How to convert R dataframe back to Pandas using rpy2?

Question

I converted a pandas dataframe to R using the code below:

import pandas as pd
import pandas.rpy.common as com
import rpy2.robjects as ro
from rpy2.robjects.packages import importr

rdf = com.convert_to_r_dataframe(df)

How do I convert rdf back to a pandas.DataFrame?

df = f(rdf)

`pandas.rpy` was removed in pandas 0.20 – Franck Dernoncourt Jul 20 '17 at 19:10 — Franck Dernoncourt, Jul 20 '17 at 19:10

score 31 · Answer 1 · edited Dec 30 '19 at 06:02

31

Since rpy2 release 2.4.0 converting data frames back and forth between rpy2 and pandas is included as an optional module. With it, no need to convert explicitly, it will be done on the fly.

The documentation contains examples (also available as a Jupyter notebook - link available near the top of the page): https://rpy2.github.io/doc/latest/html/pandas.html#interoperability-with-pandas

Note: The original answer to this question recommended the following.

from rpy2.robjects import pandas2ri
pandas2ri.activate()

If wishing to convert explicitly for any reason, the functions are pandas2ri.py2ri() and pandas2ri.ri2py() (they were pandas2ri.pandas2ri() and pandas2ri.ri2pandas()).

Note: Since rpy2 release 3.3.0 explicit conversion is done as follows

import rpy2.robjects as ro

dt = pd.DataFrame()
# To R DataFrame
r_dt = ro.conversion.py2rpy(dt)
# To pandas DataFrame
pd_dt = ro.conversion.rpy2py(r_dt)

For more details check out this link.

edited Dec 30 '19 at 06:02

Miguel Trejo

5,913
5
24
49

answered Dec 27 '13 at 22:06

lgautier

11,363
29
42

I am a noobie to Python. When I use the above functions, I lose my categorical objects and they become integers. – Shintu Joseph Apr 14 '18 at 17:17
Any ideas on changing the behaviour? @Igautier – Shintu Joseph Apr 14 '18 at 19:03
1

Which version are you using ? This answer is a bit old (4.5 years), and there were changes in the way factors are converted: https://bitbucket.org/rpy2/rpy2/commits/29e8a62de897 – lgautier May 06 '18 at 16:26
Yes Igautier, I got them working. Thanks for replying. – Shintu Joseph May 09 '18 at 03:14
@lgautier are you aware of the way we should do it now? The answer is a bit old – Perl Del Rey Feb 16 '20 at 18:08
1

@Hiyam the doc pointed by the link is few days old. It is not too old to be used. – lgautier Feb 17 '20 at 02:19

score 12 · Answer 2 · edited Jun 22 '16 at 18:08

12

As suggested by lgautier, it can be done with pandas2ri.

Here is sample code for convert rpy dataframe (rdf) to pandas dataframe (pd_df):

from rpy2.robjects import pandas2ri

pd_df = pandas2ri.ri2py_dataframe(rdf)

edited Jun 22 '16 at 18:08

tbekolay

17,201
3
40
38

answered Feb 19 '15 at 16:43

huojun

177
1
3

This operation replaces the index of the R dataframe with integers (if it was something other than integers to begin with). Anyone know how to keep the original index? – A. Slowey Feb 01 '19 at 17:00

score 9 · Answer 3 · answered Dec 17 '13 at 16:16

Given your import, it appears it is:

com.convert_robj(rdf)

For example,

In [480]: dfrm
Out[480]:
           A          B  C
0   0.454459  49.916767  1
1   0.943284  50.878174  1
2   0.974856  50.335679  2
3   0.776600  50.782104  1
4   0.553895  50.084505  1
5   0.514018  50.719019  2
6   0.915413  50.513962  0
7   0.771571  49.859855  2
8   0.068619  49.409657  0
9   0.728141  50.945174  2
10  0.388115  47.879653  1
11  0.960172  49.680258  0
12  0.015216  50.067968  0
13  0.495024  50.286287  1
14  0.565954  49.909771  1
15  0.992279  49.009696  1
16  0.179934  49.554256  0
17  0.521243  47.854791  0
18  0.551241  51.076262  1
19  0.713271  49.418503  0
20  0.801716  50.660304  1

In [481]: rdfrm = com.convert_to_r_dataframe(dfrm)

In [482]: rdfrm
Out[482]:
<DataFrame - Python:0x14905cf8 / R:0x1600ee98>
[FloatVector, FloatVector, IntVector]
  A: <class 'rpy2.robjects.vectors.FloatVector'>
  <FloatVector - Python:0xf9d0b00 / R:0x140e2620>
[0.454459, 0.943284, 0.974856, ..., 0.551241, 0.713271, 0.801716]
  B: <class 'rpy2.robjects.vectors.FloatVector'>
  <FloatVector - Python:0xf9d0878 / R:0x125aa240>
[49.916767, 50.878174, 50.335679, ..., 51.076262, 49.418503, 50.660304]
  C: <class 'rpy2.robjects.vectors.IntVector'>
  <IntVector - Python:0x11fceef0 / R:0x13f0d918>
[       1,        1,        2, ...,        1,        0,        1]

In [483]: com.convert_robj(rdfrm)
Out[483]:
           A          B  C
0   0.454459  49.916767  1
1   0.943284  50.878174  1
2   0.974856  50.335679  2
3   0.776600  50.782104  1
4   0.553895  50.084505  1
5   0.514018  50.719019  2
6   0.915413  50.513962  0
7   0.771571  49.859855  2
8   0.068619  49.409657  0
9   0.728141  50.945174  2
10  0.388115  47.879653  1
11  0.960172  49.680258  0
12  0.015216  50.067968  0
13  0.495024  50.286287  1
14  0.565954  49.909771  1
15  0.992279  49.009696  1
16  0.179934  49.554256  0
17  0.521243  47.854791  0
18  0.551241  51.076262  1
19  0.713271  49.418503  0
20  0.801716  50.660304  1

With docs:

In [475]: com.convert_robj?
Type:       function
String Form:<function convert_robj at 0x13e85848>
File:       /mnt/epd/7.3-2_pandas0.12/lib/python2.7/site-packages/pandas/rpy/common.py
Definition: com.convert_robj(obj, use_pandas=True)
Docstring:
Convert rpy2 object to a pandas-friendly form

Parameters
----------
obj : rpy2 object

Returns
-------
Non-rpy data structure, mix of NumPy and pandas objects

Rock Pereira · Answer 4 · 2019-05-26T05:48:23.953

6

Use pandas to read an rpy2 dataframe, r_df. It will avoid the deprecation warning "FutureWarning: from_items is deprecated. Use DataFrame.from_dict(dict(items), ...) instead"

type(r_df) is "rpy2.robjects.vectors.DataFrame".
type(pd_df) is "pandas.core.frame.DataFrame"

pd_df = pd.DataFrame.from_dict({ key : np.asarray(r_df.rx2(key)) for key in r_df.names })

edited May 26 '19 at 05:48

answered May 26 '19 at 05:35

Rock Pereira

471
1
4
12

Above solutions are outdated, but this seemed to work great for latest 3.2.0 version of rpy2. – COOLBEANS Oct 02 '19 at 19:08
preise and works fine with old and new verisions – alex3465 Aug 27 '21 at 15:06
This is the only method that worked for me. – Alex Vorobiev Mar 17 '23 at 21:49

score 1 · Answer 5 · edited Feb 13 '23 at 10:43

1

The other solutions seem to be outdated and did not work for me anymore.

From the docs, this is the current way to convert data from/to pandas to/from R objects.

import rpy2.robjects as ro
from rpy2.robjects import pandas2ri

From pandas to R:

with ro.default_converter + pandas2ri.converter:
  r_from_pd_df = ro.conversion.get_conversion().py2rpy(pd_df)

r_from_pd_df

From R to pandas:

with ro.default_converter + pandas2ri.converter:
  pd_from_r_df = ro.conversion.get_conversion().rpy2py(r_df)

pd_from_r_df

This only works in rpy2 version >=3.5.7.

edited Feb 13 '23 at 10:43

shs

3,683
1
6
34

answered Jan 11 '23 at 09:11

Martin Niederl

649
14
32

The last example gives me `AttributeError: __enter__` with rpy2 3.5.9. – Alex Vorobiev Mar 17 '23 at 00:48

How to convert R dataframe back to Pandas using rpy2?

5 Answers5

Linked