sabeel
/
dnet-hadoop
forked from D-Net/dnet-hadoop
1
0
Fork 0
Code Pull Requests Activity

WIP: pid cleaning

This commit is contained in:
Claudio Atzori 2023-06-09 16:47:25 +02:00
parent da7b66c542
commit 1d33074fd1
10 changed files with 573 additions and 346 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/DoiCleaningRule.java Normal file
View File

@ -0,0 +1,14 @@
package eu.dnetlib.dhp.schema.oaf.utils;
public class DoiCleaningRule {
public static String clean(final String doi) {
return doi
.toLowerCase()
.replaceAll("\\s", "")
.replaceAll("^doi:", "")
.replaceFirst(CleaningFunctions.DOI_PREFIX_REGEX, CleaningFunctions.DOI_PREFIX);
}
}

23
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/FundRefCleaningRule.java Normal file
View File

@ -0,0 +1,23 @@
package eu.dnetlib.dhp.schema.oaf.utils;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
public class FundRefCleaningRule {
public static String clean(final String fundrefId) {
String s = fundrefId
.toLowerCase()
.replaceAll("\\s", "");
Matcher m = Pattern.compile("\\d+").matcher(s);
if (m.matches()) {
return m.group();
} else {
return "";
}
}
}

89
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/GraphCleaningFunctions.java
View File

@ -162,19 +162,15 @@ public class GraphCleaningFunctions extends CleaningFunctions {
}
public static <T extends Oaf> T fixVocabularyNames(T value) {
if (value instanceof Datasource) {
// nothing to clean here
} else if (value instanceof Project) {
// nothing to clean here
} else if (value instanceof Organization) {
Organization o = (Organization) value;
if (Objects.nonNull(o.getCountry())) {
fixVocabName(o.getCountry(), ModelConstants.DNET_COUNTRY_TYPE);
}
} else if (value instanceof Relation) {
// nothing to clean here
} else if (value instanceof Result) {
if (value instanceof OafEntity) {
OafEntity e = (OafEntity) value;
Optional
.ofNullable(e.getPid())
.ifPresent(pid -> pid.forEach(p -> fixVocabName(p.getQualifier(), ModelConstants.DNET_PID_TYPES)));
if (value instanceof Result) {
Result r = (Result) value;
fixVocabName(r.getLanguage(), ModelConstants.DNET_LANGUAGES);
@ -188,6 +184,11 @@ public class GraphCleaningFunctions extends CleaningFunctions {
for (Instance i : r.getInstance()) {
fixVocabName(i.getAccessright(), ModelConstants.DNET_ACCESS_MODES);
fixVocabName(i.getRefereed(), ModelConstants.DNET_REVIEW_LEVELS);
Optional
.ofNullable(i.getPid())
.ifPresent(
pid -> pid.forEach(p -> fixVocabName(p.getQualifier(), ModelConstants.DNET_PID_TYPES)));
}
}
if (Objects.nonNull(r.getAuthor())) {
@ -208,6 +209,19 @@ public class GraphCleaningFunctions extends CleaningFunctions {
} else if (value instanceof Software) {
}
} else if (value instanceof Datasource) {
// nothing to clean here
} else if (value instanceof Project) {
// nothing to clean here
} else if (value instanceof Organization) {
Organization o = (Organization) value;
if (Objects.nonNull(o.getCountry())) {
fixVocabName(o.getCountry(), ModelConstants.DNET_COUNTRY_TYPE);
}
}
} else if (value instanceof Relation) {
// nothing to clean here
}
return value;
@ -260,6 +274,14 @@ public class GraphCleaningFunctions extends CleaningFunctions {
}
public static <T extends Oaf> T cleanup(T value, VocabularyGroup vocs) {
if (value instanceof OafEntity) {
OafEntity e = (OafEntity) value;
if (Objects.nonNull(e.getPid())) {
e.setPid(processPidCleaning(e.getPid()));
}
if (value instanceof Datasource) {
// nothing to clean here
} else if (value instanceof Project) {
@ -269,19 +291,7 @@ public class GraphCleaningFunctions extends CleaningFunctions {
if (Objects.isNull(o.getCountry()) || StringUtils.isBlank(o.getCountry().getClassid())) {
o.setCountry(ModelConstants.UNKNOWN_COUNTRY);
}
} else if (value instanceof Relation) {
Relation r = (Relation) value;
Optional<String> validationDate = doCleanDate(r.getValidationDate());
if (validationDate.isPresent()) {
r.setValidationDate(validationDate.get());
r.setValidated(true);
} else {
r.setValidationDate(null);
r.setValidated(false);
}
} else if (value instanceof Result) {
Result r = (Result) value;
if (Objects.nonNull(r.getDateofacceptance())) {
@ -394,9 +404,6 @@ public class GraphCleaningFunctions extends CleaningFunctions {
.map(GraphCleaningFunctions::cleanValue)
.collect(Collectors.toList()));
}
if (Objects.nonNull(r.getPid())) {
r.setPid(processPidCleaning(r.getPid()));
}
if (Objects.isNull(r.getResourcetype()) || StringUtils.isBlank(r.getResourcetype().getClassid())) {
r
.setResourcetype(
@ -405,13 +412,15 @@ public class GraphCleaningFunctions extends CleaningFunctions {
if (Objects.nonNull(r.getInstance())) {
for (Instance i : r.getInstance()) {
if (!vocs.termExists(ModelConstants.DNET_PUBLICATION_RESOURCE, i.getInstancetype().getClassid())) {
if (!vocs
.termExists(ModelConstants.DNET_PUBLICATION_RESOURCE, i.getInstancetype().getClassid())) {
if (r instanceof Publication) {
i
.setInstancetype(
OafMapperUtils
.qualifier(
"0038", "Other literature type", ModelConstants.DNET_PUBLICATION_RESOURCE,
"0038", "Other literature type",
ModelConstants.DNET_PUBLICATION_RESOURCE,
ModelConstants.DNET_PUBLICATION_RESOURCE));
} else if (r instanceof Dataset) {
i
@ -455,7 +464,8 @@ public class GraphCleaningFunctions extends CleaningFunctions {
});
});
if (Objects.isNull(i.getAccessright()) || StringUtils.isBlank(i.getAccessright().getClassid())) {
if (Objects.isNull(i.getAccessright())
|| StringUtils.isBlank(i.getAccessright().getClassid())) {
i
.setAccessright(
accessRight(
@ -478,7 +488,8 @@ public class GraphCleaningFunctions extends CleaningFunctions {
}
}
}
if (Objects.isNull(r.getBestaccessright()) || StringUtils.isBlank(r.getBestaccessright().getClassid())) {
if (Objects.isNull(r.getBestaccessright())
|| StringUtils.isBlank(r.getBestaccessright().getClassid())) {
Qualifier bestaccessrights = OafMapperUtils.createBestAccessRights(r.getInstance());
if (Objects.isNull(bestaccessrights)) {
r
@ -574,6 +585,20 @@ public class GraphCleaningFunctions extends CleaningFunctions {
} else if (value instanceof Software) {
}
}
} else if (value instanceof Relation) {
Relation r = (Relation) value;
Optional<String> validationDate = doCleanDate(r.getValidationDate());
if (validationDate.isPresent()) {
r.setValidationDate(validationDate.get());
r.setValidated(true);
} else {
r.setValidationDate(null);
r.setValidated(false);
}
}
return value;
@ -628,7 +653,7 @@ public class GraphCleaningFunctions extends CleaningFunctions {
.filter(sp -> !PID_BLACKLIST.contains(sp.getValue().trim().toLowerCase()))
.filter(sp -> Objects.nonNull(sp.getQualifier()))
.filter(sp -> StringUtils.isNotBlank(sp.getQualifier().getClassid()))
.map(CleaningFunctions::normalizePidValue)
.map(PidCleaner::normalizePidValue)
.filter(CleaningFunctions::pidFilter)
.collect(Collectors.toList());
}

18
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/GridCleaningRule.java Normal file
View File

@ -0,0 +1,18 @@
package eu.dnetlib.dhp.schema.oaf.utils;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
public class GridCleaningRule {
public static String clean(String grid) {
String s = grid
.replaceAll("\\s", "")
.toLowerCase();
Matcher m = Pattern.compile("\\d{4,6}\\.[0-9a-z]{1,2}").matcher(s);
return m.matches() ? "grid." + m.group() : "";
}
}

19
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/ISNICleaningRule.java Normal file
View File

@ -0,0 +1,19 @@
package eu.dnetlib.dhp.schema.oaf.utils;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
// https://www.wikidata.org/wiki/Property:P213
public class ISNICleaningRule {
public static String clean(final String isni) {
Matcher m = Pattern.compile("([0]{4}) ?([0-9]{4}) ?([0-9]{4}) ?([0-9]{3}[0-9X])").matcher(isni);
if (m.matches()) {
return String.join("", m.group(1), m.group(2), m.group(3), m.group(4));
} else {
return "";
}
}
}

19
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/PICCleaningRule.java Normal file
View File

@ -0,0 +1,19 @@
package eu.dnetlib.dhp.schema.oaf.utils;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
public class PICCleaningRule {
public static String clean(final String pic) {
Matcher m = Pattern.compile("\\d{9}").matcher(pic);
if (m.matches()) {
return m.group();
} else {
return "";
}
}
}

62
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/PidCleaner.java Normal file
View File

@ -0,0 +1,62 @@
package eu.dnetlib.dhp.schema.oaf.utils;
import java.util.Optional;
import eu.dnetlib.dhp.schema.oaf.StructuredProperty;
public class PidCleaner {
/**
* Utility method that normalises PID values on a per-type basis.
* @param pid the PID whose value will be normalised.
* @return the PID containing the normalised value.
*/
public static StructuredProperty normalizePidValue(StructuredProperty pid) {
pid
.setValue(
normalizePidValue(
pid.getQualifier().getClassid(),
pid.getValue()));
return pid;
}
public static String normalizePidValue(String pidType, String pidValue) {
String value = Optional
.ofNullable(pidValue)
.map(String::trim)
.orElseThrow(() -> new IllegalArgumentException("PID value cannot be empty"));
switch (pidType) {
// TODO add cleaning for more PID types as needed
// Result
case "doi":
return DoiCleaningRule.clean(value);
case "pmid":
return PmidCleaningRule.clean(value);
case "pmc":
return PmcCleaningRule.clean(value);
case "handle":
case "arXiv":
return value;
// Organization
case "GRID":
return GridCleaningRule.clean(value);
case "ISNI":
return ISNICleaningRule.clean(value);
case "ROR":
return RorCleaningRule.clean(value);
case "PIC":
return PICCleaningRule.clean(value);
case "FundRef":
return FundRefCleaningRule.clean(value);
default:
return value;
}
}
}

13
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/PmcCleaningRule.java Normal file
View File

@ -0,0 +1,13 @@
package eu.dnetlib.dhp.schema.oaf.utils;
public class PmcCleaningRule {
public static String clean(String pmc) {
String s = pmc
.replaceAll("\\s", "")
.toUpperCase();
return s.matches("^PMC\\d{1,8}$") ? s : "";
}
}

16
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/PmidCleaningRule.java Normal file
View File

@ -0,0 +1,16 @@
package eu.dnetlib.dhp.schema.oaf.utils;
// https://researchguides.stevens.edu/c.php?g=442331&p=6577176
public class PmidCleaningRule {
public static String clean(String pmid) {
String s = pmid
.toLowerCase()
.replaceAll("\\s", "")
.trim()
.replaceAll("^0+", "");
return s.matches("^\\d{1,8}$") ? s : "";
}
}

18
dhp-common/src/main/java/eu/dnetlib/dhp/schema/oaf/utils/RorCleaningRule.java Normal file
View File

@ -0,0 +1,18 @@
package eu.dnetlib.dhp.schema.oaf.utils;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
// https://ror.readme.io/docs/ror-identifier-pattern
public class RorCleaningRule {
public static String clean(String ror) {
String s = ror
.replaceAll("\\s", "")
.toLowerCase();
Matcher m = Pattern.compile("0[a-hj-km-np-tv-z|0-9]{6}[0-9]{2}").matcher(s);
return m.matches() ? "https://ror.org/" + m.group() : "";
}
}