Skip to content

Record relations

A relation is simply a link, a reference, from one record to another. Optionally, it can have some extra metadata to describe the relation. For example, in a library you can find a book written in English language and the same book in French. It is a nice feature for cataloguing purposes but also to show to the user that a book also exists in another language. It is natural to think then that a language relation can link these 2 books.

Understanding the data model

In InvenioILS, only document and series can have relations. Other models do not have this functionality enabled. (TODO link to domain section). Three kind of relation exists in InvenioILS: * Parent-Child relation: models the relation between a series (serial or multipart monograph) and a set of other series or documents. For example, a series (the parent) can contain a set of documents (the children), or it can contain multipart monographs (the children, but also the parent) which they contain documents (the children). InvenioILS models theses types of Parent-Child relation: * Multipart Monograph relation * Serial relation Parent-Child relations can optionally have the extra field volume that describe the relation. For example, in a serial or periodical, each issue is normally described by issue+volume: this information can be added to the relation description. TODO add a small diagram here?

  • Siblings relation: models the relation between records of the same model, where the order is not important, unordered. For example a series and a series or documents and documents. InvenioILS models theses types of Siblings relation:
  • Language relation
  • Edition relation

  • Other relation Siblings relations can optionally have the extra field note that describe the relation. This can useful to describe the Other relation and give a meaning to it.

  • Sequence relation: models the relation between records of the same model, where the order is important. For example, a journal series that at certain point in time stops the publication of volumes and issues (the previous) but it then continues in a completely new series (the next): the older series is continued by the new one, or the new series continues the older one.

Under the hood

Relations between records are stored safely in the Invenio database and they are used as source of truth when validating exiting of new relations. InvenioILS uses InvenioPIDRelations module to store the relations graph as Parent-Child relation, basically list of tuples Parent PID, Child PID and Relation Type. There is no extra metadata that can be stored at the moment for each relation. InvenioILS hides this internal implementation and adds an abstraction on top, to model not only Parent-Child relations, but also the other types described above.

Each record with relations has, at least, an extra field relations in its metadata: this field is stored in the database as a JSON Schema Reference and it is calculated "on-the-fly" by reading the PIDRelations db table when the record is fetched or indexed. This is to avoid any duplication of metadata both in the db and records metadata which could lead to data inconsistencies.

Since relations extra metadata, e.g. the volume for Parent-Child relations cannot be stored in the database, it is instead stored in the record metadata, in the relations_extra_metadata field. To avoid duplication, the extra metadata is stored only in one of the record of the relation: when fetching or indexing the second record of the relation, the extra metadata is automatically read from the first and added it on the fly. See the section below for more clear explanations.

Relations metadata

When creating a new relation, the extra metadata is always and only stored in the parent or first record metadata. Let's see an example.

Let's model a relation between the book "The lord of the rings" (PID: "J8H") and the book "The Hobbit" (PID: "M1A") of type Siblings Other, with an extra note 'same author'. Example of the REST payload: 

POST /documents/J8H/relations
{
   pid: 'M1A',
   pid_type: 'docid',
   relation_type: other',
   note: 'same author'
}
As result, the extra metadata note will be stored only in the book "The lord of the rings", the first. Let's see what happens when fetching both records from the db: 
GET /documents/J8H
{
  title: 'The lord of the rings',
  pid: 'J8H',
  edition: 'first',
  publication_year: 1954,
  ...
  relations: {
      other: {
        pid: 'M1A',
        pid_type: 'docid',
        note: 'same author',
        record_fields: {
          title: 'The Hobbit',
          edition: '1st',
          publication_year: 1937
        }
      }
  },
  relations_extra_metadata: {
    other: {
      note: 'same author',
      pid: 'M1A',
      pid_type: 'docid',
    }
  }
...
}

As expected, the first record of the relation has the field relations_extra_metadata with the extra note that we have added. The relations filled has been automatically generated by getting the information for the second record merged with the extra metadata.

GET /documents/M1A

{
  title: 'The Hobbit',
  pid: 'M1A'
  edition: '1st',
  publication_year: 1937,
  ...
  relations: {
      other: {
        pid: 'J8H',
        pid_type: 'docid',
        note: 'same author',
        record_fields: {
          title: 'The lord of the rings',
          edition: 'first',
          publication_year: 1954
        }
      }
  },
  relations_extra_metadata: {}
...
}

The second record do not contain the extra metadata (otherwise it would be a duplication), but it contains the relations field with the information of the first record and the extra metadata merged.