Validation & Persistence¶
Validation runs on a full subtree before any data is saved to the graph and has three stages:
- check data against type-based constraints
- generate identifiers, using type-based rules
- check integrity and uniqueness
Currently, data validation is mainly concerned with ensuring that
certain mandatory properties are not absent or blank. For example,
Documentary Unit items require an
identifier property. Their
descriptions likewise require both a
languageCode and a
Data validation must be done prior to generating IDs because IDs can be derived from the data itself in some circumstances. If the data is not valid, ID generation would then fail.
(See the identifiers page for more on the rationale.)
Graph IDs are generated in differently depending on the type of item. There are currently two types of ID generators:
- data and scope-based
- random UUID
For most 'first class' node types, graph IDs are generated by combining the ID of the item's parent scope(s) with a particular data value belonging to the item itself. For example, a documentary unit will typically belong to two scopes: a repository, and the country in which that repository resides. Its eventual graph ID therefore forms a path:
There are two main reasons for generating graph IDs in this manner: they are to some extent human-readable and reproducible, and it makes it possible to ensure the uniqueness of identifiers within scopes without necessitating complex graph scans.
Item types which are not 'first class', i.e. dependent on some other item, are typically assigned random UUIDs.
The integrity checking runs after ID generation and does the following:
- ensures there are no ID collisions, which would indicate an identifier is not unique within a particular scope.
- checks uniqueness constraints for properties which must be globally unique. The facility is not currently used.