Guidelines for sharing NWB extensions (NDX)


The purpose of this document is to define the requirements and strategy for sharing format extensions for NWB, so called Neurodata Extensions (NDX).


Requirements for sharing extensions

A Public NDX MUST follow the rules outlined below in order to be considered compliant. Internal NDX are also highly RECOMMENDED to follow the same rules outlined below as much as possible.

  1. Releases of extensions MUST minimally include:
  1. The complete source and namespace YAML files for the NDX, which MUST be compliant with the NWB specification language. In addition to “name”, and “schema”, the namespace specification further MUST include the “author”, “contact”, and “version” of the namespace. The version number MUST follow the NWB versioning guidelines.
  2. LICENSE file detailing the license for the NDX. Permissive licenses SHOULD be used if possible. BSD license is RECOMMENDED. In particular for release via the NDX Catalog (see Sec. 4.2), the license MUST include language giving permission to members of the GitHub organization to distribute the extensions.
  3. Release notes detailing the changes between extension versions. Release notes MAY be included in the detailed RST documentation or as a separate (or .rst) file.
  4. Requirements file detailing dependencies on other NDX and where they can be installed from. This is required as the namespace.yaml only includes the names of namespaces that are being included.
  5. (or .rst) file with an overview of what the NDX does and additional details about the NDX.
  1. Releases of extensions SHOULD further include:
  1. The sources for generating the YAML specification using the PyNWB/HDMF extensions APIs. Sharing the sources eases maintenance, testing, and reuse of the NDX.
  2. Additional documentation in Sphinx RST format describing the NDX, credits, legal, and acknowledgements.
  3. Examples and tutorials describing the use of the extension.
  1. Releases of extensions MAY further include:
  1. Custom Python plugins (e.g., Container classes for neurodata types) for PyNWB. This is RECOMMENDED to make the NDX accessible to end users and has the advantage that custom functionality for interacting with the data can be defined.
  2. Custom Matlab plugins for MatNWB for representing neurodata types.
  1. Performing releases:
  1. To ensure compliance with the semantic versioning rules, public NDX MUST be released in a persistent fashion (i.e., sources of a particular version MUST not be modified after release and MUST remain accessible).
  2. Public NDX MUST register with the NDX Catalog. To register with the catalog, a public NDX MUST create an ndx-custom-record Git repository as part of the nwb-extensions GitHub organization. An ndx-custom-record repo contains basic metadata about the NDX and information needed for install and deployment. The sources of the extension MAY be hosted in a different public ndx-custom repository.
  3. Public ndx-custom repositories SHOULD use Git tags to ensure compliance with (4.1) and NDX SHOULD follow the NDX Catalog’s sharing strategies. Note: software versions managed via Git tags MAY differ from the version of the extensions namespace. However, both MUST follow NWB versioning guidelines.
  1. Naming namespaces and repositories:
  1. The name key of the namespace for the extension SHOULD generally follow the following naming conventions:
  1. Start with the prefix “ndx-”, e.g., “ndx-cortical-surface”
  2. Use short and descriptive names
  3. Use only lower-case ASCII letters (no special characters)
  4. Use “-” to separate different parts of the name (no spaces allowed)
  1. The namespace YAML file with the specification of the namespace SHOULD have the same name as the name key of the main namespace and the .namespace.yaml suffix SHOULD be added. E.g., if the name of the namespace is “ndx-cortical-surface”, then the YAML file would be called “ndx-cortical-surface.namespace.yaml”.
  2. The ndx-custom repository of an extension SHOULD generally follow the following conventions:
  1. Have the same name as the name key of the main namespace of the NDX (e.g., “ndx-cortical-surface”)
  2. Use GitHub (preferred) or other public Git hosting service
  3. Include a description
  4. Specify at least the following GitHub topics: ndx-extension
  1. The Python package for the NDX SHOULD have the same name as the name key of the main namespace of the NDX, where hyphens are replaced with underscores (e.g., from ndx_cortical_surface import CorticalSurface).
  2. The ndx-custom-record repository as part of the NDX Catalog MUST start with the same name as the name key of the main namespace of the NDX and have the -record suffix (e.g., “ndx-cortical-surface-record”) . This strategy helps ensure that there are no name conflicts between public NDX namespace names. This repository is created automatically, so end users need not worry about this.