Whenever I am trying to install any package using pip, I am getting this import error:

guru@guru-notebook:~$ pip3 install numpy
Traceback (most recent call last):
  File "/usr/bin/pip3", line 9, in <module>
    from pip import main
ImportError: cannot import name 'main'

guru@guru-notebook:~$ cat `which pip3`
#!/usr/bin/python3
# GENERATED BY DEBIAN

import sys

# Run the main entry point, similarly to how setuptools does it, but because
# we didn't install the actual entry point from setup.py, don't use the
# pkg_resources API.
from pip import main
if __name__ == '__main__':
    sys.exit(main())

It was working fine earlier, I am not sure why it is throwing this error. I have searched about this error, but can’t find anything to fix it.

Please let me know if you need any further detail, I will update my question.

You must have inadvertently upgraded your system pip (probably through something like sudo pip install pip --upgrade)

pip 10.x adjusts where its internals are situated. The pip3 command you’re seeing is one provided by your package maintainer (presumably debian based here?) and is not a file managed by pip.

You can read more about this on pip’s issue tracker

You’ll probably want to not upgrade your system pip and instead use a virtualenv.

To recover the pip3 binary you’ll need to sudo python3 -m pip uninstall pip && sudo apt install python3-pip --reinstall.

If you want to continue in “unsupported territory” (upgrading a system package outside of the system package manager), you can probably get away with python3 -m pip ... instead of pip3.

We can clear the error by modifying the pip file.

Check the location of the file:

$ which pip

path -> /usr/bin/pip

Go to that location(/usr/bin/pip) and open terminal

Enter: $ sudo nano pip

You can see:

import sys
from pip import main
if __name__ == '__main__':
     sys.exit(main())

Change to:

import sys
from pip import __main__
if __name__ == '__main__':
     sys.exit(__main__._main())

then ctrl + o write the changes and exit

Hope this will do!!

For Ubuntu family, Debian, Linux Mint users

Thanks to Anthony’s explanation above, you can retain your original system pip (in /usr/bin/ and dist-packages/) and remove the manually-installed pip (in ~/.local/) to resolve the conflict:

$ python3 -m pip uninstall pip

Ubuntu/Debian pip v8.1.1 (16.04) from python3-pip debian package (see$ pip3 -V) shows the same search results as the latest pip v10.0.1, and installs latest modules from PyPI just fine. It has a working pip command (already in the $PATH), plus the nice --user option patched-in by default since 2016. Looking at pip release notes, the newer versions are mostly about use-case specific bug fixes and certain new features, so not everyone has to rush upgrading pip just yet. And the new pip 10 can be deployed to Python virtualenvs, anyway.

But regardless of pips, your OS allows to quickly install common Python modules (including numpy) with APT, without the need for pip, for example:
$ sudo apt install python3-numpy python3-scipy (with system dependencies)
$ sudo apt install python3-pip (Debian-patched pip, slightly older but it doesn’t matter)

Quick apt syntax reminder (please see man apt for details):
$ sudo apt update (to resync Ubuntu package index files from up-to-date sources)
$ apt search <python-package-name> (full text-search on all available packages)
$ apt show <python-package-name> (displays the detailed package description)
$ sudo apt install <python-package-name>

Package names prefixed with python- are for Python 2; and prefixed with python3- are for Python 3 (e.g. python3-pandas). There are thousands, and they undergo integration testing within Debian and Ubuntu. Unless you seek to install at per-user level (pip install --user option) or within virtualenv/venv, apt could be what you needed. These system packages are accessible from virtual envs too, as virtualenv will gracefully fall back to using system libs on import if your envs don’t have given copies of modules. Your custom-installed (with pip --user) per-user modules in ~/.local/lib will override them too.

Note, since this is a system-wide installation, you’d rarely need to remove them (need to be mindful about OS dependencies). This is convenient for packages with many system dependencies (such as with scipy or matplotlib), as APT will keep track and provide all required system libs and C extensions, while with pip you have no such guarantees.

In fact, for system-wide Python packages (in contrast to per-user, home dir level, or lower), Ubuntu expects using the APT package manager (rather than sudo pip) to avoid breaking OS: sudo pip3 targets the very same /usr/lib/python3/dist-packages directory where APT stores OS-sensitive modules. Recent Debian/Ubuntu releases depend heavily on Python 3, so its pre-installed modules are managed by apt and shouldn’t be changed.

So if you use pip3 install command, please ensure that it runs in an isolated virtual dev environment, such as with virtualenv (sudo apt install python3-virtualenv), or with Python3 built-in (-m venv), or at a per-user level (--user pip option, default in Ubuntu-provided pip since 2016), but not system-wide (never sudo pip3!), because pip interferes with the operation of the APT package manager and may affect Ubuntu OS components when a system-used python module is unexpectedly changed. Good luck!

P. S. All the above is for the ‘ideal’ solution (Debian/Ubuntu way).

If you still want to use the new pip3 v10 exclusively, there are 3 quick workarounds:

• simply open a new bash session (a new terminal tab, or type bash) – and pip3 v10 becomes available (see pip3 -V). debian’s pip3 v8 remains installed but is broken; or
• the command $ hash -d pip3 && pip3 -V to refresh pip3 pathname in the $PATH. debian’s pip3 v8 remains installed but is broken; or
• the command $ sudo apt remove python3-pip && hash -d pip3 to uninstall debian’s pip3 v8 completely, in favor of your new pip3 v10.

Note: You will always need to add --user flag to any non-debian-provided pip, unless you are in a virtualenv! (it deploys python packages to ~/.local/, default in debian/ubuntu-provided python3-pip and python-pip since 2016). Your use of pip 10 system-wide, outside of virtualenv, is not really supported by Ubuntu/Debian. Never sudo pip3!

Further details:
https://github.com/pypa/pip/issues/5221#issuecomment-382069604
https://github.com/pypa/pip/issues/5240#issuecomment-381673100

resolved in one step only.

I too faced this issue, But this can be resolved simply by 1 command without bothering around and wasting time and i have tried it on multiple systems it’s the cleanest solution for this issue. And that’s:

For python3:- sudo python3 -m pip uninstall pip && sudo apt install python3-pip --reinstall.

By this , you can simply install packages using pip3. to check use pip3 --version.

For older versions, use : sudo python -m pip uninstall pip && sudo apt install python-pip --reinstall.

By this, now you can simply install packages using pip. to check use pip --version.

Use python -m pip install instead of pip install

Example:

python -m pip install --user somepackage
python3 -m pip install --user somepackage

The pip (resp. pip3) executable is provided by your distro (python-pip package on Ubuntu 16.04) and located at /usr/bin/pip.

Therefore, it is not kept up-to date with the pip package itself as you upgrade pip, and may break.

If you just use python -m pip directly, e.g. as in:

python -m pip install --user somepackage
python3 -m pip install --user somepackage

it goes through your Python path, finds the latest version of pip and executes that file.

It relies on the fact that file is executable through import, but that is a very standard type of interface, and therefore less likely to break than the hackier Debian script.

Then I recommend adding the following aliases to your .bashrc:

pip() ( python -m pip "$@" )
pip3() ( python3 -m pip "$@" )

The Ubuntu 18.04 /usr/bin/pip3 file does:

from pip import main

and presumably main was removed from pip at some point which is what broke things.

The breaking pip commit appears to be: 95bcf8c5f6394298035a7332c441868f3b0169f4 “Move all internal APIs to pip._internal” which went into pip 18.0.

Tested in Ubuntu 16.04 after an update from pip3 9.0.1 to 18.0.

pyenv

Ultimately however, for serious Python development I would just recommend that you install your own local Python with pyenv + virtualenv, which would also get around this Ubuntu bug: https://askubuntu.com/questions/682869/how-do-i-install-a-different-python-version-using-apt-get/1195153#1195153

You can resolve this issue by reinstalling pip.

Use one of the following command line commands to reinstall pip:

Python2:

python -m pip uninstall pip && sudo apt install python-pip --reinstall

Python3:

 python3 -m pip uninstall pip && sudo apt install python3-pip --reinstall

Check if pip has been cached on another path, to do so, call $ which pip and check that the path is different from the one prompted in the error, if that’s the case run:

$ hash -r

When the cache is clear, pip will be working again. reference: http://cheng.logdown.com/posts/2015/06/14/-usr-bin-pip-no-such-file-or-directory

I’m running on a system where I have sudo apt but no sudo pip. (And no su access.) I got myself into this same situation by following the advice from pip:

You are using pip version 8.1.1, however 18.0 is available. You should consider upgrading via the ‘pip install –upgrade pip’ command.

None of the other fixes worked for me, because I don’t have enough admin privileges. However, a few things stuck with me from reading up on this:

• I shouldn’t have done this. Sure, pip told me to. It lied.
• Using –user solves a lot of issues by focusing on the user-only directory.

So, I found this command line to work to revert me back to where I was. If you were using a different version than 8.1.1, you will obviously want to change that part of the line.

python -m pip install --force-reinstall pip==8.1.1 --user

That’s the only thing that worked for me, but it worked perfectly!

Recover with python3 -m pip install --user pip==9.0.1 (or the version that worked)

Same thing happened to me on Pixelbook using the new LXC (strech). This solution is very similar to the accepted one, with one subtle difference, whiched fixed pip3 for me.

sudo python3 -m pip install --upgrade pip

That bumped the version, and now it works as expected.

I found it here … Python.org: Ensure pip is up-to-date

I met the same problem on my Ubuntu 16.04 system. I managed to fix it by re-installing pip with the following command:

curl https://bootstrap.pypa.io/get-pip.py | sudo python3

The commands above didn’t work for me but those were very helpful:

sudo apt purge python3-pip
sudo rm -rf '/usr/lib/python3/dist-packages/pip'  
sudo apt install python3-pip
cd
cd .local/lib/python3/site-packages
sudo rm -rf pip*  
cd
cd .local/lib/python3.5/site-packages
sudo rm -rf pip*  
sudo pip3 install jupyter

In ubuntu 18.04.1 Bionic Beaver, you need to log out and log back in (restart not necessary) to get the proper environment.

$ sudo apt install python-pip

$ pip --version
pip 9.0.1 from /usr/lib/python2.7/dist-packages (python 2.7)

$ pip install --upgrade pip

$ pip --version
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main

$ exit
<login>

$ pip --version
pip 18.1 from /home/test/.local/lib/python2.7/site-packages/pip (python 2.7)

I use sudo apt remove python3-pip then pip works.

 ~ sudo pip install pip --upgrade
[sudo] password for sen: 
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name 'main'
➜  ~ sudo apt remove python3-pip   
Reading package lists... Done
Building dependency tree       
Reading state information... Done
The following packages were automatically installed and are no longer required:
  libexpat1-dev libpython3-dev libpython3.5-dev python-pip-whl python3-dev python3-wheel
  python3.5-dev
Use 'sudo apt autoremove' to remove them.
The following packages will be REMOVED:
  python3-pip
0 upgraded, 0 newly installed, 1 to remove and 0 not upgraded.
After this operation, 569 kB disk space will be freed.
Do you want to continue? [Y/n] y
(Reading database ... 215769 files and directories currently installed.)
Removing python3-pip (8.1.1-2ubuntu0.4) ...
Processing triggers for man-db (2.7.5-1) ...
➜  ~ pip

Usage:   
  pip <command> [options]

For Python version 2.7 @Anthony solution works perfect, by changing python3 to python as follows:

sudo python -m pip uninstall pip && sudo apt install python-pip --reinstall

What worked for me to fix the error with using pip3 was:

sudo cp -v /usr/local/bin/pip3 /usr/bin/pip3

Everything works:

 demon@UbuntuHP:~$ pip -V
 pip 10.0.1 from /usr/local/lib/python3.5/dist-packages/pip (python 3.5)

 demon@UbuntuHP:~$ pip2 -V
 pip 10.0.1 from /home/demon/.local/lib/python2.7/site-packages/pip (python 2.7)

 demon@UbuntuHP:~$ pip3 -V
 pip 10.0.1 from /usr/local/lib/python3.5/dist-packages/pip (python 3.5)

Maybe the new 10.0.1 version of pip doesn’t update the binary in /usr/bin ? (which seems it does not)

EDIT: the same issue occurs in Ubuntu 18.04. The best solution I’ve found is to symlink the pip binaries from /home/<user/.local/bin to /usr/local/bin or /usr/bin (depending on your preference), as follows:

ln -sv /home/<user>/.local/bin/pip /usr/local/bin/pip
ln -sv /home/<user>/.local/bin/pip2 /usr/local/bin/pip2
ln -sv /home/<user>/.local/bin/pip2.7 /usr/local/bin/pip2.7
ln -sv /home/<user>/.local/bin/pip3 /usr/local/bin/pip3
ln -sv /home/<user>/.local/bin/pip3.6 /usr/local/bin/pip3.6

NOTE: replace <user> with your current running user

The associated versions (latest) are in:

Version 3.6:

/home/demon/.local/lib/python3.6/site-packages/pip (python 3.6)

Version 2.7:

/home/demon/.local/lib/python2.7/site-packages/pip (python 2.7)

Trick and works too

sudo -H pip install lxml

I had this same error, but python -m pip was still working, so I fixed it with the nuclear option sudo python -m pip install --upgrade pip. It did it for me.

For what it’s worth, I had the problem with pip (not pip2 or pip3):

$ pip -V
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main

$ pip2 -V
pip 8.1.1 from /usr/lib/python2.7/dist-packages (python 2.7)

$ pip3 -V
pip 8.1.1 from /usr/lib/python3/dist-packages (python 3.5)

Somehow (I can’t remember how) I had python stuff installed in my ~/.local directory. After I removed the pip directory from there, pip started working again.

$ rm -rf /home/precor/.local/lib/python2.7/site-packages/pip
$ pip -V
pip 8.1.1 from /usr/lib/python2.7/dist-packages (python 2.7)

Is something wrong with the packages, when it generating de file /usr/bin/pip, you have to change the import:

from pip import main

to

from pip._internal import main

That solves the problem, I’m not sure why it generated, but it saids somthing in the following issue:

After pip 10 upgrade on pyenv “ImportError: cannot import name ‘main'”

You can try this:

sudo ln -sf $( type -P pip ) /usr/bin/pip

I also run into this problem when I wanted to upgrade system pip pip3 from 9.0.1 to 19.2.3.

After running pip3 install --upgrade pip, pip version becomes 19.2.3. But main() has been moved in pip._internal in the latest version, which leaves pip3 broken.

So in file /usr/bin/pip3, replace line 9: from pip import main with from pip._internal import main. The issue will be fixed, works the same for python2-pip. (Tested on Ubuntu 18.04 distribution)

According to @Vincent H.’s answer

Please run the following commands to do the fix. After running python3 -m pip install --upgrade pip, please run the following command.

hash -r pip

Source: https://github.com/pypa/pip/issues/5221

you can simply fix the pip and pip3 paths using update-alternatives

first thing you should check is your current $PATH run echo $PATH and see is you can find /usr/local/bin which is where pip3 and pip usually are

there is a change your system is looking here /bin/pip and /bin/pip3 so i will say fix the PATH by adding to your ~/.bash_profile file so it persists

export PATH=$PATH:/usr/local/bin and then check is its fixed with which pip and which pip3

if not then use update-alternatives to fix it finally

update-alternatives --install /bin/pip3 pip3 /usr/local/bin/pip3 30

and if you want to point pip to pip3 then

update-alternatives --install /bin/pip pip /usr/local/bin/pip3 30

This Worked for me !

hash -r pip # or hash -d pip

Now, uninstall the pip installed version and reinstall it using the following commands.

python -m pip uninstall pip  # sudo
sudo apt install --reinstall python-pip

If pip is broken, use:

python -m pip install --force-reinstall pip

Hope it helps!

import main from pip._internal

from pip._internal import main

Edit the pip code from

sudo nano /usr/bin/pip3

As @cryptoboy said – check what pip/python version you have installed

 demon@UbuntuHP:~$ pip -V
 demon@UbuntuHP:~$ pip2 -V
 demon@UbuntuHP:~$ pip3 -V

and then check for no-needed libraries in your .local/lib/ folder.

I did backup of settings when I was migrating to newer Kubuntu and in had .local/lib/python2.7/ folder in my home directory. Installed python 3.6. I just removed the old folder and now everything works great!

On Debian you will need to update apt first….

sudo apt-get update -qq
sudo apt-get install python-pip -qq
sudo pip install pip --upgrade --quiet
sudo pip2 install virtualenv --quiet

If you skip ‘sudo apt-get update -qq’ your pip will become corrupt and display the ‘cannot find main’ error.

This error may be a permission one. So, test executing the command with -H flag:

sudo -H pip3 install numpy

Use the following command before the execution of any pip command

hash -d pip

It will work

Given the name of a Python package that can be installed with pip, is there any way to find out a list of all the possible versions of it that pip could install? Right now it’s trial and error.

I’m trying to install a version for a third party library, but the newest version is too new, there were backwards incompatible changes made. So I’d like to somehow have a list of all the versions that pip knows about, so that I can test them.

(update: As of March 2020, many people have reported that yolk, installed via pip install yolk3k, only returns latest version. Chris’s answer seems to have the most upvotes and worked for me)

The script at pastebin does work. However it’s not very convenient if you’re working with multiple environments/hosts because you will have to copy/create it every time.

A better all-around solution would be to use yolk3k, which is available to install with pip. E.g. to see what versions of Django are available:

$ pip install yolk3k
$ yolk -V django
Django 1.3
Django 1.2.5
Django 1.2.4
Django 1.2.3
Django 1.2.2
Django 1.2.1
Django 1.2
Django 1.1.4
Django 1.1.3
Django 1.1.2
Django 1.0.4

yolk3k is a fork of the original yolk which ceased development in 2012. Though yolk is no longer maintained (as indicated in comments below), yolk3k appears to be and supports Python 3.

Note: I am not involved in the development of yolk3k. If something doesn’t seem to work as it should, leaving a comment here should not make much difference. Use the yolk3k issue tracker instead and consider submitting a fix, if possible.

For pip >= 9.0 use

$ pip install pylibmc==
Collecting pylibmc==
  Could not find a version that satisfies the requirement pylibmc== (from 
  versions: 0.2, 0.3, 0.4, 0.5.1, 0.5.2, 0.5.3, 0.5.4, 0.5.5, 0.5, 0.6.1, 0.6, 
  0.7.1, 0.7.2, 0.7.3, 0.7.4, 0.7, 0.8.1, 0.8.2, 0.8, 0.9.1, 0.9.2, 0.9, 
  1.0-alpha, 1.0-beta, 1.0, 1.1.1, 1.1, 1.2.0, 1.2.1, 1.2.2, 1.2.3, 1.3.0)
No matching distribution found for pylibmc==

– all the available versions will be printed without actually downloading or installing any additional packages.

For pip < 9.0 use

pip install pylibmc==blork

where blork can be any string that is not a valid version number.

Update:
As of Sep 2017 this method no longer works: --no-install was removed in pip 7

Use pip install -v, you can see all versions that available

root@node7:~# pip install web.py -v
Downloading/unpacking web.py
  Using version 0.37 (newest of versions: 0.37, 0.36, 0.35, 0.34, 0.33, 0.33, 0.32, 0.31, 0.22, 0.2)
  Downloading web.py-0.37.tar.gz (90Kb): 90Kb downloaded
  Running setup.py egg_info for package web.py
    running egg_info
    creating pip-egg-info/web.py.egg-info

To not install any package, use one of following solution:

root@node7:~# pip install --no-deps --no-install flask -v                                                                                                      
Downloading/unpacking flask
  Using version 0.10.1 (newest of versions: 0.10.1, 0.10, 0.9, 0.8.1, 0.8, 0.7.2, 0.7.1, 0.7, 0.6.1, 0.6, 0.5.2, 0.5.1, 0.5, 0.4, 0.3.1, 0.3, 0.2, 0.1)
  Downloading Flask-0.10.1.tar.gz (544Kb): 544Kb downloaded

or

root@node7:~# cd $(mktemp -d)
root@node7:/tmp/tmp.c6H99cWD0g# pip install flask -d . -v
Downloading/unpacking flask
  Using version 0.10.1 (newest of versions: 0.10.1, 0.10, 0.9, 0.8.1, 0.8, 0.7.2, 0.7.1, 0.7, 0.6.1, 0.6, 0.5.2, 0.5.1, 0.5, 0.4, 0.3.1, 0.3, 0.2, 0.1)
  Downloading Flask-0.10.1.tar.gz (544Kb): 4.1Kb downloaded

Tested with pip 1.0

root@node7:~# pip --version
pip 1.0 from /usr/lib/python2.7/dist-packages (python 2.7)

You don’t need a third party package to get this information. pypi provides simple JSON feeds for all packages under

https://pypi.python.org/pypi/{PKG_NAME}/json

Here’s some Python code using only the standard library which gets all versions.

import json
import urllib2
from distutils.version import StrictVersion

def versions(package_name):
    url = "https://pypi.python.org/pypi/%s/json" % (package_name,)
    data = json.load(urllib2.urlopen(urllib2.Request(url)))
    versions = data["releases"].keys()
    versions.sort(key=StrictVersion)
    return versions

print "\n".join(versions("scikit-image"))

That code prints (as of Feb 23rd, 2015):

0.7.2
0.8.0
0.8.1
0.8.2
0.9.0
0.9.1
0.9.2
0.9.3
0.10.0
0.10.1

I came up with dead-simple bash script. Thanks to jq‘s author.

#!/bin/bash
set -e

PACKAGE_JSON_URL="https://pypi.org/pypi/${1}/json"

curl -s "$PACKAGE_JSON_URL" | jq  -r '.releases | keys | .[]' | sort -V

Update: Add sorting by version number.

You could the yolk3k package instead of yolk. yolk3k is a fork from the original yolk and it supports both python2 and 3.

https://github.com/myint/yolk

pip install yolk3k

After looking at pip’s code for a while, it looks like the code responsible for locating packages can be found in the PackageFinder class in pip.index. Its method find_requirement looks up the versions of a InstallRequirement, but unfortunately only returns the most recent version.

The code below is almost a 1:1 copy of the original function, with the return in line 114 changed to return all versions.

The script expects one package name as first and only argument and returns all versions.

http://pastebin.com/axzdUQhZ

I can’t guarantee for the correctness, as I’m not familiar with pip’s code. But hopefully this helps.

Sample output

python test.py pip

Versions of pip
0.8.2
0.8.1
0.8
0.7.2
0.7.1
0.7
0.6.3
0.6.2
0.6.1
0.6
0.5.1
0.5
0.4
0.3.1
0.3
0.2.1
0.2 dev

The code:

import posixpath
import pkg_resources
import sys
from pip.download import url_to_path
from pip.exceptions import DistributionNotFound
from pip.index import PackageFinder, Link
from pip.log import logger
from pip.req import InstallRequirement
from pip.util import Inf


class MyPackageFinder(PackageFinder):

    def find_requirement(self, req, upgrade):
        url_name = req.url_name
        # Only check main index if index URL is given:
        main_index_url = None
        if self.index_urls:
            # Check that we have the url_name correctly spelled:
            main_index_url = Link(posixpath.join(self.index_urls[0], url_name))
            # This will also cache the page, so it's okay that we get it again later:
            page = self._get_page(main_index_url, req)
            if page is None:
                url_name = self._find_url_name(Link(self.index_urls[0]), url_name, req) or req.url_name

        # Combine index URLs with mirror URLs here to allow
        # adding more index URLs from requirements files
        all_index_urls = self.index_urls + self.mirror_urls

        def mkurl_pypi_url(url):
            loc = posixpath.join(url, url_name)
            # For maximum compatibility with easy_install, ensure the path
            # ends in a trailing slash.  Although this isn't in the spec
            # (and PyPI can handle it without the slash) some other index
            # implementations might break if they relied on easy_install's behavior.
            if not loc.endswith('/'):
                loc = loc + '/'
            return loc
        if url_name is not None:
            locations = [
                mkurl_pypi_url(url)
                for url in all_index_urls] + self.find_links
        else:
            locations = list(self.find_links)
        locations.extend(self.dependency_links)
        for version in req.absolute_versions:
            if url_name is not None and main_index_url is not None:
                locations = [
                    posixpath.join(main_index_url.url, version)] + locations

        file_locations, url_locations = self._sort_locations(locations)

        locations = [Link(url) for url in url_locations]
        logger.debug('URLs to search for versions for %s:' % req)
        for location in locations:
            logger.debug('* %s' % location)
        found_versions = []
        found_versions.extend(
            self._package_versions(
                [Link(url, '-f') for url in self.find_links], req.name.lower()))
        page_versions = []
        for page in self._get_pages(locations, req):
            logger.debug('Analyzing links from page %s' % page.url)
            logger.indent += 2
            try:
                page_versions.extend(self._package_versions(page.links, req.name.lower()))
            finally:
                logger.indent -= 2
        dependency_versions = list(self._package_versions(
            [Link(url) for url in self.dependency_links], req.name.lower()))
        if dependency_versions:
            logger.info('dependency_links found: %s' % ', '.join([link.url for parsed, link, version in dependency_versions]))
        file_versions = list(self._package_versions(
                [Link(url) for url in file_locations], req.name.lower()))
        if not found_versions and not page_versions and not dependency_versions and not file_versions:
            logger.fatal('Could not find any downloads that satisfy the requirement %s' % req)
            raise DistributionNotFound('No distributions at all found for %s' % req)
        if req.satisfied_by is not None:
            found_versions.append((req.satisfied_by.parsed_version, Inf, req.satisfied_by.version))
        if file_versions:
            file_versions.sort(reverse=True)
            logger.info('Local files found: %s' % ', '.join([url_to_path(link.url) for parsed, link, version in file_versions]))
            found_versions = file_versions + found_versions
        all_versions = found_versions + page_versions + dependency_versions
        applicable_versions = []
        for (parsed_version, link, version) in all_versions:
            if version not in req.req:
                logger.info("Ignoring link %s, version %s doesn't match %s"
                            % (link, version, ','.join([''.join(s) for s in req.req.specs])))
                continue
            applicable_versions.append((link, version))
        applicable_versions = sorted(applicable_versions, key=lambda v: pkg_resources.parse_version(v[1]), reverse=True)
        existing_applicable = bool([link for link, version in applicable_versions if link is Inf])
        if not upgrade and existing_applicable:
            if applicable_versions[0][1] is Inf:
                logger.info('Existing installed version (%s) is most up-to-date and satisfies requirement'
                            % req.satisfied_by.version)
            else:
                logger.info('Existing installed version (%s) satisfies requirement (most up-to-date version is %s)'
                            % (req.satisfied_by.version, applicable_versions[0][1]))
            return None
        if not applicable_versions:
            logger.fatal('Could not find a version that satisfies the requirement %s (from versions: %s)'
                         % (req, ', '.join([version for parsed_version, link, version in found_versions])))
            raise DistributionNotFound('No distributions matching the version for %s' % req)
        if applicable_versions[0][0] is Inf:
            # We have an existing version, and its the best version
            logger.info('Installed version (%s) is most up-to-date (past versions: %s)'
                        % (req.satisfied_by.version, ', '.join([version for link, version in applicable_versions[1:]]) or 'none'))
            return None
        if len(applicable_versions) > 1:
            logger.info('Using version %s (newest of versions: %s)' %
                        (applicable_versions[0][1], ', '.join([version for link, version in applicable_versions])))
        return applicable_versions


if __name__ == '__main__':
    req = InstallRequirement.from_line(sys.argv[1], None)
    finder = MyPackageFinder([], ['http://pypi.python.org/simple/'])
    versions = finder.find_requirement(req, False)
    print 'Versions of %s' % sys.argv[1]
    for v in versions:
        print v[1]

You can use this small Python 3 script (using only standard library modules) to grab the list of available versions for a package from PyPI using JSON API and print them in reverse chronological order. Unlike some other Python solutions posted here, this doesn’t break on loose versions like django‘s 2.2rc1 or uwsgi‘s 2.0.17.1:

#!/usr/bin/env python3

import json
import sys
from urllib import request    
from pkg_resources import parse_version    

def versions(pkg_name):
    url = f'https://pypi.python.org/pypi/{pkg_name}/json'
    releases = json.loads(request.urlopen(url).read())['releases']
    return sorted(releases, key=parse_version, reverse=True)    

if __name__ == '__main__':
    print(*versions(sys.argv[1]), sep='\n')

Save the script and run it with the package name as an argument, e.g.:

python versions.py django

3.0a1
2.2.5
2.2.4
2.2.3
2.2.2
2.2.1
2.2
2.2rc1
...

https://pypi.python.org/pypi/Django/ – works for packages whose maintainers choose to show all packages https://pypi.python.org/simple/pip/ – should do the trick anyhow (lists all links)

This works for me on OSX:

pip install docker-compose== 2>&1 \
| grep -oE '(\(.*\))' \
| awk -F:\  '{print$NF}' \
| sed -E 's/( |\))//g' \
| tr ',' '\n'

It returns the list one per line:

1.1.0rc1
1.1.0rc2
1.1.0
1.2.0rc1
1.2.0rc2
1.2.0rc3
1.2.0rc4
1.2.0
1.3.0rc1
1.3.0rc2
1.3.0rc3
1.3.0
1.3.1
1.3.2
1.3.3
1.4.0rc1
1.4.0rc2
1.4.0rc3
1.4.0
1.4.1
1.4.2
1.5.0rc1
1.5.0rc2
1.5.0rc3
1.5.0
1.5.1
1.5.2
1.6.0rc1
1.6.0
1.6.1
1.6.2
1.7.0rc1
1.7.0rc2
1.7.0
1.7.1
1.8.0rc1
1.8.0rc2
1.8.0
1.8.1
1.9.0rc1
1.9.0rc2
1.9.0rc3
1.9.0rc4
1.9.0
1.10.0rc1
1.10.0rc2
1.10.0

Or to get the latest version available:

pip install docker-compose== 2>&1 \
| grep -oE '(\(.*\))' \
| awk -F:\  '{print$NF}' \
| sed -E 's/( |\))//g' \
| tr ',' '\n' \
| gsort -r -V \
| head -1

1.10.0rc2

Keep in mind gsort has to be installed (on OSX) to parse the versions. You can install it with brew install coreutils

My project luddite has this feature.

Example usage:

>>> import luddite
>>> luddite.get_versions_pypi("python-dateutil")
('0.1', '0.3', '0.4', '0.5', '1.0', '1.1', '1.2', '1.4', '1.4.1', '1.5', '2.0', '2.1', '2.2', '2.3', '2.4.0', '2.4.1', '2.4.2', '2.5.0', '2.5.1', '2.5.2', '2.5.3', '2.6.0', '2.6.1', '2.7.0', '2.7.1', '2.7.2', '2.7.3', '2.7.4', '2.7.5', '2.8.0')

It lists all versions of a package available, by querying the json API of https://pypi.org/

I didn’t have any luck with yolk, yolk3k or pip install -v but so I ended up using this (adapted to Python 3 from eric chiang’s answer):

import json
import requests
from distutils.version import StrictVersion

def versions(package_name):
    url = "https://pypi.python.org/pypi/{}/json".format(package_name)
    data = requests.get(url).json()
    return sorted(list(data["releases"].keys()), key=StrictVersion, reverse=True)

>>> print("\n".join(versions("gunicorn")))
19.1.1
19.1.0
19.0.0
18.0
17.5
0.17.4
0.17.3
...

Alternative solution is to use the Warehouse APIs:

https://warehouse.readthedocs.io/api-reference/json/#release

For instance for Flask:

import requests
r = requests.get("https://pypi.org/pypi/Flask/json")
print(r.json()['releases'].keys())

will print:

dict_keys(['0.1', '0.10', '0.10.1', '0.11', '0.11.1', '0.12', '0.12.1', '0.12.2', '0.12.3', '0.12.4', '0.2', '0.3', '0.3.1', '0.4', '0.5', '0.5.1', '0.5.2', '0.6', '0.6.1', '0.7', '0.7.1', '0.7.2', '0.8', '0.8.1', '0.9', '1.0', '1.0.1', '1.0.2'])

Simple bash script that relies only on python itself (I assume that in the context of the question it should be installed) and one of curl or wget. It has an assumption that you have setuptools package installed to sort versions (almost always installed). It doesn’t rely on external dependencies such as:

• jq which may not be present;
• grep and awk that may behave differently on Linux and macOS.

curl --silent --location https://pypi.org/pypi/requests/json | python -c "import sys, json, pkg_resources; releases = json.load(sys.stdin)['releases']; print(' '.join(sorted(releases, key=pkg_resources.parse_version)))"

A little bit longer version with comments.

Put the package name into a variable:

PACKAGE=requests

Get versions (using curl):

VERSIONS=$(curl --silent --location https://pypi.org/pypi/$PACKAGE/json | python -c "import sys, json, pkg_resources; releases = json.load(sys.stdin)['releases']; print(' '.join(sorted(releases, key=pkg_resources.parse_version)))")

Get versions (using wget):

VERSIONS=$(wget -qO- https://pypi.org/pypi/$PACKAGE/json | python -c "import sys, json, pkg_resources; releases = json.load(sys.stdin)['releases']; print(' '.join(sorted(releases, key=pkg_resources.parse_version)))")

Print sorted versions:

echo $VERSIONS

My take is a combination of a couple of posted answers, with some modifications to make them easier to use from within a running python environment.

The idea is to provide a entirely new command (modeled after the install command) that gives you an instance of the package finder to use. The upside is that it works with, and uses, any indexes that pip supports and reads your local pip configuration files, so you get the correct results as you would with a normal pip install.

I’ve made an attempt at making it compatible with both pip v 9.x and 10.x.. but only tried it on 9.x

https://gist.github.com/kaos/68511bd013fcdebe766c981f50b473d4

#!/usr/bin/env python
# When you want a easy way to get at all (or the latest) version of a certain python package from a PyPi index.

import sys
import logging

try:
    from pip._internal import cmdoptions, main
    from pip._internal.commands import commands_dict
    from pip._internal.basecommand import RequirementCommand
except ImportError:
    from pip import cmdoptions, main
    from pip.commands import commands_dict
    from pip.basecommand import RequirementCommand

from pip._vendor.packaging.version import parse as parse_version

logger = logging.getLogger('pip')

class ListPkgVersionsCommand(RequirementCommand):
    """
    List all available versions for a given package from:

    - PyPI (and other indexes) using requirement specifiers.
    - VCS project urls.
    - Local project directories.
    - Local or remote source archives.

    """
    name = "list-pkg-versions"
    usage = """
      %prog [options] <requirement specifier> [package-index-options] ...
      %prog [options] [-e] <vcs project url> ...
      %prog [options] [-e] <local project path> ...
      %prog [options] <archive url/path> ..."""

    summary = 'List package versions.'

    def __init__(self, *args, **kw):
        super(ListPkgVersionsCommand, self).__init__(*args, **kw)

        cmd_opts = self.cmd_opts

        cmd_opts.add_option(cmdoptions.install_options())
        cmd_opts.add_option(cmdoptions.global_options())
        cmd_opts.add_option(cmdoptions.use_wheel())
        cmd_opts.add_option(cmdoptions.no_use_wheel())
        cmd_opts.add_option(cmdoptions.no_binary())
        cmd_opts.add_option(cmdoptions.only_binary())
        cmd_opts.add_option(cmdoptions.pre())
        cmd_opts.add_option(cmdoptions.require_hashes())

        index_opts = cmdoptions.make_option_group(
            cmdoptions.index_group,
            self.parser,
        )

        self.parser.insert_option_group(0, index_opts)
        self.parser.insert_option_group(0, cmd_opts)

    def run(self, options, args):
        cmdoptions.resolve_wheel_no_use_binary(options)
        cmdoptions.check_install_build_global(options)

        with self._build_session(options) as session:
            finder = self._build_package_finder(options, session)

            # do what you please with the finder object here... ;)
            for pkg in args:
                logger.info(
                    '%s: %s', pkg,
                    ', '.join(
                        sorted(
                            set(str(c.version) for c in finder.find_all_candidates(pkg)),
                            key=parse_version,
                        )
                    )
                )


commands_dict[ListPkgVersionsCommand.name] = ListPkgVersionsCommand

if __name__ == '__main__':
    sys.exit(main())

Example output

./list-pkg-versions.py list-pkg-versions pika django

pika: 0.5, 0.5.1, 0.5.2, 0.9.1a0, 0.9.2a0, 0.9.3, 0.9.4, 0.9.5, 0.9.6, 0.9.7, 0.9.8, 0.9.9, 0.9.10, 0.9.11, 0.9.12, 0.9.13, 0.9.14, 0.10.0b1, 0.10.0b2, 0.10.0, 0.11.0b1, 0.11.0, 0.11.1, 0.11.2, 0.12.0b2
django: 1.1.3, 1.1.4, 1.2, 1.2.1, 1.2.2, 1.2.3, 1.2.4, 1.2.5, 1.2.6, 1.2.7, 1.3, 1.3.1, 1.3.2, 1.3.3, 1.3.4, 1.3.5, 1.3.6, 1.3.7, 1.4, 1.4.1, 1.4.2, 1.4.3, 1.4.4, 1.4.5, 1.4.6, 1.4.7, 1.4.8, 1.4.9, 1.4.10, 1.4.11, 1.4.12, 1.4.13, 1.4.14, 1.4.15, 1.4.16, 1.4.17, 1.4.18, 1.4.19, 1.4.20, 1.4.21, 1.4.22, 1.5, 1.5.1, 1.5.2, 1.5.3, 1.5.4, 1.5.5, 1.5.6, 1.5.7, 1.5.8, 1.5.9, 1.5.10, 1.5.11, 1.5.12, 1.6, 1.6.1, 1.6.2, 1.6.3, 1.6.4, 1.6.5, 1.6.6, 1.6.7, 1.6.8, 1.6.9, 1.6.10, 1.6.11, 1.7, 1.7.1, 1.7.2, 1.7.3, 1.7.4, 1.7.5, 1.7.6, 1.7.7, 1.7.8, 1.7.9, 1.7.10, 1.7.11, 1.8a1, 1.8b1, 1.8b2, 1.8rc1, 1.8, 1.8.1, 1.8.2, 1.8.3, 1.8.4, 1.8.5, 1.8.6, 1.8.7, 1.8.8, 1.8.9, 1.8.10, 1.8.11, 1.8.12, 1.8.13, 1.8.14, 1.8.15, 1.8.16, 1.8.17, 1.8.18, 1.8.19, 1.9a1, 1.9b1, 1.9rc1, 1.9rc2, 1.9, 1.9.1, 1.9.2, 1.9.3, 1.9.4, 1.9.5, 1.9.6, 1.9.7, 1.9.8, 1.9.9, 1.9.10, 1.9.11, 1.9.12, 1.9.13, 1.10a1, 1.10b1, 1.10rc1, 1.10, 1.10.1, 1.10.2, 1.10.3, 1.10.4, 1.10.5, 1.10.6, 1.10.7, 1.10.8, 1.11a1, 1.11b1, 1.11rc1, 1.11, 1.11.1, 1.11.2, 1.11.3, 1.11.4, 1.11.5, 1.11.6, 1.11.7, 1.11.8, 1.11.9, 1.11.10, 1.11.11, 1.11.12, 2.0, 2.0.1, 2.0.2, 2.0.3, 2.0.4

I need to install psycopg2 v2.4.1 specifically. I accidentally did:

 pip install psycopg2

Instead of:

 pip install psycopg2==2.4.1

That installs 2.4.4 instead of the earlier version.

Now even after I pip uninstall psycopg2 and attempt to reinstall with the correct version, it appears that pip is re-using the cache it downloaded the first time.

How can I force pip to clear out its download cache and use the specific version I’m including in the command?

If using pip 6.0 or newer, try adding the --no-cache-dir option.

If using pip older than pip 6.0, upgrade it with pip install -U pip.

Clear the cache directory where appropriate for your system

Linux and Unix

~/.cache/pip  # and it respects the XDG_CACHE_HOME directory.

OS X

~/Library/Caches/pip

Windows

%LocalAppData%\pip\Cache

From documentation at https://pip.pypa.io/en/latest/reference/pip_install.html#caching:

Starting with v6.0, pip provides an on-by-default cache which functions similarly to that of a web browser. While the cache is on by default and is designed do the right thing by default you can disable the cache and always access PyPI by utilizing the --no-cache-dir option.

pip can install a package ignoring the cache, like this

pip --no-cache-dir install scipy

On Ubuntu, I had to delete /tmp/pip-build-root.

(pip maintainer here!)

Since pip 6.0 (back in 2014!), pip install, pip download and pip wheel commands can be told to avoid using the cache with the --no-cache-dir option. (eg: pip install --no-cache-dir <package>)

Since pip 10.0 (back in 2018!), a pip config command was added, which can be used to configure pip to always ignore the cache — pip config set global.cache-dir false configures pip to not use the cache “globally” (i.e. in all commands).

Since pip 20.1, pip has a pip cache command to manage the contents of pip’s cache.

• pip cache purge removes all the wheel files in the cache.
• pip cache remove matplotlib selectively removes files related to a matplotlib from the cache.

In summary, pip provides a lot of ways to tweak how it uses the cache:

• pip install --no-cache-dir <package>: install a package without using the cache, for just this run.
• pip config set global.cache-dir false: configure pip to not use the cache “globally” (in all commands)
• pip cache remove matplotlib: removes all wheel files related to matplotlib from pip’s cache.
• pip cache purge: to clear all files from pip’s cache.

The specific issue of “installing the wrong version due to caching” issue mentioned in the question was fixed in pip 1.4 (back in 2013!):

Fix a number of issues related to cleaning up and not reusing build directories. (#413, #709, #634, #602, #939, #865, #948)

If you like to set the --no-cache-dir option by default, you can put this into pip.conf:

[global]
no-cache-dir = false

The location of pip.conf depends on your OS. See the documentation for more info.

I just had a similar problem and found that the only way to get pip to upgrade the package was to delete the $PWD/build (%CD%\build on Windows) directory that might have been left over from a previously unfinished install or a previous version of pip (it now deletes the build directories after a successful install).

On archlinux pip cache is located at ~/.cache/pip, I could solve my issue by removing the http folder inside it.

On my mac I had to remove the cache directory ~/Library/Caches/pip/

Since pip 20.1b1, which was released on 21 April 2020 and “added pip cache command for inspecting/managing pip’s wheel cache”, it is possible to issue this command:

pip cache purge

The reference guide is here:
https://pip.pypa.io/en/stable/reference/pip_cache/
The corresponding pull request is here.

On Windows 7, I had to delete %HOMEPATH%/pip.

If using virtualenv, look for the build directory under your environments root.

I had to delete %TEMP%\pip-build On Windows 7

On Mac OS (Mavericks), I had to delete /tmp/pip-build/

A better way to do it is to delete the cache and rebuild it. In this way, if you install it again for other virtualenv, it will use the cache instead of building every time when you install it.

For example, when you install it, it will say it uses cached wheel,

Processing <some_prefix>/Library/Caches/pip/wheels/d0/c4/e4/e49fd07bca8dda00dd6b4bbc606aa05a25aacb00d45747a47a/horovod-0.19.3-cp37-cp37m-macosx_10_9_x86_64.wh

Just delete that one and restart your install.

(…) it appears that pip is re-using the cache (…)

I’m pretty sure that’s not what’s happening. Pip used to (wrongly) reuse build directory not cache. This was fixed in version 1.4 of pip which was released on 2013-07-23.

I know the obvious answer is to use virtualenv and virtualenvwrapper, but for various reasons I can’t/don’t want to do that.

So how do I modify the command

pip install package_name

to make pip install the package somewhere other than the default site-packages?

Use:

pip install --install-option="--prefix=$PREFIX_PATH" package_name

You might also want to use --ignore-installed to force all dependencies to be reinstalled using this new prefix. You can use --install-option to multiple times to add any of the options you can use with python setup.py install (--prefix is probably what you want, but there are a bunch more options you could use).

The –target switch is the thing you’re looking for:

pip install --target=d:\somewhere\other\than\the\default package_name

But you still need to add d:\somewhere\other\than\the\default to PYTHONPATH to actually use them from that location.

-t, –target <dir>
Install packages into <dir>. By default this will not replace existing files/folders in <dir>.
Use –upgrade to replace existing packages in <dir> with new versions.

Upgrade pip if target switch is not available:

On Linux or OS X:

pip install -U pip

On Windows (this works around an issue):

python -m pip install -U pip

Instead of the --target option or the --install-options option, I have found that the following works well (from discussion on a bug regarding this very thing at https://github.com/pypa/pip/issues/446):