Using Data Virtualization, the ThinkData Platform is exceptionally good at indexing and managing data tables sourced from external warehouses. But users do not need to connect to data in order to create a dataset.
Using the platform, it is possible to create what we call "Metadata Datasets", datasets that are not connected to actual data tables. These metadata datasets are indexed by the platform's search engine, and can have most of the metadata associated with a connected dataset.
As with any dataset, a metadata dataset only requires a source and a title in order to be created. Remember that a data source is the semantic origin of the dataset. Due to this, metadata datasets and connected datasets can originate from the same source.
Platform users create metadata datasets to accurately index their production data in the platform, even before all data connections have occurred. This enables data stewards to analyze and explore the organization's data catalog quickly and efficiently. Metadata datasets can be easily upgraded to connected datasets by editing the dataset and choosing a preconfigured connection.
Metadata datasets are very useful for organizations seeking to index their data environment and create business-facing data catalogs.
Of course, metadata datasets are only as useful as the metadata you add to them. Review the metadata addition section of the knowledge hub to learn about the types of metadata you can add to a dataset.
Remember: The only things a dataset needs in order to be created are a source and a title. Remember that a data source may connect to live data using a data connection, but that the source itself is merely a description of where the data comes from.
Step 1: Select "+ Create dataset" from the Search or Data tab in the sidebar
Step 2: Give your dataset a title. You can change this at any time
Step 3: Choose or create a source for your dataset. This is where the data lives. In the case of a metadata dataset, the data source is an additional descriptive field and not a physical connection to underlying data.
Step 4: (Optional) Add additional metadata as needed. This might include adding description, classifications, templated metadata, or custom metadata fields.
Step 5: (Optional) Define a data dictionary. Using the Platform, you can manually create a data dictionary that includes all the columns in the metadata dataset. If you eventually connect the metadata dataset to a connected data table, the two data dictionaries will merge. If they are different, the columns in the connected table will be prioritized.
Step 6: Select "Create dataset" to save the metadata dataset. The return to the dataset to change the metadata or add data, select the Edit button on the dataset page. You will notice that metadata datasets do not have a table view (there is no data to query), and that data profiling and export have been disabled. Data profiling is an analysis of physical data structure, and only applies to connected datasets. Data export is not available for metadata datasets.