Limitations
We are working hard to create a unified API from the most popular dataframe libraries. However, all the technologies are in different development stages. Many issues have been flying under the radar, but here, we want to highlight some of the most important ones.
Right now, the main limitations are as follows:
- Creating a UDF for string processing in cuDF and Dask-cuDF since they are not supported yet: https://github.com/rapidsai/cudf/issues/7301.
- cuDF, Dask-cuDF, and Vaex database connections are handled using Dask, which needs to load data as pandas dataframes and then convert them into the appropriate format based on a certain engine.
- Regex cuDF support is limited. For example, it still can handle lowercase and uppercase characters at the same time: https://github.com/rapidsai/cudf/issues/5217.