Query cuDF DataFrames¶
Everything in BlazingSQL is processed as a cudf.DataFrame, much like almost
everything in the RAPIDS ecosystem. Therefore you can easily build a BlazingSQL
table off of a cudf DataFrame, by only providing the table name and the
DataFrame itself.
import blazingsql
import cudf
bc = blazingsql.BlazingContext()
table_gdf = cudf.read_csv('...file_path/table.csv')
bc.create_table('table_name', table_gdf)
Parameters¶
table_name - string. Required. Name of the table you are creating.
cudf.DataFrame -
cudf.DataFrame. Required. The DataFrame you wish to query.
Query dask_cudf DataFrames¶
When using a Dask client to make BlazingSQL work in a distributed context, you can also create a BlazingSQL table from a dask_cudf DataFrame.
import blazingsql
import dask_cudf
from dask.distributed import Client
from dask_cuda import LocalCUDACluster
cluster = LocalCUDACluster()
client = Client(cluster)
bc = BlazingContext(dask_client=client, network_interface='lo')
table_ddf = dask_cudf.read_parquet('...file_path/table.parquet')
bc.create_table('table_name', table_ddf)
Parameters¶
table_name - string. Required. Name of the table you are creating.
dask_cudf.DataFrame -
dask_cudf.DataFrame. Required. The DataFrame you wish to query.