Check if module exists, if not install it

You can launch pip install %s"%keyring in the except part to do this but I don’t recommend it. The correct way is to package your application using distutils so that when it’s installed, dependencies will be pulled in.

Not all modules can be installed so easily. Not all of them have easy-install support, some can only be installed by building them.. others require some non-python prerequisites, like gcc, which makes things even more complicated (and forget about it working well on Windows).

So I would say you could probably make it work for some predetermined modules, but there’s no chance it’ll be something generic that works for any module.

This approach of dynamic import work really well in cases you just want to print a message if module is not installed. Automatically installing a module SHOULDN’T be done like issuing pip via subprocess. That’s why we have setuptools (or distribute).

We have some great tutorials on packaging, and the task of dependencies detection/installation is as simple as providing install_requires=[ 'FancyDependency', 'otherFancy>=1.0' ]. That’s just it!

But, if you really NEED to do by hand, you can use setuptools to help you.

from pkg_resources import WorkingSet , DistributionNotFound
working_set = WorkingSet()

# Printing all installed modules
print tuple(working_set)

# Detecting if module is installed
try:
    dep = working_set.require('paramiko>=1.0')
except DistributionNotFound:
    pass

# Installing it (anyone knows a better way?)
from setuptools.command.easy_install import main as install
install(['django>=1.2'])

import pip

def import_or_install(package):
    try:
        __import__(package)
    except ImportError:
        pip.main(['install', package])       

This code simply attempt to import a package, where package is of type str, and if it is unable to, calls pip and attempt to install it from there.

You can use os.system as follows:

import os

package = "package_name"

try:
    __import__package
except:
    os.system("pip install "+ package)

NOTE: Ipython / Jupyter specific solution.

While using notebooks / online kernels, I usually do it using systems call.

try:
  import keyring
except:
  !pip install keyring
  import keyring

P.S. One may wish to call conda install or mamba install instead.

I made an import_neccessary_modules() function to fix this common issue.

# ======================================================================================
# == Fix any missing Module, that need to be installed with PIP.exe. [Windows System] ==
# ======================================================================================
import importlib, os
def import_neccessary_modules(modname:str)->None:
    '''
        Import a Module,
        and if that fails, try to use the Command Window PIP.exe to install it,
        if that fails, because PIP in not in the Path,
        try find the location of PIP.exe and again attempt to install from the Command Window.
    '''
    try:
        # If Module it is already installed, try to Import it
        importlib.import_module(modname)
        print(f"Importing {modname}")
    except ImportError:
        # Error if Module is not installed Yet,  the '33[93m' is just code to print in certain colors
        print(f"33[93mSince you don't have the Python Module [{modname}] installed!")
        print("I will need to install it using Python's PIP.exe command.33[0m")
        if os.system('PIP --version') == 0:
            # No error from running PIP in the Command Window, therefor PIP.exe is in the %PATH%
            os.system(f'PIP install {modname}')
        else:
            # Error, PIP.exe is NOT in the Path!! So I'll try to find it.
            pip_location_attempt_1 = sys.executable.replace("python.exe", "") + "pip.exe"
            pip_location_attempt_2 = sys.executable.replace("python.exe", "") + "scriptspip.exe"
            if os.path.exists(pip_location_attempt_1):
                # The Attempt #1 File exists!!!
                os.system(pip_location_attempt_1 + " install " + modname)
            elif os.path.exists(pip_location_attempt_2):
                # The Attempt #2 File exists!!!
                os.system(pip_location_attempt_2 + " install " + modname)
            else:
                # Neither Attempts found the PIP.exe file, So i Fail...
                print(f"33[91mAbort!!!  I can't find PIP.exe program!")
                print(f"You'll need to manually install the Module: {modname} in order for this program to work.")
                print(f"Find the PIP.exe file on your computer and in the CMD Command window...")
                print(f"   in that directory, type    PIP.exe install {modname}33[0m")
                exit()


import_neccessary_modules('art')
import_neccessary_modules('pyperclip')
import_neccessary_modules('winsound')

I tried installing transformers using the below method and it worked fine. Similarly, you can just replace your library name instead of "transformers".

import pip
try:
    from transformers import pipeline
except ModuleNotFoundError:
    pip.main(['install', "transformers"])
    from transformers import pipeline

I tried this in a new virtual envoirnment with no packages installed and it installed the necessary package i.e. opencv-python
Example is given below

import os

try:
    import cv2
except ImportError:
    os.system('pip install opencv-python')

Here is my approach. The idea is loop until python has already installed all modules by built in module as "pip" .

import pip

while True:
    
    try:
        #import your modules here. !
        import seaborn
        import bokeh

        break

    except ImportError as err_mdl:
        
        print((err_mdl.name))
        pip.main(['install', err_mdl.name])