Currently Timefence accepts Parquet, CSV, SQL queries, and pandas DataFrames. Polars is increasingly popular in the ML/data engineering space, and accepting polars.DataFrame and polars.LazyFrame natively (without going through Parquet) would reduce friction.
Since DuckDB can scan Polars DataFrames directly, this should be straightforward to wire up.
Scope:
- Accept
polars.DataFrame / polars.LazyFrame in Source(df=...) and Labels(df=...)
- Return
polars.DataFrame from build() when input is Polars
- Keep polars as an optional dependency
Currently Timefence accepts Parquet, CSV, SQL queries, and pandas DataFrames. Polars is increasingly popular in the ML/data engineering space, and accepting
polars.DataFrameandpolars.LazyFramenatively (without going through Parquet) would reduce friction.Since DuckDB can scan Polars DataFrames directly, this should be straightforward to wire up.
Scope:
polars.DataFrame/polars.LazyFrameinSource(df=...)andLabels(df=...)polars.DataFramefrombuild()when input is Polars