Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 13 Next »

Background

OpenMRS has a mechanism to map concepts to external sources (concept_map tables); however, OpenMRS does not yet have a means to map metadata to external sources. We have tried using UUIDs to uniquely (and globally) specify metadata across systems; however, this has failed us in several ways: (1) UUIDs end up needing to be manually entered by developers and implementers; (2) there are many cases where metadata created separately (with different UUIDs) are actually referring to the same thing; and, (3) modules introducing their own metadata have a difficult time properly working with existing metadata.

Goals

The goal of Metadata Mapping is to solve many of the problems around metadata management by providing an easy & explicit way for metadata within a system (encounter types, location, etc.) to be mapped to external vocubularies.

Use case examples:

  • Developers and implementers can refer to metadata by its more human-friendly mapping (e.g., source + code) instead of having to manually manipulate UUIDs.
  • Implementations with multiple servers no longer need to synchronize metadata UUIDs across systems; rather, tools can be designed to use metadata mappings to define sameness of metadata across servers.
  • Modules can use metadata mappings to allow administrators to create new metadata or map to existing metadata as needed.

A "1.0" version of Metadata Mapping would include a MetadataService within openmrs-core (could be introduced by a module initially or for the foreseeable future) and REST API endpoints to manage & search metadata mappings.

Assumptions

  • Source provides the "namespace" of codes and should be universally unique. We should encourage use of fully qualified names (modules defining metadata should use their fully qualified module ID, organizations should use their domain name, etc.); otherwise, managing the uniqueness of metadata sources (e.g., creating a registry) is out of scope for this effort.
  • We want to map from human-friendly source + code to the actual metadata (not just a row in a database table), so we map to Class and instance information (as "reference") instead of just linking to a row in a metadata table.
  • All metadata mapping are "SAME AS" relationships. Other types of mapping relationships are out of scope for this effort.

Data Model Design

The following design was proposed on the Design Forum 2015-06-22:

Notes on data model design:

  • metadata_source is used to define a unique source (authority) for each namespace of metadata terms.
    • name should be fully qualified and universally unique.
  • metadata_term table provides both the term and its mapping to local metadata.
    • code should be unique within the given source.
    • metadata_class refers to the Java class for the metadata.
    • metadata_reference is a unique reference to the metadata within the class (e.g., uuid)
  • metadata_set is used to define relating grouping of metadata similar to what OpenMRS has traditionally done within global properties and similar to FHIR's ValueSet for metadata terms.
    • sort_weight is used to optionally give members of a metadata set a reliable sequence.

API Design

The common operations for metadata mapping will include these use cases:

  • Manage mappings (add or remove mappings, should support bulk operations)
  • Search for mappings
    • By source
    • By code
    • By metadata object (given an OpenmrsMetadata instance, return any associated mappings)
  • Search for metadata (given a mapping UUID or source+code, return the OpenmrsMetadata instance)

API Call examples:

Fetch a location with the given source and code:
Location location = metadataService.getItem(Location.class, "SOME-SOURCE", "CODE");
// should be implemented as: generic <T> getItem(T type, String source, String code);
Fetch all visit types from the given source:
List<VisitType> visitTypes = metadataService.getItems(VisitType.class, "SOME-SOURCE");
// should be implemented as: generic <T> getItems(T type, String source);
Fetch all visit types for the given set:
List<VisitType> visitTypes = metadataSerivce
  .getItems(VisitType.class, "SOME-SOURCE", "CODE");
// should be implemented as: generic <T> getItems(T type, String source, String code); 
// source = metadataSet.source.name, code = metadataSet.code
Retire term:
List<MetadataTerm> terms = metadataService.getMetadataTerms(location);
metadataService.retireMapping(terms.get(0), "some reason");
Create term: 
MetadataSource source = metadataService.getSourceByName("SOME-SOURCE");
MetadataTerm term = new MetadataTerm(source, "CODE", location);
metadataService.saveMetadataTerm(term);
Create terms in bulk:
List<MetadataTerm> terms = Arrays.asList(mapping1, mapping2, mapping3);
metadataService.saveMetadataTerms(terms);
Get term by source and code:
MetadataTerm term = metadataService.getMetdataTerm(mappingSource, "CODE");
Get terms by source:
List<MetadataTerm> terms = metadataService.getMetadataTerms(source);
Get term by uuid:
MetadataTerm mapping = metadataService.getMetadataTermByUuid(String uuid);
Get source by uuid:
MetadataSource source = metadataService.getMetadataSourceByUuid(String uuid);
MetadataTerm, MetadataSource, MetadataSet should support all CRUD operations.

 

 

 

  • No labels