[Rd] Inconsistent results from .C()

[Robin Evans]

Hello,

I've run into a problem which is both maddening and rather hard to
replicate, therefore I wondered if someone might know of a plausible
explanation for it.  I couldn't find anything in the archives, though
maybe I'm searching for the wrong thing.

I'm calling some C code using .C, and get the vector I'm interested in
back as the 7th location in the returned list.  However I find that if
I try to inspect the contents of this entry in the list in some ways,
I get one answer, and if I look at it in others I get a different
answer.  It's quite possible that there's something wrong with the C
code, but this doesn't seem to explain why this phenomenon would occur
in R.

The problem does not always occur - I have to run the code a few times
and then call the console when it does, but the commands below show
what can happen when it does.  I apologise for not being able to get a
cleaner example.  Full code and output is below, but here is a
stylised version:

The following all give one answer (which is the wrong answer as far as
I'm concerned) :
 * printing the whole list :
          .C(...)     # and looking at the 7th entry
 * applying c() to the 7th element of the list
          c(.C(...)[[7]])
 * assigning the 7th element to a vector:
          x = .C(...)[[7]];
          x

these give a different answer (which is the answer I would hope the C
code returns):
 * using dput on the 7th entry:
          dput(.C(...)[[7]])
 * applying c() and then dput()
          dput(c(.C(...)[[7]]))
 * just printing the 7th entry of the list
          .C(...)[[7]]

The answers are consistent in the sense that I always get the same
answers from running the same command in the console.  I have tried
inspecting the returned objects to see if the objects are somehow of a
different class than I expect, or are just being printed oddly, but
have not found anything untoward.

Any suggestions would be much appreciated!

Regards,

Robin


# THESE COMMANDS GIVE ONE ANSWER
# [the correct answer always begins with 1, the incorrect with -1]

> .C("oneMargin", c(1L,1L,1L,1L,1L), c(1L,1L,1L,1L,1L), 1L, c(2L,2L,2L,2L,2L), 5L, ptr1, ptr2)[[7]]
[1]  1 -1 -1  1 -1  1  1 -1 -1  1  1 -1  1 -1 -1  1 -1  1  1 -1  1 -1
-1  1  1 -1 -1  1 -1  1  1 -1

> dput(.C("oneMargin", c(1L,1L,1L,1L,1L), c(1L,1L,1L,1L,1L), 1L, c(2L,2L,2L,2L,2L), 5L, ptr1, ptr2)[[7]])
c(1L, -1L, -1L, 1L, -1L, 1L, 1L, -1L, -1L, 1L, 1L, -1L, 1L, -1L,
-1L, 1L, -1L, 1L, 1L, -1L, 1L, -1L, -1L, 1L, 1L, -1L, -1L, 1L,
-1L, 1L, 1L, -1L)

> x=dput(c(.C("oneMargin", c(1L,1L,1L,1L,1L), c(1L,1L,1L,1L,1L), 1L, c(2L,2L,2L,2L,2L), 5L, ptr1, ptr2)[[7]]))
c(1L, -1L, -1L, 1L, -1L, 1L, 1L, -1L, -1L, 1L, 1L, -1L, 1L, -1L,
  -1L, 1L, -1L, 1L, 1L, -1L, 1L, -1L, -1L, 1L, 1L, -1L, -1L, 1L,
  -1L, 1L, 1L, -1L)
> x
[1]  1 -1 -1  1 -1  1  1 -1 -1  1  1 -1  1 -1 -1  1 -1  1  1 -1  1 -1
-1  1  1 -1 -1  1 -1  1  1 -1

# THESE ALL GIVE A DIFFERENT ONE!

> .C("oneMargin", c(1L,1L,1L,1L,1L), c(1L,1L,1L,1L,1L), 1L, c(2L,2L,2L,2L,2L), 5L, ptr1, ptr2)

# (OTHER ELEMENTS OF LIST REMOVED)
[[7]]
[1] -1 -1  1  1  1  1 -1 -1  1  1 -1 -1 -1 -1  1  1  1  1 -1 -1 -1 -1
1  1 -1 -1  1  1  1  1 -1 -1

> c(.C("oneMargin", c(1L,1L,1L,1L,1L), c(1L,1L,1L,1L,1L), 1L, c(2L,2L,2L,2L,2L), 5L, ptr1, ptr2)[[7]])
[1] -1 -1  1  1  1  1 -1 -1  1  1 -1 -1 -1 -1  1  1  1  1 -1 -1 -1 -1
1  1 -1 -1  1  1  1  1 -1 -1
> x = .C("oneMargin", c(1L,1L,1L,1L,1L), c(1L,1L,1L,1L,1L), 1L, c(2L,2L,2L,2L,2L), 5L, ptr1, ptr2)[[7]]
> x
[1] -1 -1  1  1  1  1 -1 -1  1  1 -1 -1 -1 -1  1  1  1  1 -1 -1 -1 -1
1  1 -1 -1  1  1  1  1 -1 -1


--
Robin Evans
Statistical Laboratory
University of Cambridge
blog: itsastatlife.blogspot.com
web: www.statslab.cam.ac.uk/~rje42

Causal Inference Workshop July 15th:
http://www.statslab.cam.ac.uk/~rje42/uai13/main.htm