在Python中使用rpy2包：用户代码未处理Rruntime错误

Question

I'm trying to create an r environment in python using the rpy2 package. I'm trying to create a GLRLM of a ROI I have extracted from an image. I had some problems getting the package to load, but I finally managed to do so. Now I get the following error when running the code:

RuntimeError was unhandled by user code

At this line:

ro.r('glrlmatrix <- radiomics:::glrlm(tissue)')

I'm suspecting it cannot find the function "glrlm". I tried running the code in R and it works fine. Anyone has an idea of what to do?

tissue = pd.DataFrame(rgbRoi[...,0])
rdf= pandas2ri.py2ri(tissue)
ro.globalenv['tissue'] = rdf
radiomics=importr("radiomics",lib_loc="C:/Users/nka/Documents/R/win-library/3.3");
ro.r('tissue <- as.matrix(tissue)')
ro.r('print(dim(tissue))')
ro.r('library(radiomics)')
ro.r('glrlmatrix <- radiomics:::glrlm(tissue)')
ro.r('glrlmatrix[0,]    <- 0')                           ### Assign zero value to first row which belongs to mask region
ro.r('glrlfeature       <- array(NA,dim=c(11,1))')
ro.r('glrlfeature[1,1]  <- radiomics:::glrlm_GLN(glrlmatrix)')
ro.r('glrlfeature[2,1]  <- radiomics:::glrlm_HGLRE(glrlmatrix)')
ro.r('glrlfeature[3,1]  <- radiomics:::glrlm_LRE(glrlmatrix)')
ro.r('glrlfeature[4,1]  <- radiomics:::glrlm_LRHGLE(glrlmatrix)')
ro.r('glrlfeature[5,1]  <- radiomics:::glrlm_LRLGLE(glrlmatrix)')
ro.r('glrlfeature[6,1]  <- radiomics:::glrlm_LGLRE(glrlmatrix)')
ro.r('glrlfeature[7,1]  <- radiomics:::glrlm_RLN(glrlmatrix)')
ro.r('glrlfeature[8,1]  <- radiomics:::glrlm_RP(glrlmatrix)')
ro.r('glrlfeature[9,1]  <- radiomics:::glrlm_SRE(glrlmatrix)')
ro.r('glrlfeature[10,1] <- radiomics:::glrlm_SRHGLE(glrlmatrix)')
ro.r('glrlfeature[11,1] <- radiomics:::glrlm_SRLGLE(glrlmatrix)')
glr = ro.r.matrix(ro.r('glrlfeature'))
glr = np.array(glr)

Answer 1

When doing radiomics:::glrlm you are accessing the symbol glrlm in the namespace radiomics while it is not exported (use :: to only access exported symbols). This should be working in R and when parsing/evaluating an R string with ro.r . However, symbols are generally not exported by an R package because they not meant to be accessed directly by regular users.

Also The message RuntimeError was unhandled by user code does not seem to originate from rpy2. Is this really all there is to report about the error ?

Beside this, you may considering migrating some of the R strings to Python (if you think that this is helping with legibility):

# this is implicitly loading the R package
radiomics=importr("radiomics",
                  lib_loc="C:/Users/nka/Documents/R/win-library/3.3");
# the line below is then unnecessary because of the above
# ro.r('library(radiomics)')

# binding the symbol "tissue" in R's globalenv to the matrix conversion
# of the data frame of the same name can be written more like Python
# code:
base = importr("base")
tissue_mat = base.as_matrix(tissue)
ro.globalenv['tissue'] = tissue_mat
# the following should work, unless the R code in `radiomics::glrlm`
# cannot handle anonymous objects
glrlmatrix = radiomics.glrlm(tissue_mat)
# glrmatrix is a Python/rpy2 objects that also be used as a numpy array
# and modified in-place
# (see http://rpy2.readthedocs.io/en/version_2.8.x/numpy.html#from-rpy2-to-numpy)