CODEWITHSUNDEEP
pythonrrpy2
Stereo
Stereo

Reputation: 1193

Convert Python binary data to R using rpy2

My goal is to get bytes data from Python to R to unserialize in R. The following code provides the test

import rpy2.robjects as ro
rcode = 'serialize(iris, NULL)'
r_res = ro.r(rcode)
print(type(r_res[0]))
# <class 'bytes'>

# Works up to here, not sure what how to get the 'bytes' type back into R

# Got 24 from the Rinternals.h file where it indicates RAWSXP
rawsxp_rinternals = 24
r_vec = ro.SexpVector(r_res[0], rawsxp_rinternals)

This yields the following error:

Error while converting to Bytes element 0.

Ideally I would want to achieve the following

  1. get the raw data back into R
  2. unserialize the data

Upvotes: 0

Views: 237

Answers (2)

Yakov Goldberg
Yakov Goldberg

Reputation: 21

I found the following works for me:

R code:

library(stringi)

foo <- function(binary_data) {
    typeof(binary_data) # raw
    # to decode use rawToChar if encoding is utf-8
    # or stri_conv(binary_data, "from_encoding", "to_encoding"), from the lib stringi
    stri_conv(binary_data, "utf8") # "my text"
}

Python code:

import rpy2.robjects as ro

text = "my text"
binary = text.encode("utf8")

r_raw_vector = ro.rinterface.ByteSexpVector(binary)
ro.r.foo(data=r_raw_vector)

Upvotes: 1

lgautier
lgautier

Reputation: 11545

R's serialize() is returning a list of byte vectors. This is the input expected by unserialize(). The following will "just work":

ro.r('unserialize')(r_res)

Otherwise, building an rpy2 Vector (for an R RAWSXP vector) can be achieved like for other vectors: 

>>> ro.rinterface.str_typeint(r_res.typeof)
'RAWSXP'
>>> r_res2 = ro.vectors.Vector(r_res)
>>> ro.rinterface.str_typeint(r_res2.typeof)
'RAWSXP'
>>> r_res3 = ro.vectors.Vector([r_res[0]])
>>> ro.rinterface.str_typeint(r_res3.typeof)
'RAWSXP'

Upvotes: 1

Related Questions