Selecting a row from a dataframe based on condition in R - mysql

Question

I am trying to subset a dataframe based on each user_id and order_date.
If ecomm_id and pulse_id exists in the row for that userid and for order_date, that row should be selected to new dataframe.
Else only one row with no ecomm_id must be selected to the new data frame and all other rows must be discarded.
Sample data:
userid returning device store_n testid ecomm_id pulse_id order_date
1.00 1 0 9328 Experience E 1 23 7/25/2015
1.00 1 0 NA Experience E NA NA 7/25/2015
2.00 1 1 NA Experience C NA NA 7/14/2015
3.00 1 0 3486 Experience F 2 86 7/23/2015
3.00 1 0 NA Experience F NA NA 7/24/2015
3.00 1 0 NA Experience F NA NA 7/24/2015
Expected Output:
userid returning device store_n testid ecomm_id pulse_id order_date
1.00 1 0 9328 Experience E 1 23 7/25/2015
2.00 1 1 NA Experience C NA NA 7/14/2015
3.00 1 0 3486 Experience F 2 86 7/23/2015
3.00 1 0 NA Experience F NA NA 7/24/2015

Answer 1

Hope this helps!
df <- data.frame(userid=c(1,1,2,3,3,3),
returning=c(1,1,1,1,1,1),
device=c(0,0,1,0,0,0),
store_n=c(9328,NA,NA,3486,NA,NA),
testid=c('Experience E','Experience E','Experience C','Experience F','Experience F','Experience F'),
ecomm_id=c(1,NA,NA,2,NA,NA),
pulse_id=c(23,NA,NA,86,NA,NA),
order_date=c('7/25/2015','7/25/2015','7/14/2015','7/23/2015','7/24/2015','7/24/2015')
)
library(dplyr)
df1 <- unique(df) %>% group_by(userid,order_date) %>% summarise(count=n())
df1 <- merge(unique(df),df1,on=c(userid,order_date))
final_df <- df1[!(is.na(df1$ecomm_id) & is.na(df1$pulse_id) & df1$count > 1),-ncol(df1)]
Don't forget to let us know if it solved your problem :)

Answer 2

With data.table, this becomes a concise "one-liner":
library(data.table)
setDT(DT)[order(ecomm_id), .SD[1], keyby = .(userid, order_date)]
userid order_date returning device store_n testid tid ecomm_id pulse_id
1: 1.00 7/25/2015 1 0 9328 Experience E 1 23
2: 2.00 7/14/2015 1 1 NA Experience C NA NA
3: 3.00 7/23/2015 1 0 3486 Experience F 2 86
4: 3.00 7/24/2015 1 0 NA Experience F NA NA
By ordering by ecomm_id, the NA entries are moved to the bottom. Now, for each combination of userid and order_date the first element within that group is picked.
Note that this assumes that there is at most one entry per group in case of non-NA ecomm_ids because the OP has specified:
If ecomm_id and pulse_id exists in the row for that userid and for order_date, that row should be selected to new dataframe.