Closed Jonathanyu2014 closed 4 years ago
I was unaware of this problem as continuous integration (CI) hasn't yet indicated this error. My CI testing suggests rpy2 version 2.9.4 should be fine. My guess is that api changes from rpy2 version 2 to version 3 broke the code.
Thanks for reply. I install rpy2 v2.9.4, however, there still was something wrong. I used 2020plus v1.2.3. The command line is "python 2020plus.py --out-dir=result_compare classify -f test.features.txt" . The error is as follows:
Running Random forest . . .
AN ERROR HAS OCCURRED: check the log file
Type: <class 'ValueError'>
Exception: Buffer for this type not yet supported.
Traceback:
File "2020plus.py", line 275, in
The python modules I used were : numpy==1.17.2 pandas==0.25.2 scipy==1.3.1 scikit-learn==0.19.2 matplotlib==3.1.1 PyYAML==5.1.2 pysam==0.15.3 rpy2==2.9.4 tzlocal==2.0.0 probabilistic2020==1.2.3 snakemake==5.7.4
This is actually a bug in rpy2 itself (see https://github.com/conda-forge/rpy2-feedstock/issues/28). Apparently, rpy2 2.9.x does not work with numpy 1.17.
Here are some version numbers that worked in a clean install environment: scipy-0.19.1 pandas-0.25.2 python-3.6.9 matplotlib-3.1.1 numpy-1.15.4 pysam-0.15.3
Many thanks! Finally it worked for me. I reinstalled scipy-0.19.1 and numpy-1.15.4.
When I ran 'python 2020plus.py classify', the error showed as follows:
Type: <class 'AttributeError'> Exception: module 'rpy2.robjects.pandas2ri' has no attribute 'py2ri'
I used rpy2 v3.1.0. I suggested the author specified a rpy2 version that could solve this problem. Thanks.