The openMINDS umbrella

In summary, openMINDS is the overall umbrella for a set of distributed GitHub repositories, each defining a particular metadata model for neuroscience research products.

The main (or central) openMINDS GitHub repository ingests all these GitHub repositories as git-submodules. Furthermore it stores the openMINDS vocabulary (vocab), providing general definitions and references for types and properties used in schemas across all openMINDS repositories (cf. below). And last but not least, it holds the schema representations for all supported metadata formats created by the openMINDS integration pipeline (cf. below).

For this to work smoothly for the existing, but also for all new openMINDS metadata models, the corresponding openMINDS submodules (GitHub repositories) have to meet the following requirements:

(1) The openMINDS metadata model has to be located on a public GitHub repository and published under an MIT license.

(2) The GitHub repository should have at least one version branch (e.g., "v1").

(3) The version branch should have the following main directory folders: schemas (required), tests (recommended),  examples (recommended), and img (optional).

(4) The schemas folder should contain the schemas of that metadata model implemented in the openMINDS schema template syntax (cf. below). The directory of the schemas can be further structured or flat.

(5) The tests folder should contain test-instances (JSON-LDs) for the schemas in a flat directory. The file names for these test-instances should follow the convention of

<<XXX>>-<<YYY>>.jsonld

for files that should pass the tests, and

<<XXX>>-<<YYY>>-nok.jsonld

for files that should fail the test. In both cases, <<XXX>> should be replaced with the label of the schema that is tested, and <<YYY>> with a user defined label for what aspect is tested (e.g., person-withoutCI.jsonld).

(6) The examples folder should contain examples for valid instance collections for that metadata model. Each example should receive its own directory (folder) with a README.md describing the example, and an metadataCollection subfolder containing the openMINDS instances (JSON-LDs). This subfolder can be further structured or flat.

(7) The img folder should contain image files used on that GitHub repository (e.g., the logo of the new openMINDS metadata model). The directory of the images can be further structured or flat.

The openMINDS vocabulary

Located under the folder vocab in the main openMINDS GitHub directory, the openMINDS vocabulary is semi-automatically gathered and stored in dedicated JSON files (types.json and properties.json). The openMINDS integration pipeline makes sure that both files are updated with each commit to any of the GitHub repositories for the openMINDS metadata models. For this reason, the files always contain an up-to-date list of all schemas and properties in use. This setup not only allows us to centrally maintain general definitions and references for schema types and properties across all openMINDS metadata models, but also to keep several technical aspects for the openMINDS schemas hidden from the naive user. How this works is explained in the following.

The types.json file is an associative array listing all existing openMINDS schemas. For each openMINDS schema a nested associative array is created providing the respective display label, the general description, and a list of references to corresponding schemas of other initiatives:

With each new schema committed to one of the openMINDS metadata models, a new entry is appended to the types.json file, with the display label automatically derived from the respective schema type and the remaining attributes provided with a null value. Once an entry for a schema is made in the types.json file, all attributes (name, description, and translatableTo) can be manually edited. All manual editions will be preserved and not overwritten when the file is updated again with a new commit. In case a schema is deleted from the openMINDS metadata models, the corresponding entry in the types.json is marked as being deprecated (additional attribute; "deprecated": true). It only can be permanently removed from the types.json file, if the entry is manually deleted.

Similar to the types.json file, the properties.json file is an associative array listing all properties across all existing openMINDS schemas. For each openMINDS property a nested associative array is created providing the respective display label, the general description, the label for displaying the reversed link of that property, the list of schemas in which the property is used, and a list of references to matching schema properties of other initiatives:

The keywords of those nested dictionaries are pre-defined to consistently capture for all schema types and properties their namespace, their occurrence (cf. "schemas" in properties.json), their general description (cf. "description" in types.json and properties.json), and possible references to related or matching schema types (cf. "translatableTo" in types.json) and properties  (cf. ""sameAs"" in properties.json) of other metadata initiatives (e.g., schema.org). 

This setup also allows us to define some values/entries to be automatically filled in by the openMINDS integration pipeline with each commit to one of the openMINDS repositories ("name", "schemas") and others to be manually editable later on ("description", "translatableTo", "sameAs", "nameForReverseLink"). 

For security, outdated entries in those openMINDS vocabulary files (e.g., because the namespace of the schema type or property changed or the schema type or property was deleted) are not automatically deleted, but kept and marked as being deprecated. After evaluation, deprecated schema types or properties can be deleted manually from openMINDS vocabulary.

With that, the openMINDS vocab reflects always an up-to-date status of the schema types and properties in use across all openMINDS metadata models, while providing the opportunity to centrally review and maintain their consistency and references. 

The openMINDS schema template syntax

All openMINDS metadata models are defined using a light-weighted schema template syntax. Although this schema template syntax is inspired by JSON-Schema, it outsources most schema technicalities to be handled in the openMINDS integration pipeline, making the openMINDS schemas more human-readable, especially for untrained eyes. 

The few remaining customized technical properties which need additional interpretation or translation to a formal schema languages (e.g. JSON-Schema) have an underscore as prefix (e.g., "_type"). Within the openMINDS integration pipeline (cf. below), the schema template syntax is interpreted, extended and flexibly translated to various formal schema languages. All further specifications of the openMINDS schema template syntax are described below.

All openMINDS schemas need to have the extension .schema.tpl.json and each schema is defined as a nested associative array (dictionary) with the following conceptual structure:

"_type" defines the schema type (or namespace) with the depicted naming convention, where the label of the respective openMINDS metadata model (e.g., "core") and the schema name (format: UpperCamelCase; e.g. "Person") have to be specified. Obviously, the schema name should be meaningful and provide some insides into what metadata content the schema covers. 

Under "properties" a nested associative array is defined, where each key defines the property name (format: lowerCamelCase; e.g. "givenName"). The corresponding value is again a nested associative array defining the expected data "type" (cf. below) and the "_instructions" for entering the correct metadata for the respective property. 

Under "required" a list of property names can be provided that are obligatory to be present in a correctly instantiated metadata instance of the respective schema. If none of the properties are required, this key-value pair does not have to be specified.

Now, depending on the expected data type additional constraints can be made for the metadata entry of a respective property. Currently, the openMINDS schema template syntax supports the following data types: "string", "integer", "float", "boolean", "array" and "object".

The openMINDS integration pipeline

(coming soon) If you'd like to learn more about the openMINDS integration pipeline, especially if you'd like to contribute to it, please get in touch with us (the openMINDS development team) via the issues on the openMINDS or openMINDS_generator GitHub or the support email: openminds@ebrains.eu