Samvera::NestingIndexer
The Samvera::NestingIndexer gem is responsible for indexing the graph relationship of objects. It maps a PreservationDocument to an IndexDocument by mapping a PreservationDocument's direct parents into the paths to get from a root document to the given PreservationDocument.
Background
This is a sandbox to work through the reindexing strategy as it relates to CurateND Collections. At this point the code is separate to allow for raid testing and prototyping (no sense spinning up SOLR and Fedora to walk an arbitrary graph).
Concepts
As we are indexing objects, we have two types of documents:
- PreservationDocument - a light-weight representation of a Fedora object
- IndexDocument - a light-weight representation of a SOLR document object
We have four attributes to consider for indexing the graph:
- id - the unique identifier for a document
- parent_ids - the ids for all of the parents of a given document
- pathnames - the paths to traverse from a root document to the given document
- ancestors - the pathnames of each of the ancestors
See Samvera::NestingIndexer::Documents::IndexDocument for further discussion.
To reindex a single document, we leverage the Samvera::NestingIndexer.reindex_relationships method.
Examples
Given the following PreservationDocuments:
| PID | Parents |
|---|---|
| A | - |
| B | - |
| C | A |
| D | A, B |
| E | C |
If we were to reindex the above PreservationDocuments, we will generate the following IndexDocuments:
| PID | Parents | Pathnames | Ancestors |
|---|---|---|---|
| A | - | [A] | [] |
| B | - | [B] | [] |
| C | A | [A/C] | [A] |
| D | A, B | [A/D, B/D] | [A, B] |
| E | C | [A/C/E] | [A/C] |
For more scenarios, look at the Reindex PID and Descendants specs.
Adapters
An AbstractAdapter provides the method interface for others to build against.
The InMemory adapter is a reference implementation (and used to ease testing overhead).
CurateND has implemented the following adapter for its LibraryCollection indexing.
To define the adapter for your application:
# In an application initializer (e.g. config/samvera_indexer_config.rb)
Samvera::NestingIndexer.configure do |config|
config.adapter = MyCustomAdapter
end
To best ensure you have implemented the adapter to spec:
# In the spec for MyCustomAdapter
require 'samvera/nesting_indexer/adapters/interface_behavior_spec'
RSpec.describe MyCustomAdapter
it_behaves_like 'a Samvera::NestingIndexer::Adapter'
end
See CurateND for our adaptor configuration.
Considerations
Given a single object A, when we reindex A, we:
- Find the parent objects of A to calculate the ancestors and pathnames
- Iterate through each descendant, in a breadth-first process, to reindex it (and each descendant's descendants).
This is a potentially time consumptive process and should not be run within the request cycle.