dplyr group_by summarise inconsistent number of rows

Question

I have been following the tutorial on DataCamp . I have the following line of code, that when I run it produces a different value for "drows"

hflights %>% 
group_by(UniqueCarrier, Dest) %>% 
summarise(rows= n(), drows = n_distinct(rows))

First time:

Source: local data frame [234 x 4]
Groups: UniqueCarrier [?]

        UniqueCarrier  Dest  rows drows
                <chr> <chr> <int> <int>
1             AirTran   ATL   211    86
2             AirTran   BKG    14     6
3              Alaska   SEA    32    18
4            American   DFW   186    74
5            American   MIA   129    57
6      American_Eagle   DFW   234   101
7      American_Eagle   LAX    74    34
8      American_Eagle   ORD   133    56
9  Atlantic_Southeast   ATL    64    28
10 Atlantic_Southeast   CVG     1     1
# ... with 224 more rows

Second time:

   Source: local data frame [234 x 4]
Groups: UniqueCarrier [?]

        UniqueCarrier  Dest  rows drows
                <chr> <chr> <int> <int>
1             AirTran   ATL   211   125
2             AirTran   BKG    14    13
3              Alaska   SEA    32    29
4            American   DFW   186   118
5            American   MIA   129    76
6      American_Eagle   DFW   234   143
7      American_Eagle   LAX    74    47
8      American_Eagle   ORD   133    85
9  Atlantic_Southeast   ATL    64    44
10 Atlantic_Southeast   CVG     1     1
# ... with 224 more rows

Third time:

Source: local data frame [234 x 4]
Groups: UniqueCarrier [?]

        UniqueCarrier  Dest  rows drows
                <chr> <chr> <int> <int>
1             AirTran   ATL   211    88
2             AirTran   BKG    14     7
3              Alaska   SEA    32    16
4            American   DFW   186    79
5            American   MIA   129    61
6      American_Eagle   DFW   234    95
7      American_Eagle   LAX    74    31
8      American_Eagle   ORD   133    67
9  Atlantic_Southeast   ATL    64    31
10 Atlantic_Southeast   CVG     1     1
# ... with 224 more rows

My question is why does this value constantly change? What is it doing?

Answer 1

Apparently this is normal behaviour, see this issue here. https://github.com/tidyverse/dplyr/issues/2222 .

This is because values in list columns are compared by reference, so n_distinct() treats them as different unless they really point to the same object:

So the internal storage of the df changes the way the thing works. Hadley's comment in that issue seems to say it might be a bug (in the sense of unwanted behaviour), or it might be expected behaviour they need to document better.