEmpty()) { throw new CollectorServiceException("Param 'baseurl' is null or empty"); }
|
| 20 |
return new FilesystemIterable(interfaceDescriptor); |
|
| 21 |
|
|
| 22 |
if (StringUtils.isNotBlank(fromDate) && !fromDate.matches("\\d{4}-\\d{2}-\\d{2}")) { throw new CollectorServiceException("Invalid date (YYYY-MM-DD): " + fromDate); }
|
|
| 23 |
return new FilesystemIterable(interfaceDescriptor, fromDate); |
|
| 21 | 24 |
} |
| 22 | 25 |
|
| 23 | 26 |
} |
| modules/dnet-modular-collector-service/trunk/src/main/java/eu/dnetlib/data/collector/plugins/filesystem/FilesystemIterable.java | ||
|---|---|---|
| 51 | 51 |
|
| 52 | 52 |
private boolean setObjIdentifierFromFileName = false; |
| 53 | 53 |
|
| 54 |
private String fromDate; |
|
| 55 |
|
|
| 54 | 56 |
/** |
| 55 | 57 |
* Instantiates a new filesystem iterable. |
| 56 | 58 |
* |
| 57 | 59 |
* @param descriptor the descriptor |
| 58 | 60 |
* @throws CollectorServiceException the collector service exception |
| 59 | 61 |
*/ |
| 60 |
public FilesystemIterable(final InterfaceDescriptor descriptor) throws CollectorServiceException {
|
|
| 62 |
public FilesystemIterable(final InterfaceDescriptor descriptor, final String fromDate) throws CollectorServiceException {
|
|
| 61 | 63 |
try {
|
| 62 | 64 |
final String baseUrl = descriptor.getBaseUrl(); |
| 63 | 65 |
URL basePath = new URL(baseUrl); |
| ... | ... | |
| 71 | 73 |
if (descriptor.getParams().containsKey("setObjIdentifierFromFileName")) {
|
| 72 | 74 |
setObjIdentifierFromFileName = Boolean.parseBoolean(descriptor.getParams().get("setObjIdentifierFromFileName"));
|
| 73 | 75 |
} |
| 76 |
this.fromDate = fromDate; |
|
| 74 | 77 |
} catch (MalformedURLException e) {
|
| 75 | 78 |
throw new CollectorServiceException("Filesystem collector failed! ", e);
|
| 76 | 79 |
} |
| ... | ... | |
| 83 | 86 |
*/ |
| 84 | 87 |
@Override |
| 85 | 88 |
public Iterator<String> iterator() {
|
| 86 |
final FileSystemIterator fsi = new FileSystemIterator(baseDir.getAbsolutePath(), extensions); |
|
| 89 |
final FileSystemIterator fsi = new FileSystemIterator(baseDir.getAbsolutePath(), extensions, fromDate);
|
|
| 87 | 90 |
return Iterators.transform(fsi, inputFileName -> {
|
| 88 | 91 |
FileInputStream fileInputStream = null; |
| 89 | 92 |
try {
|
| modules/dnet-modular-collector-service/trunk/src/main/java/eu/dnetlib/data/collector/plugins/filesystem/FileSystemIterator.java | ||
|---|---|---|
| 4 | 4 |
import java.nio.file.Files; |
| 5 | 5 |
import java.nio.file.Path; |
| 6 | 6 |
import java.nio.file.Paths; |
| 7 |
import java.nio.file.attribute.FileTime; |
|
| 8 |
import java.time.*; |
|
| 9 |
import java.time.format.DateTimeFormatter; |
|
| 10 |
import java.time.temporal.TemporalAccessor; |
|
| 7 | 11 |
import java.util.Iterator; |
| 8 | 12 |
import java.util.Set; |
| 9 | 13 |
|
| 14 |
import eu.dnetlib.data.collector.rmi.CollectorServiceException; |
|
| 15 |
import eu.dnetlib.data.collector.rmi.CollectorServiceRuntimeException; |
|
| 10 | 16 |
import org.apache.commons.io.FilenameUtils; |
| 11 | 17 |
import org.apache.commons.lang3.StringUtils; |
| 12 | 18 |
import org.apache.commons.logging.Log; |
| ... | ... | |
| 30 | 36 |
private Iterator<Path> pathIterator; |
| 31 | 37 |
private String current; |
| 32 | 38 |
|
| 39 |
private boolean incremental = false; |
|
| 40 |
private LocalDate fromDate = null; |
|
| 41 |
private final DateTimeFormatter simpleDateTimeFormatter = DateTimeFormatter |
|
| 42 |
.ofPattern("yyyy-MM-dd")
|
|
| 43 |
.withZone(ZoneId.systemDefault()); |
|
| 44 |
|
|
| 33 | 45 |
public FileSystemIterator(final String baseDir, final String extensions) {
|
| 46 |
this(baseDir,extensions, null); |
|
| 47 |
} |
|
| 48 |
|
|
| 49 |
public FileSystemIterator(final String baseDir, final String extensions, final String fromDate) {
|
|
| 34 | 50 |
if(StringUtils.isNotBlank(extensions)) {
|
| 35 | 51 |
this.extensions = Sets.newHashSet(extensions.split(","));
|
| 36 | 52 |
} |
| 53 |
this.incremental = StringUtils.isNotBlank(fromDate); |
|
| 54 |
if (incremental) {
|
|
| 55 |
//I expect fromDate in the format 'yyyy-MM-dd'. See class eu.dnetlib.msro.workflows.nodes.collect.FindDateRangeForIncrementalHarvestingJobNode . |
|
| 56 |
this.fromDate = LocalDate.parse(fromDate, simpleDateTimeFormatter); |
|
| 57 |
log.debug("fromDate string: " + fromDate + " -- parsed: " + this.fromDate.toString());
|
|
| 58 |
} |
|
| 37 | 59 |
try {
|
| 38 | 60 |
this.pathIterator = Files.newDirectoryStream(Paths.get(baseDir)).iterator(); |
| 39 | 61 |
this.current = walkTillNext(); |
| 40 |
} catch (IOException e) {
|
|
| 62 |
} catch (Exception e) {
|
|
| 41 | 63 |
log.error("Cannot initialize File System Iterator. Is this path correct? " + baseDir);
|
| 42 |
throw new RuntimeException("Filesystem collection error.", e);
|
|
| 64 |
e.printStackTrace(); |
|
| 65 |
throw new CollectorServiceRuntimeException("Filesystem collection error.", e);
|
|
| 43 | 66 |
} |
| 67 |
|
|
| 44 | 68 |
} |
| 45 | 69 |
|
| 46 | 70 |
@Override |
| ... | ... | |
| 79 | 103 |
} |
| 80 | 104 |
} else {
|
| 81 | 105 |
if (extensions.isEmpty() || extensions.contains(FilenameUtils.getExtension(nextFilePath.toString()))) {
|
| 82 |
log.debug("Returning: " + nextFilePath.toString());
|
|
| 83 |
return nextFilePath.toString(); |
|
| 106 |
if(incremental){
|
|
| 107 |
try {
|
|
| 108 |
final FileTime lastModifiedTime = Files.getLastModifiedTime(nextFilePath); |
|
| 109 |
if(lastModifiedTime.toInstant().isAfter(fromDate.atStartOfDay().toInstant(ZoneOffset.UTC))){
|
|
| 110 |
log.debug("Returning: " + nextFilePath.toString());
|
|
| 111 |
return nextFilePath.toString(); |
|
| 112 |
} |
|
| 113 |
else {
|
|
| 114 |
log.debug("File "+nextFilePath.toString()+" has not changed.");
|
|
| 115 |
} |
|
| 116 |
} catch (Exception e) {
|
|
| 117 |
throw new CollectorServiceRuntimeException(e); |
|
| 118 |
} |
|
| 119 |
} |
|
| 120 |
else {
|
|
| 121 |
log.debug("Returning: " + nextFilePath.toString());
|
|
| 122 |
return nextFilePath.toString(); |
|
| 123 |
} |
|
| 84 | 124 |
} |
| 85 | 125 |
} |
| 86 | 126 |
} |
Also available in: Unified diff
supporting incremental harvesting with the fs collector plugin