wiki:OAIHarvester

Context Navigation

Version 4 (modified by Menzo Windhouwer, 9 years ago) (diff)
More on future plans

Responsible for this page: Menzo Windhouwer?.
Last content check: 28-10-2015

Purpose

The purpose of this page is to collect relevant information about the OAI Harvester.

Project: OAI Harvester

The OAI Harvester manages the regular harvests of CMD records from endpoints provided by the CLARIN centers and additionally harvests of OLAC and DC records from various other endpoints.

Project: OAI Harvester

People

Menzo Windhouwer?

Getting code

You can browse the code https://github.com/TheLanguageArchive/oai-harvest-manager
Git clone from: https://github.com/TheLanguageArchive/oai-harvest-manager.git

Usage

The deployment package contains a script to start the app, run-harvester.sh (for Unix systems including Mac OS X; we can add a Windows batch file if anyone wants it). The simplest usage is:

run-harvester.sh config.xml

where config.xml is the configuration file you wish to use. Additionally, parameters can be defined on the command line. For example:

run-harvester.sh timeout=30 config.xml

will set the connection timeout to 30 seconds. This value will override the timeout value defined in config.xml, if any. The first parameter that does not contain = is taken as the configuration file name.

Configuration

The behaviour of the app is determined by a single configuration file. The configuration file is composed of four sections:

settings, where options such as directory paths and timeouts are set;
directories, where output paths are defined;
actions, the most complex section, where actionSequences of actions can be defined for different metadata formats (actions include semantic transformations and saving intermediary or final results into a file); and
providers, where endpoints for the providers to be harvested are listed.

To get a clear idea of the structure of the configuration file, see the sample configuration files.

System Requirements

Dependencies

This application does not itself contain an implementation of the OAI-PMH protocol; it uses the OCLC harvester2 library for OAI-PMH requests.

Building and Deploying

mvn clean package assembly:assembly

The above build process creates a package named oai-harvest-manager-x.y.z.tar.gz (where x.y.z is a version number). This package can be deployed where needed.

Interfaces

Design

Tickets

https://github.com/TheLanguageArchive/oai-harvest-manager/issues

Status, Planning and Roadmap

Status: active

Planning and roadmap:

switch to the ListRecords scenario, where batches of records are requested from the providers
get rid of OCLC harvester2 library, which prevents specific timeouts etc. per endpoint
get rid of always building a DOM, which blows up memory consumption
create a new OAI harvester viewer

The new OAI harvester viewer

This new viewer should provide some advantages over the current viewer:

paged listing of records harvested
jump from records to the OAI request it originates from (in the ListRecords scenario a request can contain multiple records which cannot be handled by the current viewer)
keep some statistics over past runs, so a warning can be send when a 'sudden' drop in the number of records is experienced
can also provide access to the archived harvests

Additionaly the viewer can also be an access point for tools to assess the quality of the CMD records:

run XSD validation on the records/a record
run Schematron rules agains the records/a record, e.g., to check against best practices
run the VLO importer to see if the records/a record would be included in the VLO and which facet values it would deliver
check the profiles used, e.g., in CMDI 1.2 one could check if deprecated profiles are used or now already how well they cover the VLO facets
calculate a quality score (see LREC 2014 paper)

These tools could by run by default on all records or allow to select a specific record to check, but also allow the upload of a record.

More in the OAI domain we could also trigger a run against a OAI validator, e.g, http://validator.oaipmh.com/, and/or allow to trigger a harvest for a specific endpoint. The latter might need a specific setup/installation to not interfere with the periodic CLARIN harvest.