General Framework

Introduction

This documentation explains the features and the usage of the importer framework.
Enabling the framework can be achieved by uncommenting the following step in item-submission.xml.
Implementation specific or additional configuration can be found in their related documentation, if any.
Please refer to subdivisions of this documentation for specific implementations of the framework. 

<step>
   <heading>submit.progressbar.lookup</heading>
   <processing-class>org.dspace.submit.step.XMLUIStartSubmissionLookupStep</processing-class>
   <xmlui-binding>org.dspace.app.xmlui.aspect.submission.submit.StartSubmissionLookupStep</xmlui-binding>
   <workflow-editable>true</workflow-editable>
</step>

Features

• lookup publications from remote sources
• Support for multiple implementations

Abstraction of input format

The importer framework does not enforce a specific input format. Each importer implementation defines which input format it expects from a remote source. The import framework uses generics to achieve this. Each importer implementation will have a type set of the record type it receives from the remote source's response. This type set will also be used by the framework to use the correct MetadataFieldMapping for a certain implementation. Read Implementation of an import source for more information and how to enable the framework.

Transformation to DSpace item

The framework produces an 'ImportRecord' that is completely decoupled from DSPace. It contains a set of metadata DTO's that contain the notion of schema,element and qualifier. The specific implementation is responsible for populating this set. It is then very simple to create a DSPace item from this list.

Relation with BTE

While there is some overlap between this framework and BTE, this framework supports some features that are hard to implement using the BTE. It has explicit support to deal with network failure and throttling imposed by the data source. It also has explicit support for distinguishing between network caused errors and invalid requests to the source. Furthermore the framework doesn't impose any restrictions on the format in which the data is retrieved. It uses java generics to support different source record types. A reference implementation of using XML records is provided for which a set of metadata can be generated from any xpath expression (or composite of xpath expressions). Unless 'advanced' processing is necessary (e.g. lookup of authors in an LDAP directory) this metadata mapping can be simply configured using spring. No code changes necessary. A mixture of advanced and simple (xpath) mapping is also possible.

This design is also in line with the roadmap to create a Modular Framework as detailed in https://wiki.duraspace.org/display/DSPACE/Design+-+Module+Framework+and+Registry This modular design also allows it to be completely independent of the user interface layer, be it JSPUI, XMLUI, command line or the result of the new UI projects: https://wiki.duraspace.org/display/DSPACE/Design+-+Single+UI+Project

Implementation of an import source

Each importer implementation must at least implement interface org.dspace.importer.external.service.components.MetadataSource and implement the inherited methods.

One can also choose to implement class org.dspace.importer.external.service.components.AbstractRemoteMetadataSource next to the MetadataSource interface. This class contains functionality to handle request timeouts and to retry requests.

A third option is to implement class org.dspace.importer.external.service.AbstractImportSourceService. This class already implements both the MetadataSource interface and Source class. AbstractImportSourceService has a generic type set 'RecordType'. In the importer implementation this type set should be the class of the records received from the remote source's response (e.g. when using axiom to get the records from the remote source's XML response, the importer implementation's type set isorg.apache.axiom.om.OMElement).

Implementing the AbstractImportSourceService allows the importer implementation to use the framework's build-in support to transform a record received from the remote source to an object of class *org.dspace.importer.external.datamodel.ImportRecord containing DSpace metadata fields, as explained here: Metadata mapping.

Inherited methods

Method getImportSource() should return a unique identifier. Importer implementations should not be called directly, but class org.dspace.importer.external.service.ImportService should be called instead. This class contains the same methods as the importer implementations, but with an extra parameter 'url'. This url parameter should contain the same identifier that is returned by the getImportSource() method of the importer implementation you want to use.

framework is used by both the REST API and User Interface to help enhance or enrich submissions.  One examples usage is in Importing Items via basic bibliographic formats (Endnote, BibTex, RIS, CSV, etc) and online services (arXiv, PubMed, CrossRef, CiNii, etc)

Features

• lookup publications from remote sources
• Support for multiple implementations

Abstraction of input format

The importer framework does not enforce a specific input format. Each importer implementation defines which input format it expects from a remote source. The import framework uses generics to achieve this. Each importer implementation will have a type set of the record type it receives from the remote source's response. This type set will also be used by the framework to use the correct MetadataFieldMapping for a certain implementation. Read "Implementation of an import source" below for more information and how to enable the framework.

Editing Metadata Mapping

At a simple level, metadata mapping configurations are all in Spring configs in [dspace.dir]/config/spring/api/

In that directory, you'll find a mapping file per import source, e.g. "arxiv-integration.xml", "bibtex-integration.xml", "endnote-integration.xml", "pubmed-integration.xml", etc.

There are two different mapping types.

1. First, mapping from a file-based import (e.g. bibtex, endnote, ris, etc) to a DSpace metadata field.

   1. The list of all of the enabled mappings can be found in a "MetadataFieldConfig" <util:map>, usually at the top of the config file.

      Code Block

      <util:map id="bibtexMetadataFieldMap" key-type="org.dspace.importer.external.metadatamapping.MetadataFieldConfig" value-type="org.dspace.importer.external.metadatamapping.contributor.MetadataContributor"> <description>Defines which metadatum is mapped on which metadatum. Note that while the key must be unique it only matters here for postprocessing of the value. The mapped MetadatumContributor has full control over what metadatafield is generated. </description> <!-- These entry tags are the enabled mappings. The "value-ref" must map to a <bean> ID. --> <entry key-ref="dcTitle" value-ref="bibtexTitleContrib" /> <entry key-ref="dcAuthors" value-ref="bibtexAuthorsContrib" /> <entry key-ref="dcJournal" value-ref="bibtexJournalContrib" /> <entry key-ref="dcIssued" value-ref="bibtexIssuedContrib" /> <entry key-ref="dcJissn" value-ref="bibtexJissnContrib" /> </util:map>

      
      

   2. Each field in the file is mapped to a DSpace metadata field in a "SimpleMetadataContributor" bean definition.  NOTE: a large number of DSpace defined metadata fields are already configured as MetadataFieldConfig beans in the "dublincore-metadata-mapper.xml" Spring Config in the same directory. These may be reused in other configurations.

      Code Block

      <!-- This example bean for BibTex says the "title" key in the BibTex" file should be mapped to the DSpace metadata field defined in the "dcTitle" bean. This "dcTitle" bean is found in "dublincore-metadata-mapper.xml" and obviously maps to "dc.title" --> <bean id="bibtexTitleContrib" class="org.dspace.importer.external.metadatamapping.contributor.SimpleMetadataContributor"> <property name="field" ref="dcTitle"/> <property name="key" value="title" /> </bean>

      
      

2. Second, mapping from an external API query import (e.g. arxiv, pubmed, etc) to a DSpace metadata field.

   1. Similar to above, The list of all of the enabled mappings can be found in a "MetadataFieldConfig" <util:map>, usually at the top of the config file.

      Code Block

      <util:map id="arxivMetadataFieldMap" key-type="org.dspace.importer.external.metadatamapping.MetadataFieldConfig" value-type="org.dspace.importer.external.metadatamapping.contributor.MetadataContributor"> <description>Defines which metadatum is mapped on which metadatum. Note that while the key must be unique it only matters here for postprocessing of the value. The mapped MetadatumContributor has full control over what metadatafield is generated. </description> <!-- These entry tags are the enabled mappings. The "value-ref" must map to a <bean> ID. --> <entry key-ref="arxiv.title" value-ref="arxivTitleContrib"/> <entry key-ref="arxiv.summary" value-ref="arxivSummaryContrib"/> <entry key-ref="arxiv.published" value-ref="arxivPublishedContrib"/> <entry key-ref="arxiv.arxiv.doi" value-ref="arxivDoiContrib"/> <entry key-ref="arxiv.arxiv.journal_ref" value-ref="arxivJournalContrib"/> <entry key-ref="arxiv.category.term" value-ref="arxivCategoryTermContrib"/> <entry key-ref="arxiv.author.name" value-ref="arxivAuthorContrib"/> <entry key-ref="arxiv.identifier.other" value-ref="arxivOtherContrib"/> </util:map>

      
      

   2. Each field in the file is mapped to a DSpace metadata field, usually in a "SimpleXPathMetadatumContributor" bean definition which also uses a "MetadataFieldConfig" bean.  NOTE: a large number of DSpace defined metadata fields are already configured as MetadataFieldConfig beans in the "dublincore-metadata-mapper.xml" Spring Config in the same directory. These may be reused in other configurations.

      Code Block

      <!-- This first bean define an XPath query ("ns:title") to map to a field (ID="arxiv.title") in DSpace --> <bean id="arxivTitleContrib" class="org.dspace.importer.external.metadatamapping.contributor.SimpleXpathMetadatumContributor"> <property name="field" ref="arxiv.title"/> <property name="query" value="ns:title"/> <property name="prefixToNamespaceMapping" ref="arxivBasePrefixToNamespaceMapping"/> </bean> <!-- This second bean then defines which DSpace field to use when "arxiv.title" is references. In other words, between these two beans, the "ns:title" XPath query value is saved to "dc.title". --> <bean id="arxiv.title" class="org.dspace.importer.external.metadatamapping.MetadataFieldConfig"> <constructor-arg value="dc.title"/> </bean>

      
      

Transformation to DSpace Item

The framework produces an 'ImportRecord' that is completely decoupled from DSpace. It contains a set of metadata DTO's that contain the notion of schema,element and qualifier. The specific implementation is responsible for populating this set. It is then very simple to create a DSpace item from this list.

Implementation of an import source for External Sources

Each external source/API importer implementation must at least implements org.dspace.importer.external.service.components.QuerySource, which provides the query method used by the framework to retrieve data from the remote source (e.g. Pubmed, ArXiv, etc). Each external source importer must implements, according to the provider APIs, the declared methods.
An useful abstract for remote sources is org.dspace.importer.external.service.components.AbstractRemoteMetadataSource.This class contains functionality to handle request timeout and to retry requests. Using this abstract, the query method must implements java.util.concurrent.Callable.

Implementation of an import source for files

Each file importer implementation must at least implements org.dspace.importer.external.service.components.FileSource, which provides the basic methods used by the framework to parse and load data from the file (e.g. CSV, Endnote, etc).

Each importer must implements the method:

public List<ImportRecord> getRecords(InputStream inputStream) throws FileSourceException;

This method is responsible to transform the input data into an ImportRecord list, which will then managed by the top layer of the framework.

The conversion from raw data to an ImportRecord could be done using the framework too, using the metadata mapping structure (see below).

File sources needs to know which file extensions they have to supports. This is done by the default method isValidSourceForFile in FileSource, and is controlled by the entries in the list returned by declared method public List<String> getSupportedExtensions();

An useful abstract for file source is org.dspace.importer.external.service.components.AbstractPlainMetadataSource. It should be used whenever it is possible to model the data in the file as a list of key-value lists (e.g. for CSV files, any row is a key value list).

Mapping raw data to Metadata

The framework core is a mid-layer component which allow the conversion of raw data into metadata (ImportRecord) using xml configurable spring beans.

The core of this approach is org.dspace.importer.externa.service.AbstractImportMetadataSourceService. Any service that wants to generate metadata from raw data should go through this abstract.

Our service then should extends AbstractImportMetadataSourceService, and use transformSourceRecords to transform raw data into ImportRecords.

The most relevant concept in the framework is private MetadataFieldMapping<RecordType, MetadataContributor<RecordType>> metadataFieldMapping. This is the place where the framework take the mapping between row data and the associated metadatum. This map must be injected in the service, and will be used by transformSourceRecords to convert the data.

RecordType is a generic type, which rapresent a single entry of the list of data, and will be mapped to a single ImportRecord. Any metadatum
will be mapped to a specific field in the RecordType using a Contributor as described in Metadata mapping.

Inherited methods

Spring configuration for External Sources

In order to make the live import providers available, them must be mapped as spring beans into dspace-api/src/main/resources/spring/spring-dspace-
addon-import-services.xml.

This is an example of a provider which allow to import both files and remote source.

<bean id="PubmedImportService"
      class="org.dspace.importer.external.pubmed.service.PubmedImportMetadataSourceServiceImpl" scope="singleton">
    <property name="metadataFieldMapping" ref="PubmedMetadataFieldMapping"/>
    <property name="supportedExtensions">
        <list>
            <value>xml</value>
        </list>
    </property>
    ...
</bean>

Here is defined the service responsible to fetch and transform the data PubmedImportMetadataSourceServiceImpl, which is an extension
of AbstractImportMetadataSourceService as described above.

The field metadataFieldMapping is an instance of Map<MetadataFieldConfig,MetadataContributor> and contains the effective
mapping.

supportedExtensions is the file extension this provider supports.

To expose this provider as Live Import provider, we need to construct a bean of type org.dspace.external.provider.impl.LiveImportDataProvider in the following way

<bean id="pubmedLiveImportDataProvider" class="org.dspace.external.provider.impl.LiveImportDataProvider">
    <property name="metadataSource" ref="PubmedImportService"/>
    <property name="sourceIdentifier" value="pubmed"/>
    <property name="recordIdMetadata" value="dc.identifier.other"/>
</bean>

where metadataSource is the bean referencing to live import service as described in “Metadata mapping”, sourceIdentifier the name of
the provider in the live import framework and recordIdMetadata the metadatum used as id of the ImportRecord

...

.

Metadata mapping

When using an implementation of AbstractImportSourceService, a mapping of remote record fields to DSpace metadata fields can be created.

first create an implementation of class AbstractMetadataFieldMapping with the same type set used for the importer implementation.

...

    <bean id="dc.title" class="org.dspace.importer.external.metadatamapping.MetadataFieldConfig">
        <constructor-arg value="dc.title"/>
    </bean></bean>

NOTE: A large number of these MetadataFieldConfig definitions are already provided out-of-the-box in  [dspace.dir]/config/spring/api/dublincore-metadata-mapper.xml  This allows most service-specific Spring configurations to just reuse those existing MetadataFieldConfig definitions 

Now this metadata field can be used to create a mapping. To add a mapping for the "dc.title" field declared above, a new spring bean configuration of a class class org.dspace.importer.external.metadatamapping.contributor.MetadataContributor needs to be added. This interface contains a type argument. The type needs to match the type used in the implementation of AbstractImportSourceService. The responsibility of each MetadataContributor implementation is to generate a set of metadata from the retrieved document. How it does that is completely opaque to the AbstractImportSourceService but it is assumed that only one entity (i.e. item) is fed to the metadatum contributor.

...

This bean expects 2 property values:

...

:

...

• field: A reference to the configured spring bean of the DSpace metadata field. e.g. the "dc.title" bean declared above.
• query: The xpath expression used to select the record value returned by the remote source

    <bean id="titleContrib" class="org.dspace.importer.external.metadatamapping.contributor.SimpleXpathMetadatumContributor">
        <property name="field" ref="dc.title"/>
        <property name="query" value="dc:title"/>
    </bean>

...

• .
  
  

    <bean id="lastNameContribtitleContrib" class="org.dspace.importer.external.metadatamapping.contributor.SimpleXpathMetadatumContributor">
        <property name="field" ref="dc.contributor.authortitle"/>
        <property name="query" value="x:authors/x:author/x:surname"dc:title"/>
    </bean>

Multiple record fields can also be combined into one value. To implement a combined mapping first create a SimpleXpathMetadatumContributor as explained above for each part of the field.

/>
    </bean>
    <bean id="firstNameContriblastNameContrib" class="org.dspace.importer.external.metadatamapping.contributor.SimpleXpathMetadatumContributor">
        <property name="field" ref="dc.contributor.author"/>
        <property name="query" value="x:authors/x:author/x:given-name="dc.contributor.author"/>
    </bean>

Note that namespace prefixes used in the xpath queries are configured in bean "FullprefixMapping" in the same spring file.

...

...

    <

...

property name="query" value

...

="x:authors/x:author/x:surname"/>
    </bean>
    <

...

bean id="firstNameContrib" class="org.dspace.importer.external.metadatamapping.contributor.SimpleXpathMetadatumContributor">
        <

...

property 

...

name="

...

field" 

...

ref="dc.contributor.author"/>
        <

...

property name="query" value="

...

x:

...

authors/x:author/x:given-name"/>
    </

...

bean>

Note that namespace prefixes used in the xpath queries are configured in bean "FullprefixMapping" in the same spring file

...

.

    <util:

map id="

FullprefixMapping" 

key-type="

java.lang.String" value-type="java.

lang.

String">

        <

description>Defines the namespace mappin for the SimpleXpathMetadatum contributors</description>
        <

entry key="http://purl.org/dc/elements/1.1/" value="dc"/>
        <

entry key="http://www.w3.org/2005/Atom" value="x"/>
    </util:map>

    <beanutil:list id="authorContribcombinedauthorList" classvalue-type="org.dspace.importer.external.metadatamapping.contributor.CombinedMetadatumContributor".MetadataContributor" list-class="java.util.LinkedList">

        <ref bean="lastNameContrib"/>
        <propertyref namebean="separator" value=", firstNameContrib"/>
        <property name="metadatumContributors" ref="combinedauthorList"/>
        <property name="field" ref="dc.contributor.author"/>
    </bean>

/util:list>

    <

...

bean id="authorContrib" class="org.dspace.importer.external.metadatamapping.contributor.

...

CombinedMetadatumContributor">
        

...

<property name="separator" value

...

="

...

, "/>
        <

...

property 

...

name="

...

metadatumContributors" 

...

ref="

...

combinedauthorList"/>
        <

...

property name="field" ref="dc.contributor.author

...

"/>
    </

...

Note that the single field mappings used for the combined author mapping are not added to this list.

Framework Sources Implementations

PubMed Integration

Introduction

First read the base documentation on external importing This documentation explains the implementation of the importer framework using PubMed (http://www.ncbi.nlm.nih.gov/pubmed) as an example.

Enabling PubMed Lookup (XMLUI Only)

bean>

    <util:map id="org.dspace.importer.external.metadatamapping.MetadataFieldConfig"
              value-type="org.dspace.importer.external.metadatamapping.contributor.MetadataContributor">
        <entry key-ref="dc.title" value-ref="titleContrib"/>

...

...

...

...

...

...

<entry key-ref="dc.contributor.author" value-ref="authorContrib"/>
    </util:map>

Available Metadata Contributor classes