Serialization and Transfer

[domoritz]

The data that is being visualized with Vega and Vega-Lite needs to be loaded into a browser. Right now, we serialize this data into text formats (e.g., JSON/CSV) which incurs a large overhead. Thus, for larger datasets, data transfer is often the main bottleneck in Vega. We plan to integrate more efficient binary formats (such as Apache Arrow) into the Vega runtime, to improve transfer in JupyterLab and other data science environments. See vega/vega#1300 for details.