[Rd] Issue when calling deparse(substitute(x)) from C with "anonymous" R vectors ?

[Laurent Gautier]

Dear list,


When calling R from C, what appears like a spurious error can be 
triggered during the execution of chisq.test(x, y).

This is happening when the following conditions are met:

- x and y are "anonymous" C-level R vectors (they do not have a symbol),
but they are protected from garbage collection

- x and y are "not too small" (it was experienced as soon as they are 
longer than 17 elements).


The error is

Error in names(dimnames(x)) <- DNAME :
   'names' attribute [4] must be the same length as the vector [2]

and can be traced to the use of deparse(substitute(x)) and 
deparse(substitute(y)) in the R code.

What seems to be happening is that the deparse(substitute(x)) call
gives a character vector of length > 1 as soon as x is "not so small".

To demonstrate this, I am using rpy2 (as the problem was found when using it
http://sourceforge.net/mailarchive/forum.php?thread_name=4B043CA1.9050500%40salilab.org&forum_name=rpy-list
), but it will likely be present in other bridges to R as well.


#using R-2.10 and rpy2-2.1.dev
 >>> import rpy2.robjects as robjects

 >>> f = robjects.r('''function(x) return(deparse(substitute(x)))''')

 >>> tuple(f(robjects.FloatVector(range(17))))
('c(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)',)
# length 1
 >>> tuple(f(robjects.FloatVector(range(18))))
('c(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17', ')')
# length 2 !!


Does it seem to others like an issue present in other bridges as well ?



L.