test 1960 still failing (grouping sets output mismatch)

[MichaelChirico]

@jangorecki this came up earlier in #3168: #3168 (comment)

The output on my machine differs from that on yours and on Travis/Codecov...

It's still going on... if we can figure out and squash before CRAN release that'd be great

DT = data.table(
  color = c("yellow", "red", "green", "red", "green", "red",
            "yellow", "yellow", "green", "green", "green", "yellow",
            "red", "yellow", "red", "green", "yellow", "red", "yellow",
            "red", "green", "yellow", "green", "green"),
  year = structure(c(15340, 15340, 14975, 15706, 15706, 15340,
                     16436, 15340, 15340, 14975, 16436, 15706,
                     16436, 15340, 14975, 14975, 16071, 15340,
                     15706, 16071, 15706, 15340, 16436, 16071), class = "Date"),
  status = structure(c(4L, 3L, 4L, 3L, 2L, 1L, 3L, 4L, 4L, 3L, 4L, 4L,
                       4L, 4L, 1L, 3L, 3L, 2L, 1L, 2L, 3L, 4L, 2L, 4L),
                     .Label = c("active", "archived", "inactive", "removed"),
                     class = "factor"),
  amount = c(1L, 4L, 2L, 3L, 1L, 5L, 1L, 1L, 4L, 2L, 3L, 1L,
             5L, 4L, 2L, 2L, 4L, 3L, 3L, 2L, 4L, 4L, 1L, 2L),
  value = c(2.5, 2, 3, 3, 2.5, 3.5, 2.5, 3.5, 3, 2.5, 3.5, 2.5, 2,
            2.5, 3, 3, 3, 3, 3, 3, 2, 2.5, 3, 3)
)
groupingsets(DT[ , .(amount, value)], j = 5, by = character(0L),
             sets = list(character()), id=TRUE)

Output in test.data.table() (also for R --vanilla):

Running test id 1960      Test 1960 ran without errors but failed check that x equals y:
> x = ans 
   grouping V1  [Key= Types=int,dou Classes=int,num]
1:       NA  5                                      
> y = data.table(grouping = 0L, V1 = 5) 
   grouping V1  [Key= Types=int,dou Classes=int,num]
1:        0  5                                      
Column 'grouping': 'is.NA' value mismatch: 0 in current 1 in target

My sessionInfo():

R version 3.5.1 (2018-07-02)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.6

Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/3.5/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.5/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] xts_0.11-2        zoo_1.8-4         bit64_0.9-7       bit_1.1-14       
[5] data.table_1.11.9

loaded via a namespace (and not attached):
[1] compiler_3.5.1  tools_3.5.1     grid_3.5.1      lattice_0.20-38

Related: #3173

[jangorecki]

I don't think it is an urgent issue for 1.12.0, using grouping sets and constant j doesn't make much sense.
You seems to have some extra packages attached, be sure to try on clean data.table-only session.
Grouping ID column is created using the following, check if you are getting NA or 0 already there

by = character(0L)
sets = list(character())
lapply(sets, function(by.set) strtoi(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""), base=2L))
#[[1]]
#[1] 0

you can also try putting browser() to investigate this branch

data.table/R/groupingsets.R

Lines 79 to 82 in 6346a88

	if (!is.data.table(empty)) {
	if (length(empty)>0) empty = empty[0L] # fix for #3173 when no grouping and j constant
	empty = setDT(list(empty)) # improve after #648, see comment in aggregate.set
	}

and then finally also

data.table/R/groupingsets.R

Lines 108 to 111 in 6346a88

	if (id) {
	# integer bit mask of aggregation levels: http://www.postgresql.org/docs/9.5/static/functions-aggregate.html#FUNCTIONS-GROUPING-TABLE
	set(r, j = "grouping", value = strtoi(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""), base=2L))
	}

removing bug as of now as it is possible it is machine dependent

[MichaelChirico]

@jangorecki extra packages are attached by test.data.table(), are they supposed to be un-attached by the time test.data.table() finishes?

This was run via

$ R --vanilla
> library(data.table)
> test.data.table()

And yes, indeed we've narrowed the issue:

by = character(0L)
sets = list(character())
lapply(sets, function(by.set) strtoi(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""), base=2L))
#[[1]]
# [1] NA

and in fact:

by = character(0L)
sets = list(character())
lapply(sets, function(by.set) {
    dput(by.set)
    dput(by %chin% by.set + 1L)
    dput(c("1", "0")[by %chin% by.set + 1L])
    dput(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""))
    dput(strtoi(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""), base=2L))
    strtoi(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""), base=2L)
})
# character(0)
# integer(0)
# character(0)
# ""
# NA_integer_
# [[1]]
# [1] NA

From what I see I'm amazed other machines are coming out with 0!

[jangorecki]

Manual of strtoi

An integer vector of the same length as x. Values which cannot be interpreted as integers or would overflow are returned as NA_integer_.

maybe macOS is not equipped with proper strtol library, you can try to debug that outside of R

[MichaelChirico]

I have a remote machine where 0L is coming out:

by = character(0L)
sets = list(character())
lapply(sets, function(by.set) {
    dput(by.set)
    dput(by %chin% by.set + 1L)
    dput(c("1", "0")[by %chin% by.set + 1L])
    dput(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""))
    dput(strtoi(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""), base=2L))
    strtoi(paste(c("1", "0")[by %chin% by.set + 1L], collapse=""), base=2L)
})

# character(0)
# integer(0)
# character(0)
# ""
# 0L
# [[1]]
# [1] 0

so it's really down to strtoi output, not data.table...

sessionInfo() on a successful machine:

R version 3.4.1 (2017-06-30)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 16.04.5 LTS

Matrix products: default
BLAS: /opt/conda/lib/R/lib/libRblas.so
LAPACK: /opt/conda/lib/R/lib/libRlapack.so

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
 [9] LC_ADDRESS=C               LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] data.table_1.11.9

loaded via a namespace (and not attached):
[1] compiler_3.4.1 tools_3.4.1

[jangorecki]

yes,

strtoi("", base=2L)
#[1] 0

[jangorecki]

try pasting this code into strtol.c

#include <stdio.h>
#include <stdlib.h>

int main () {
   char str[5] = "";
   char *ptr;
   int ans = strtol(str, &ptr, 2);
   printf("ans: %d\n", ans);
   printf("leftover: %s\n", ptr);
   return(0);
}

and

gcc strtol.c -o strtol.out
./strtol.out
#ans: 0
#leftover: 

[MichaelChirico]

Hmm I get ans 0 from the C program...

Maybe time to raise a bug on R-devel... at very least the documentation needs to be clarified...

[jangorecki]

Values which cannot be interpreted as integers or would overflow are returned as NA_integer_.

doc is correct, the problem is lack of consistency to C lib

[jangorecki]

anyway we can ignore that and just add extra branch for !nzchar

[MichaelChirico]

Yes sounds like the right solution. And I'll follow up with r-devel separately.

[mattdowle]

I'm not fully following. It sounds like some sort of change is needed to data.table in this 1.12.0 to ensure tests pass consistently on all platforms?

[MichaelChirico]

@mattdowle yes, a very simple fix I think, testing now

[MichaelChirico]

r-devel thread for the record:

http://r.789695.n4.nabble.com/strtoi-output-of-empty-string-inconsistent-across-platforms-td4754736.html