groupingsets() adding an extra column with named vectors?

[sindribaldur]

groupingsets(
  data.table(iris),
  j = mean(Sepal.Length),
  by = c('Sp' = 'Species'),
  sets = list('Species')
)
     Sp    V1    Species
1: <NA> 5.006     setosa
2: <NA> 5.936 versicolor
3: <NA> 6.588  virginica

I'm using latest development version of data.table and R 4.1.0. This seems not to have been the behaviour about three months ago when I last ran the code that now broke.

[jangorecki]

Could you check verbose=T (if there is)? I think it may not be related to groupingsets, it calls ordinary dt code under the hood.

[sindribaldur]

Setting options(datatable.verbose = TRUE) I get:

Argument 'by' after substitute: by
i clause present and columns used in by detected, only these subset: [Species]
Detected that j uses these columns: [Sepal.Length]
lapply optimization is on, j unchanged as 'mean(Sepal.Length)'
Old mean optimization changed j from 'mean(Sepal.Length)' to '.External(Cfastmean, Sepal.Length, FALSE)'
Making each group and running j (GForce FALSE) ... 
  memcpy contiguous groups took 0.000s for 1 groups
  eval(j) took 0.000s for 1 calls
0.010s elapsed (0.020s cpu) 
Argument 'by' after substitute: by.set
Detected that j uses these columns: [Sepal.Length]
Finding groups using forderv ... forder.c received 150 rows and 1 columns
0.000s elapsed (0.000s cpu) 
Finding group sizes from the positions (can be avoided to save RAM) ... 0.000s elapsed (0.000s cpu) 
lapply optimization is on, j unchanged as 'mean(Sepal.Length)'
GForce optimized j to 'gmean(Sepal.Length)'
Making each group and running j (GForce TRUE) ... gforce initial population of grp took 0.002
gforce assign high and low took 0.000
This gmean took (narm=FALSE) ... gather took ... 0.000s
0.000s
gforce eval took 0.001
0.020s elapsed (0.000s cpu) 

[jangorecki]

Hm, it was split function that had verbose arg. Grouping sets doesn't provide this level of debug.

[sindribaldur]

I removed development version and installed again from CRAN and the issue is gone.

[ben-schwen]

For reproduceability and can confirm that there is an issue/changing behavior

# 1.14.2
groupingsets(
  data.table(iris),
  j = mean(Sepal.Length),
  by = c('Sp' = 'Species'),
  sets = list('Species')
)
#       Species    V1
#        <fctr> <num>
# 1:     setosa 5.006
# 2: versicolor 5.936
# 3:  virginica 6.588

# current dev 1.14.3
groupingsets(
  data.table(iris),
  j = mean(Sepal.Length),
  by = c('Sp' = 'Species'),
  sets = list('Species')
)
#        Sp    V1    Species
#    <fctr> <num>     <fctr>
# 1:   <NA> 5.006     setosa
# 2:   <NA> 5.936 versicolor
# 3:   <NA> 6.588  virginica

Doing some digging current behavior was introduced by #4713