This software implements the metadata model that datalad and datalad-metalad will use in the future (datalad-metalad>=0.3.0) to handle metadata.
The metadata model is defined by the API of the top-level classes. Those are:
MetadataRootRecord -- holds top-level metadata information for a single
version of a datalad dataset
UUIDSet -- holds metadata root records for a set of datasets that are
identified by their UUIDs and their version.
TreeVersionList -- holds metadata root records and a sub-dataset tree for a
dataset version and its sub-datasets
Metadata -- represents metadata for a single item, i.e. dataset or file.
Metadata is associated with extractor names and extraction parameters.
DatasetTree -- a representation of the sub-dataset hierarchy of a dataset
FileTree -- a representation of the file-tree of a dataset
...
Because of the large size of some datalad-datasets, e.g. tens of thousands of sub-datasets and hundres of millions of files, the implementation allows focus-based operations on individual parts of the potentially very large metadata model. The implementation uses the proxy-pattern, that means, it loads, modifies, and saves only the minimal necessary model elements that are necessary to operate on the metadata-information that the user is interested in.
The model elements have to be persisted on a storage backend. How the model is mapped on storage backends is defined by the storage layer, that is to a large degree independent of the model layer. The intention is to support multiple storage backends in the past.
Currently only one storage backend is supported:
git-mapping -- a storage backend that stores a metadata model in a
git repository. The model objects are stored outside of existing branches.
They are referenced by datalad-specific git-references under refs/datalad/*