# Dataframes/Polars Roadmap ## 0.94.x * [Transition to a lazy dataframe-based system](https://hackmd.io/mE6PiA_qTUSCtLu3Ztworw): * This is a simplification that @elferherrera had originally wanted to add. When the dataframes crate was first introduced, lazy dataframes were not yet available. This is why the API is more eager focused with lazy dataframes feeling a little tacked on. * This aligns the polars plugin more closely with how the python polars API works. * This will dramatically simplify the code base, as all the conversions back and forth between lazy and eager dataframes will no longer be necessary. * Removal of the nu-cmd-dataframe crate ## 0.95.x * Utilize streams for `polars into-nu` * Implement additional string functions * [Command unification/refactor](https://hackmd.io/@nucore/ryrusmlW0) ## 0.93.x * Introduction of `nu_plugin_polars` * Upgrade to polars 0.39.x * Deprecation of the `nu-cmd-dataframes` crate: * No new changes will be made to this crate beyond fixing compilation breakages and library upgrades. It is slated for removal in 0.95. * Removal of `polars dtypes`: * This command has been largely duplicated by `polars schema`, which is easier to work with as it can be applied as a schema to other commands. * Introduction of polars internal cache commands: * Move `polars ls` to `polars store-ls` for consistency. * Addition of `polars store-rm` to remove items in the cache. * Addition of `polars store-get` to pull an item from the cache. This command will require adding a simple reference counter to the cache entries. ## Backlog * [Cloud Storage Support](https://docs.pola.rs/user-guide/io/cloud-storage/) * [Polars Optimized with GPU chips](https://pola.rs/posts/polars-on-gpu/) * Bug: [Incorrect dtype of object of polars into-df when numeric and null values are mixed.](https://github.com/nushell/nushell/issues/12726) * [polars sample with quantiles specified doesn't work.](https://github.com/nushell/nushell/issues/12730)