Support bigint

[bbimber]

This is related to this ticket: https://www.labkey.org/ONPRC/Support%20Tickets/issues-details.view?issueId=54682

I dont recall how LabKey does testing on RLabkey; however, it would be a good idea to add a test case with some kind of large integer (e.g., 3.157524392E9) and SelectRows.

[bbimber]

@cnathe: I think the current changes should pass for you. the query API tests are working locally for me.

[cnathe]

@cnathe: I think the current changes should pass for you. the query API tests are working locally for me.

Yes, that works much better. I think this approach to "only use bigint when needed" makes sense for this first pass. Thanks.

[cnathe]

I'm going to get this merges so that I can create a new Rlabkey version with this change and the other change that was just merged to develop last week.

[bbimber]

Hey @cnathe: Having used this a little more, I want to point out some suboptimal behaviors in R. I dont think there is a perfect solution, since R doesnt suport 64-bit integers well (as far as I can see).

Take this simple example:

library(bit64)

int_vec <- as.integer64(c(1,2,3,4))

# This loop returns the following, rather than 1-4:
#[1] 4.940656e-324
#[1] 9.881313e-324
#[1] 1.482197e-323
#[1] 1.976263e-323
for (y in int_vec) {
  print(y)
}

# This returns the correct values
lapply(int_vec, function(x) {
  print(x)
})

There doesnt seem to be a good solution for this problem. Because the final version of this PR only uses integer64 when necessary, the impact of this problem is scoped only to situations that would have failed in the prior version anyway. Without integer64, all large integers would be reported as NAs, so neither approach is seamless.

I'm not advocating changes now, but I think if this becomes a problem, we could drop bit64::integer64 and use R's numeric() class:

labkey-api-r/Rlabkey/R/makeDF.R

Line 22 in d3e44f2

return(bit64::as.integer64(x))