Skip to content
This repository has been archived by the owner on Nov 30, 2023. It is now read-only.

Latest commit

 

History

History
103 lines (70 loc) · 5.3 KB

README.md

File metadata and controls

103 lines (70 loc) · 5.3 KB
Raw

Introduction

The pass-indexer keeps an Elasticsearch index up to date with resources in a Fedora repository.

Design

The pass-indexer monitors a JMS queue for messages about creation, deletion, and modification of Fedora resources. (Fedora must be configured appropriately to setup this queue.)

The Elasticsearch index is created on startup if it does not exist with a set configuration. (That configuration default can be changed.) If the index does exist, the configuration is retrieved from the index. In either case the mapping must match the documents which will be indexed. The Elasticsearch document is the compact JSON-LD representation of that resource without server triples. If a key on the document is not present in the mapping for the index, then the key is removed from the document and a warning is logged.

When there is a message about a resource of a type being monitored, the indexer either creates a corresponding document in Elasticsearch from the Fedora resource, updates such a document, or deletes the document. Only messages about a resource of a type which matches a configured prefix, PI_TYPE_PREFIX, are handled. The id of the Elasticsearch document is the safe URL base64 encoding of resource path. This lets both the document be created and updated with the same PUT.

Auto-completion (suggestion) support

Suggestion is supported on field of type completion. In order to allow auto-completion on the values of a field NAME, add a mapping for NAME_suggest of type completion. The field NAME_suggest will be automatically populated with the contents of NAME by the indexer. The content of NAME will have a completion for the substring starting at each token. So if the content is "token1 token2 token3", there will be completions, "token1 token2 token3", "token2 token3", and "token3".

A consequence of the auto-completion is that NAME_suggest fields will appear in the _source document returned by Elasticsearch. See https://www.elastic.co/guide/en/elasticsearch/reference/current/search-request-source-filtering.html for options.

You can search like below to find journName with a word beginning with "R".

curl -X POST "http://localhost:9200/pass/_search?pretty" -H 'Content-Type: application/json' -d'
{
    "suggest": {
        "my-suggest" : {
            "prefix" : "R", 
            "completion" : { 
                "field" : "journalName_suggest" 
            }
        }
    }
}
'

Handling Fedora URIs

In PASS, a Fedora resource can be addressed in two different ways, by a public URI or a private URI. The public URI must pass through Shibboleth and can be used by the public at large. The private URI is used by the backend services and allows them to avoid Shibboleth.

A custom normalizer is defined to handle fields containing fedora URIs mapping them to their Fedora resource path. This allows searches to be done using either the public or private URI.

Pass Indexer Test Note:

In the integration test class PassIndexerIT, some tests may fail with an error message saying that some PUT requests fail because resources are system (fcrepo) managed. These errors are caused by 409 errors from Fedora See the Fedora documentation about the Containers RESTful API.

Award Number normalization

A custom normalizer is defined to help searching award numbers.

  • Ignore case
  • Strip out whitespace, dashes, and periods
  • Ignore the first 3 characters and zero padding in NIH award numbers.

Command line indexing tool

The command line tool runs the indexer with the configuration either specifies as environment variable or system properties.

Configuration properties:

The PI_FEDORA_USER and PI_FEDORA_PASS are the credentials used to connect to Fedora with basic auth.

The PI_FEDORA_JMS_USER and PI_FEDORA_JMS_PASSWORD are credentials used to connect to the activemq broker, if it is secured.

The PI_ES_INDEX is the index where Fedora documents are sent. PI_ES_CONFIG is the configuration used to create an index if it does not exist. It must be set even if not used. It's value may be a file path or a classpath resource or a URL.

Command line reindexing tool

This tool reads objects from Fedora and writes them to an Elasticsearch index. Records are not removed. If the index does not exist, it will be created with the correct configuration.

The PASS_FEDORA_BASEURL and PI_ES_INDEX environment variables must be set. A db is used to keep track of the indexing. The db writes out a db to the current working directory when run with no arguments. When run with a single argument, it adds record to the db.

The results table of the db can be examined with sqlite to check for errors.