In WinVector/replyr: Patches to Use 'dplyr' on Remote Data Sources

library('dplyr')
library('sparklyr')
library('replyr')

sc <- sparklyr::spark_connect(version='2.2.0', 
                              master = "local")

d <- copy_to(sc, 
             data.frame(x=1:7, group=floor((1:7)/3)),
             name= 'd')

print(d)

pieces <- replyr_split(d, 'group', partitionMethod = 'extract')
print(pieces)

f <- function(pi) {
  ni <- replyr_nrow(pi)
  mutate(pi, n=ni)
}

pieces <- lapply(pieces, f)
print(pieces)

recovered <- replyr_bind_rows(pieces) %>%
  arrange(x)
print(recovered)

r2 <- d %>%
  gapply('group', f, partitionMethod = 'extract') %>%
  arrange(x)
print(r2)

spark_disconnect(sc)
rm(list=ls()); gc() # disconnect