Weird behavior of group_by + lag

[tongcx]

library(dplyr)
df = read.csv(text ='
   name time
1   Rob    3
2  Pete    2
3   Rob    5
4  John    3
5   Rob    2
6  Pete    3
7  John    2
8  Pete    4
9  John    1
10 Pete    1
11  Rob    4
12  Rob    1
', sep='')
df %>% group_by(name) %>% mutate(lag_time = lag(time))

Gets me

   name time lag_time
1   Rob    3       NA
2  Pete    2       NA
3   Rob    5        2
4  John    3       NA
5   Rob    2        3
6  Pete    3        2
7  John    2        3
8  Pete    4        2
9  John    1        4
10 Pete    1        1
11  Rob    4        1
12  Rob    1        4

Should row 3 be Rob 5 3? Is this a bug or intended behavior?

[pkanavos]

I can reproduce the problem. It wasn't there with 0.3.2, appeared after upgrading to 0.4.1

[alex23lemm]

I can reproduce the issue as well. With the upgrade to version 0.4.1 the parts of my code using lag on grouped data started producing wrong results.

Here is another SO question showing the current issue.

[hadley]

@romainfrancois can you take a look please?