bug in group_by

[sfischme]

Try this:

            time type loc                             call     p cnt
1  1304945116000    a   b 1fb646244dbd6c807005811838b1e5ec 53115   0
2  1304945116000    a   b 6b3dafa1ff257c3891fc87abc78e137e     1   0
3  1304945116000    a   b 6e6f8c68c86173aa7e98b6375e752f89    NA   0
4  1304945116000    a   b d02922f7a545c15632c9ecb39e6da9a2    NA   0
5  1304945116000    a   b 11ce203ca480e56f03a7d1a9591de0e9    NA   0
6  1304945116000    a   b 7dabc9db21bbdc080780f493fe93f2af 53115   0
7  1304945116001    a   b 6b3dafa1ff257c3891fc87abc78e137e     1   1
8  1304945116001    a   b 6e6f8c68c86173aa7e98b6375e752f89    NA   1
9  1304945116001    a   b d02922f7a545c15632c9ecb39e6da9a2    NA   1
10 1304945116001    a   b 11ce203ca480e56f03a7d1a9591de0e9    NA   1

# now compare these two, they should be the same:
q.2 %>% group_by(time, type, loc, call) %>% summarize(n=n()) %>% filter(n>1)
q.2 %>% data.table() %>% group_by(time, type, loc, call) %>% summarize(n=n()) %>% filter(n>1)

[DavidArenburg]

groub_by is not data.table function...

[sfischme]

You are right, this produces the expected output:

q.2[,n:=nrow(.SD),by=list(time,type,loc,call)]

[eantonya]

@sfischme simply use .N instead of nrow(.SD)

[arunsrinivasan]

Convert time column to bit64::integer64 and try again. Or do setNumericRounding(0L) and try again. Read ?setNumericRounding to understand why we do this (including the examples). For large (integer) numbers, we recommend integer64.

I'll add that we recommend bit4::integer64 to ?setNumericRounding and then close the issue.