Open plusmid opened 4 months ago
Hi @plusmid, currently, pygwalker does not utilize the pandas API for dataframe-related computations.
When you pass a dataframe to pygwalker, it converts it into either a duckdb format or an in-browser memory JavaScript array. Therefore, the performance remains consistent across various dataframes.
We are exploring more native computation methods using the dataframe API. However, it appears that in most scenarios, using duckDB or integrating pygwalker with Snowflake yields quite good performance.
https://github.com/h2oai/datatable It has higher performance reading and manipulating big csv data than pandas/polars/modin. But I can't walk this datatable.frame type.