Though Jupyter notebooks are a useful tool for working with datasets of all shapes and sizes, until now a user has had to write their own custom code within the notebook in order to marshal and load large or remote datasets. Recently, significant progress has been made towards adding tools to the JupyterLab UI that will enable working interactively with such datasets.
notebook servercontents api always eagerly creates a local copy of any loaded datacontents apijupyterlab-remote-data: Ian R Roseremotecontentmanager) within the notebook server itself@jupyterlab/hdf5: my own contributioncontents api by serving its data via its own custom server extensionh5py@phosphor/datagrid, which can be set up to fetch data only as needed@jupyterlab/hdf5 can be used to work with very large (TB) files in remote environmentsjupyterlab-data-explorer: Saul Shanabrook, et al@jupyterlab/hdf5 includes support for data-explorercontents apijupyterlab-data-explorer is the leading candidate to replace the filebrowserdata-explorer and then consume the data within notebook code