careful when filtering with many groups in dplyr

dplyr_filter_ungroup.R

library(dplyr)

# create a dummy dataframe with 100,000 groups and 1,000,000 rows
# and partition by group_id
df <- data.frame(group_id=sample(1:1e5, 1e6, replace=T),
                 val=sample(1:100, 1e6, replace=T)) %>%
      group_by(group_id)

# filter rows with a value of 1 naively
system.time(df %>% filter(val == 1))

#   user  system elapsed
#  1.447   0.017   1.476

# ungroup before filtering for a huge speedup
system.time(df %>% ungroup() %>% filter(val == 1))

#   user  system elapsed
#  0.007   0.003   0.010