How to uninstall a package installed with pip install –user
Question:
There is a --user option for pip which can install a Python package per user:
pip install --user [python-package-name]
I used this option to install a package on a server for which I do not have root access. What I need now is to uninstall the installed package on the current user. I tried to execute this command:
pip uninstall --user [python-package-name]
But I got:
no such option: --user
How can I uninstall a package that I installed with pip install --user, other than manually finding and deleting the package?
I’ve found this article
pip cannot uninstall from per-user site-packages directory
which describes that uninstalling packages from user directory does not supported. According to the article if it was implemented correctly then with
pip uninstall [package-name]
the package that was installed will be also searched in user directories. But a problem still remains for me. What if the same package was installed both system-wide and per-user?
What if someone needs to target a specific user directory?
Answers:
Having tested this using Python 3.5 and pip 7.1.2 on Linux, the situation appears to be this:
-
pip install --user somepackage installs to $HOME/.local, and uninstalling it does work using pip uninstall somepackage.
-
This is true whether or not somepackage is also installed system-wide at the same time.
-
If the package is installed at both places, only the local one will be uninstalled. To uninstall the package system-wide using pip, first uninstall it locally, then run the same uninstall command again, with root privileges.
-
In addition to the predefined user install directory, pip install --target somedir somepackage will install the package into somedir. There is no way to uninstall a package from such a place using pip. (But there is a somewhat old unmerged pull request on Github that implements pip uninstall --target.)
-
Since the only places pip will ever uninstall from are system-wide and predefined user-local, you need to run pip uninstall as the respective user to uninstall from a given user’s local install directory.
The answer is Not possible yet. You have to remove it manually.
example to uninstall package ‘oauth2client’ on MacOS:
pip uninstall oauth2client
As @thomas-lotze has mentioned, currently pip tooling does not do that as there is no corresponding –user option. But what I find is that I can check in ~/.local/bin and look for the specific pip#.# which looks to me like it corresponds to the –user option.
In my case:
antho@noctil: ~/.l/bin$ pwd
/home/antho/.local/bin
antho@noctil: ~/.l/bin$ ls pip*
pip pip2 pip2.7 pip3 pip3.5
And then just uninstall with the specific pip version.
I am running Anaconda version 4.3.22 and a python3.6.1 environment, and had this problem. Here’s the history and the fix:
pip uninstall opencv-python # -- the original step. failed.
ImportError: DLL load failed: The specified module could not be found.
I did this into my python3.6 environment and got this error.
python -m pip install opencv-python # same package as above.
conda install -c conda-forge opencv # separate install parallel to opencv
pip-install opencv-contrib-python # suggested by another user here. doesn't resolve it.
Next, I tried downloading python3.6 and putting the python3.dll in the folder and in various folders. nothing changed.
finally, this fixed it:
pip uninstall opencv-python
(the other conda-forge version is still installed) This left only the conda version, and that works in 3.6.
>>>import cv2
>>>
working!
Be careful though, for those who using pip install --user some_pkg inside a virtual environment.
$ path/to/python -m venv ~/my_py_venv
$ source ~/my_py_venv/bin/activate
(my_py_venv) $ pip install --user some_pkg
(my_py_venv) $ pip uninstall some_pkg
WARNING: Skipping some_pkg as it is not installed.
(my_py_venv) $ pip list
# Even `pip list` will not properly list the `some_pkg` in this case
In this case, you have to deactivate the current virtual environment, then use the corresponding python/pip executable to list or uninstall the user site packages:
(my_py_venv) $ deactivate
$ path/to/python -m pip list
$ path/to/python -m pip uninstall some_pkg
Note that this issue was reported few years ago. And it seems that the current conclusion is: --user is not valid inside a virtual env’s pip, since a user location doesn’t really make sense for a virtual environment.
I strongly recommend you to use virtual environments for python package installation. With virtualenv, you prevent any package conflict and total isolation from your python related userland commands.
To delete all your package installed globally follow this;
It’s possible to uninstall packages installed with --user flag. This one worked for me;
pip freeze --user | xargs pip uninstall -y
For python 3;
pip3 freeze --user | xargs pip3 uninstall -y
But somehow these commands don’t uninstall setuptools and pip. After those commands (if you really want clean python) you may delete them with;
pip uninstall setuptools && pip uninstall pip
Now you have clean python environment. You can create virtualenv and install the package inside them.
There is a --user option for pip which can install a Python package per user:
pip install --user [python-package-name]
I used this option to install a package on a server for which I do not have root access. What I need now is to uninstall the installed package on the current user. I tried to execute this command:
pip uninstall --user [python-package-name]
But I got:
no such option: --user
How can I uninstall a package that I installed with pip install --user, other than manually finding and deleting the package?
I’ve found this article
pip cannot uninstall from per-user site-packages directory
which describes that uninstalling packages from user directory does not supported. According to the article if it was implemented correctly then with
pip uninstall [package-name]
the package that was installed will be also searched in user directories. But a problem still remains for me. What if the same package was installed both system-wide and per-user?
What if someone needs to target a specific user directory?
Having tested this using Python 3.5 and pip 7.1.2 on Linux, the situation appears to be this:
-
pip install --user somepackageinstalls to$HOME/.local, and uninstalling it does work usingpip uninstall somepackage. -
This is true whether or not
somepackageis also installed system-wide at the same time. -
If the package is installed at both places, only the local one will be uninstalled. To uninstall the package system-wide using
pip, first uninstall it locally, then run the same uninstall command again, withrootprivileges. -
In addition to the predefined user install directory,
pip install --target somedir somepackagewill install the package intosomedir. There is no way to uninstall a package from such a place usingpip. (But there is a somewhat old unmerged pull request on Github that implementspip uninstall --target.) -
Since the only places
pipwill ever uninstall from are system-wide and predefined user-local, you need to runpip uninstallas the respective user to uninstall from a given user’s local install directory.
The answer is Not possible yet. You have to remove it manually.
example to uninstall package ‘oauth2client’ on MacOS:
pip uninstall oauth2client
As @thomas-lotze has mentioned, currently pip tooling does not do that as there is no corresponding –user option. But what I find is that I can check in ~/.local/bin and look for the specific pip#.# which looks to me like it corresponds to the –user option.
In my case:
antho@noctil: ~/.l/bin$ pwd
/home/antho/.local/bin
antho@noctil: ~/.l/bin$ ls pip*
pip pip2 pip2.7 pip3 pip3.5
And then just uninstall with the specific pip version.
I am running Anaconda version 4.3.22 and a python3.6.1 environment, and had this problem. Here’s the history and the fix:
pip uninstall opencv-python # -- the original step. failed.
ImportError: DLL load failed: The specified module could not be found.
I did this into my python3.6 environment and got this error.
python -m pip install opencv-python # same package as above.
conda install -c conda-forge opencv # separate install parallel to opencv
pip-install opencv-contrib-python # suggested by another user here. doesn't resolve it.
Next, I tried downloading python3.6 and putting the python3.dll in the folder and in various folders. nothing changed.
finally, this fixed it:
pip uninstall opencv-python
(the other conda-forge version is still installed) This left only the conda version, and that works in 3.6.
>>>import cv2
>>>
working!
Be careful though, for those who using pip install --user some_pkg inside a virtual environment.
$ path/to/python -m venv ~/my_py_venv
$ source ~/my_py_venv/bin/activate
(my_py_venv) $ pip install --user some_pkg
(my_py_venv) $ pip uninstall some_pkg
WARNING: Skipping some_pkg as it is not installed.
(my_py_venv) $ pip list
# Even `pip list` will not properly list the `some_pkg` in this case
In this case, you have to deactivate the current virtual environment, then use the corresponding python/pip executable to list or uninstall the user site packages:
(my_py_venv) $ deactivate
$ path/to/python -m pip list
$ path/to/python -m pip uninstall some_pkg
Note that this issue was reported few years ago. And it seems that the current conclusion is: --user is not valid inside a virtual env’s pip, since a user location doesn’t really make sense for a virtual environment.
I strongly recommend you to use virtual environments for python package installation. With virtualenv, you prevent any package conflict and total isolation from your python related userland commands.
To delete all your package installed globally follow this;
It’s possible to uninstall packages installed with --user flag. This one worked for me;
pip freeze --user | xargs pip uninstall -y
For python 3;
pip3 freeze --user | xargs pip3 uninstall -y
But somehow these commands don’t uninstall setuptools and pip. After those commands (if you really want clean python) you may delete them with;
pip uninstall setuptools && pip uninstall pip
Now you have clean python environment. You can create virtualenv and install the package inside them.