Thanks Laurent, that makes sense. I noticed someone reported this problem earlier (I probably should have looked there first!), but didn't follow it up so their issue got closed. I reopened it, hope that's ok. See issue #47 rpy2 -> numpy mangles arrays.
Thanks again, Robert -----Original Message----- From: Laurent Gautier [mailto:lgaut...@gmail.com] Sent: Tuesday, 2 November 2010 4:49 PM To: Denham Robert Cc: RPy help, support and design discussion list Subject: Re: [Rpy] converting r array to numpy array This is a bug then. The problem is in the way strides are computed (buffer.c in rpy/rinterface), and a fix seems to be: static void sexp_strides(SEXP sexp, Py_intptr_t *strides, Py_ssize_t itemsize, Py_intptr_t *shape, int nd) { /* Set the buffer 'strides', that is a vector or Py_intptr_t * containing the offset (in bytes) when progressing along * each dimension. */ int i; Py_intptr_t cumul_strides = 1; strides[0] = itemsize; for (i = 1; i < nd; i++) { strides[i] = shape[i-1] * strides[i-1]; } } I just committed that change to the branch 2.2.x. but I'd like to have this tested a little before porting it back to 2.1.x. Can you fill one on the bitbucket tracker ? (so others know about the issue) Thanks, L. On 01/11/10 22:36, Denham Robert wrote: > Yeah, sorry, this is the output: > In [3]: import numpy > In [4]: import rpy2 > In [5]: import rpy2.robjects as robjects In [7]: marr = > robjects.r.array(robjects.IntVector(range(54)),dim= > robjects.IntVector([9,2,3])) > > In [8]: marrpy = numpy.array(marr) > In [9]: print marrpy.flatten() > ------> print(marrpy.flatten()) > [ 0 2 4 9 11 13 1 3 5 10 12 14 2 4 6 11 13 15 3 5 7 12 14 > 16 > 4 > 6 8 13 15 17 5 7 9 14 16 18 6 8 10 15 17 19 7 9 11 16 18 20 > 8 10 > 12 17 19 21] > > In [10]: print robjects.r['as.vector'](marr) > -------> print(robjects.r['as.vector'](marr)) > [1] 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 > 21 > 22 23 24 > [26] 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 > 47 48 49 > [51] 50 51 52 53 > > > In [15]: print marrpy.flatten('F') > -------> print(marrpy.flatten('F')) > [ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 2 3 4 5 6 > 7 > 8 > 9 10 11 12 13 14 15 16 17 18 19 4 5 6 7 8 9 10 11 12 13 14 15 > 16 > 17 > 18 19 20 21] > > > I was prepared for differences due to column/row ordering, but this > doesn't seem to be the problem. > > Works fine on matrices: > > In [11]: mat = robjects.r.matrix(robjects.IntVector(range(54)),nrow=9) > In [12]: matpy = numpy.array(mat) > In [13]: print mat > -------> print(mat) > [,1] [,2] [,3] [,4] [,5] [,6] > [1,] 0 9 18 27 36 45 > [2,] 1 10 19 28 37 46 > [3,] 2 11 20 29 38 47 > [4,] 3 12 21 30 39 48 > [5,] 4 13 22 31 40 49 > [6,] 5 14 23 32 41 50 > [7,] 6 15 24 33 42 51 > [8,] 7 16 25 34 43 52 > [9,] 8 17 26 35 44 53 > > > In [14]: print matpy > -------> print(matpy) > [[ 0 9 18 27 36 45] > [ 1 10 19 28 37 46] > [ 2 11 20 29 38 47] > [ 3 12 21 30 39 48] > [ 4 13 22 31 40 49] > [ 5 14 23 32 41 50] > [ 6 15 24 33 42 51] > [ 7 16 25 34 43 52] > [ 8 17 26 35 44 53]] > > > > Regards, Robert > > > > -----Original Message----- > From: Laurent Gautier [mailto:lgaut...@gmail.com] > Sent: Monday, 1 November 2010 4:58 PM > To: Denham Robert > Cc: RPy help, support and design discussion list > Subject: Re: [Rpy] converting r array to numpy array > > On 01/11/10 07:28, Denham Robert wrote: >> I got caught out converting a 3d array in R to a numpy array. I did >> something like: >> >> import numpy >> import rpy2 >> import rpy2.robjects as robjects >> >> marr = robjects.r.array(robjects.IntVector(range(54)),dim= >> robjects.IntVector([9,2,3])) >> marrpy = numpy.array(marr) >> print marrpy.flatten() >> print robjects.r['as.vector'](marr) >> >> Which clearly doesn't work (at least for me, using rpy2.1.7, R-10.0 >> on 64bit suse linux). > Could you be more specific about "does not work" ? > If you are after getting the same output for the two print calls, this > is not happening because arrays are being stored "column-major" by R > by while numpy has "row-major" as a default. > Try: > > print marrpy.flatten('F') > > > >> The matrix version does work, eg >> >> mat = robjects.r.matrix(robjects.IntVector(range(54)),nrow=9) >> matpy = numpy.array(mat) >> print mat >> print matpy >> >> I checked the documentation, and found examples for vectors and >> matrices, but not for arrays, but also no warnings. So, I was just >> wondering, is this supposed to work? If it is, a hint on why it >> doesn't work for me would be good. If it isn't supposed to work, I >> was wondering if it were possible to throw an error or something. > > It is supposed to work, but not claim about an absence of bugs is made. > > > Best, > > > Laurent > > >> Regards, >> Robert >> >> >> >> Robert Denham >> Environmental Statistician, Remote Sensing Centre Environment and >> Resource Sciences Phone 07 3896 9899 (ext 69899) >> Email: robert.den...@derm.qld.gov.au >> www.derm.qld.gov.au<http://www.derm.qld.gov.au/> >> >> Department of Environment and Resource Management QCCE Building, 80 >> Meiers Rd, Indooroopilly QLD 4068 >> >> >> +----------------------------------------------------------------+ >> Think B4U Print >> 1 ream of paper = 6% of a tree and 5.4kg CO2 in the atmosphere >> 3 sheets of A4 paper = 1 litre of water >> +----------------------------------------------------------------+ >> >> >> --------------------------------------------------------------------- >> - >> -------- Nokia and AT&T present the 2010 Calling All Innovators-North >> America contest Create new apps& games for the Nokia N8 for consumers >> in U.S. and Canada $10 million total in prizes - $4M cash, 500 >> devices, nearly $6M in marketing Develop with Nokia Qt SDK, Web >> Runtime, or Java and Publish to Ovi Store >> http://p.sf.net/sfu/nokia-dev2dev >> _______________________________________________ >> rpy-list mailing list >> rpy-list@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/rpy-list ------------------------------------------------------------------------------ Achieve Improved Network Security with IP and DNS Reputation. Defend against bad network traffic, including botnets, malware, phishing sites, and compromised hosts - saving your company time, money, and embarrassment. Learn More! http://p.sf.net/sfu/hpdev2dev-nov _______________________________________________ rpy-list mailing list rpy-list@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rpy-list
