allow.cartesian when i has no duplicate values

[nigmastar]

Please consider the following:

> dt <- data.table(id=rep(letters[1:2], 2), var = rnorm(4), key="id")
> dt
   id       var
1:  a 0.9609685
2:  a 0.1432707
3:  b 1.1276582
4:  b 0.8051821

> dt[letters[1:3], list(var)]
Error in vecseq(f__, len__, if (allow.cartesian) NULL else as.integer(max(nrow(x),  : 
  Join results in 5 rows; more than 4 = max(nrow(x),nrow(i)). Check for duplicate key values in i, each of which join to the same group in x over and over again. [...]

> dt[letters[1:3], list(var), by=.EACHI]
   id       var
1:  a 0.9609685
2:  a 0.1432707
3:  b 1.1276582
4:  b 0.8051821
5:  c        NA

The second join results in 5 rows too, shouldn't both joins above be consistent? (Maybe both like the second)

I also wander, the concept behind the implementation of allow.cartesian is simply 1) output rows has not to be more than max(nrow(x),nrow(i)) or 2) to avoid duplicates in key values of i?

> sessionInfo()
R version 3.1.1 (2014-07-10)
Platform: x86_64-w64-mingw32/x64 (64-bit)
# data.table installed today from github

[arunsrinivasan]

Thanks @nigmastar. We'll take a look at this asap.

Regarding your last question: I think this SO post should help.

[nigmastar]

Hi Arun,

thanks for the link. It helped. So, (besides what you say in the link about the error only for 'huge' results) don't you think that when i has no duplicates, like above, no error should be thrown?

Thanks,

Michele.

[arunsrinivasan]

There've been quite a few issues filed on this topic now. Without giving too much thought, it does appear to me that i having no duplicates need not provide an error.

Bumping all allow.cartesian=. issues to 1.9.4 milestone.

[mgriebe]

Maybe the error can mention "nomatch=0" as a possible solution.