Right now replyr considers its materialization operator to be dplyr::compute(). For Sparklyr we need to look into both spaklyr::sdf_checkpoint() (a new op in the development version) and user-specified sparklyr::spark_write_parquet() plus sparklyr::spark_read_parquet().
Right now
replyrconsiders its materialization operator to bedplyr::compute(). ForSparklyrwe need to look into bothspaklyr::sdf_checkpoint()(a new op in the development version) and user-specifiedsparklyr::spark_write_parquet()plussparklyr::spark_read_parquet().