#### the following code tested sucessfully on R 3.5.2 Windows 10
#### kaggle currently provide 3.4.2, hence not supporting sparklyr installation
#### hence uncomment and run from lines 5 to 17 in your latest R

#    library(sparklyr)
#    sc <- spark_connect(master = "local")
    

#    parquet_path <- "../input/train.parquet"
    
    # reading only two columns i.e. signals first
#    traindt <- spark_read_parquet(
#      sc, name = "traindt", path = "train.parquet",
#      memory = FALSE
#    ) %>% select("0", "1") %>%
#      sdf_persist()
    
#    traindt <- collect(traindt)

### read from local copy of above
traindt <- readRDS("../input/train2cols/traindt.RDS")
head(traindt)

#    spark_disconnect(sc)