My python project runs on a conda virtual environment. I install packages on the environment with conda install
whenever that package is available for installing that way, otherwise, I install it with pip install
.
To make the project installation easier for other developers, I export the list of packages that are used on the conda environment to a file: conda env export > conda_environment_export.yml
.
However, since a while, the package export does no longer contain the packages installed by pip.
What can I do to make pip packages appear again in the export?
My conda export file as it was some weeks ago:
name: sensor_gateway
channels:
- defaults
dependencies:
- aiohttp=2.3.9=py36_0
- async-timeout=2.0.0=py36hc3e01a3_0
- certifi=2018.1.18=py36_0
- chardet=3.0.4=py36h420ce6e_1
- jinja2=2.10=py36h292fed1_0
- markupsafe=1.0=py36h0e26971_1
- multidict=3.3.2=py36h72bac45_0
- pip=9.0.1=py36h226ae91_4
- pymysql=0.7.11=py36hf59f3ba_0
- python=3.6.4=h6538335_1
- pytz=2018.3=py36_0
- setuptools=38.4.0=py36_0
- sqlalchemy=1.2.1=py36hfa6e2cd_0
- vc=14=h0510ff6_3
- vs2015_runtime=14.0.25123=3
- wheel=0.30.0=py36h6c3ec14_1
- wincertstore=0.2=py36h7fe50ca_0
- yarl=0.14.2=py36h27d1bf2_0
- pip:
- aiohttp-jinja2==0.16.0
- aiomysql==0.0.12
- attrs==17.4.0
- idna==2.6
- idna-ssl==1.0.0
- python-mimeparse==1.6.0
prefix: C:\ProgramData\Anaconda3\envs\sensor_gateway
The conda export now:
name: sensor_gateway
channels:
- defaults
dependencies:
- aiohttp=2.3.9=py36_0
- async-timeout=2.0.0=py36hc3e01a3_0
- certifi=2018.4.16=py36_0
- chardet=3.0.4=py36h420ce6e_1
- icc_rt=2017.0.4=h97af966_0
- intel-openmp=2018.0.0=8
- jinja2=2.10=py36h292fed1_0
- markupsafe=1.0=py36h0e26971_1
- mkl=2018.0.2=1
- mkl_fft=1.0.1=py36h452e1ab_0
- mkl_random=1.0.1=py36h9258bd6_0
- multidict=3.3.2=py36h72bac45_0
- numpy=1.14.2=py36h5c71026_1
- pip=9.0.1=py36h226ae91_4
- pymysql=0.7.11=py36hf59f3ba_0
- python=3.6.4=h6538335_1
- pytz=2018.3=py36_0
- setuptools=38.4.0=py36_0
- simplejson=3.14.0=py36hfa6e2cd_0
- sqlalchemy=1.2.1=py36hfa6e2cd_0
- vc=14=h0510ff6_3
- vs2015_runtime=14.0.25123=3
- wheel=0.30.0=py36h6c3ec14_1
- wincertstore=0.2=py36h7fe50ca_0
- yarl=0.14.2=py36h27d1bf2_0
prefix: C:\ProgramData\Anaconda3\envs\sensor_gateway
I suspect that updating the version of conda (4.4.10-py36_0
--> 4.5.2-py36_0
) may have caused this problem on this project. On another project, I have the same issue, but there I updated pip (9.0.1-py36h226ae91_4
--> 10.0.1-py36_0
). Even after going back to pip 9.0.1 there, the problem remains. Is my suspicion correct? And what can I do to fix this again?
UPDATE:
The only proposed answer right now links to this question and it does not seem to solve my problem:
1. On a Windows command terminal, opened as administrator:
where pip
output (where
as windows-equivalent command to which
on unix/linux):
C:\ProgramData\Anaconda3\Scripts\pip.exe
C:\Users\Sander\AppData\Local\Programs\Python\Python36\Scripts\pip.exe
pip freeze
output:
Lists all packages I ever installed on my 'normal' python (= everything by default installed, not inside any dedicated virtual environments). The package list is too long to include here.
2. On a Windows command terminal, opened with administrator privileges, and after executing activate sensor_gateway
:
where pip
output:
C:\ProgramData\Anaconda3\envs\sensor_gateway\Scripts\pip.exe
C:\ProgramData\Anaconda3\Scripts\pip.exe
C:\Users\Sander\AppData\Local\Programs\Python\Python36\Scripts\pip.exe
pip freeze
output:
lists all packages installed with pip on my environment correctly.
So it seems that when my virtual environment is activated, there is a pip installed in that environment, and it must be that pip that returns the list of installed packages when asked for, not another pip install, since the pip package list on the environment is correct.
You can install pip in the current conda environment with the command conda install pip , as discussed in Using pip in an environment. If there are instances of pip installed both inside and outside the current conda environment, the instance of pip installed inside the current conda environment is used.
Use pip installed with conda , e.g. ~/anaconda/bin/pip . Use it to install packages into a conda environment, as well as to see the union of packages installed with this pip and with conda install .
Conda analyzes each package for compatible dependencies, and how to install them without conflict. If there is a conflict, Conda will let you know that the installation cannot be completed. By comparison, Pip installs all package dependencies regardless of whether they conflict with other packages already installed.
in terminal, type : conda list to obtain the packages installed using conda. There may be some overlap of these lists as pip may recognize packages installed by conda (but maybe not the other way around, IDK). There is a useful source here, including how to update or upgrade packages..
I was able to solve this problem just by running this in a command window with admin privileges:
activate <env-name>
conda update pip
After that, exports of the available packages on my environment contain again both the packages installed with conda and those with pip.
I suspect I may have caused this package export problem because I had followed the instruction that pip gave earlier on, when I was executing a package install:
"You are using pip version 9.0.3, however version 10.0.1 is available. You should consider upgrading via the 'python -m pip install --upgrade pip' > command."
So updating with python -m pip install --upgrade pip
broke my package exports, but conda update pip
fixed it.
For me this only appeared using the conda package pip 18.1_py36_0 (didnt try any other versions). So i downgraded pip in the virtual environment to version 10 via
conda install pip=10
After this the export was listing the pip packages again.
Also see this pull request for conda https://github.com/conda/conda/pull/7612 which fixes the issue for conda 4.6.2. It also was backported to conda 4.5.10. Currently there is no conda 4.4.x backport.
At the moment of writing this post, because of an issue with reading pip lists by anaconda, conda export command only exports packages installed via conda install packagename
and ignores the pip ones.
You can try reinstalling your packages via conda install
and see if it generates the environment.yml file correctly.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With