- The Issue / Potential Confusion
- How Windows Decides Which Executable to Execute
Quite often, I get comments from people telling me that a module doesn't work, that they have installed the library/module:
pip install xyz
And then executed
python file.py which says something like the following:
Traceback (most recent call last): File "file.py", line 1, in <module> import xyz ModuleNotFoundError: No module named 'xyz'
Two things could have happened here:
- Installing the
xyzmodule failed. This would have clearly been stated when executing
pip install xyz.
- When using
pip, the module has been installed in one distribution of Python and then a different distribution of Python has been used to try and import the module.
Point 2 is what I will be focusing on in this post and how to fix confusion between more than one installed distribution of Python.
Aside from the situation outlined above, there are better methods of proving that this is the case. The best method in my opinion, would be to identify where the
python executables are and check if they are in the same Python distribution. To do this, you can use the
where command in Windows.
For example, to identify what executable is executed when I execute
python, I would execute:
This will print one or more paths. When I execute it for example, the following is printed:
Most people will find that they have only one path in this returned list. If you have more than one printed, use the top one; I discuss why more than one path may be returned later.
To find the location of pip, I would then execute:
Which would then output one or more paths. In my case, it has output:
Using the paths returned from these two
where calls, you can then see if they both belong to the same distribution or not.
Typically a Python distribution has a root folder which contains python.exe and a Script folder which contains pip.exe. Using my example, you can see that my paths follow the pattern as the Python root is "C:\Python37\".
In the case you have paths that do not have a common root folder like:
> where python C:\path\to\distribution1\python.exe > where pip C:\a\different\path\to\distribution2\Scripts\pip.exe
We can see that they do not follow the Python root folder structure I described. We can also see they do not share the same parent directory so they must be from different distributions.
When you execute a command in a terminal, the terminal has to figure out which executable to call. As shown above, we can use the
where command to identify which executable is found and executed.
where prints the paths of executables that the terminal has identified with the name you provided; the top file path is always the one that is executed.
The terminal has found these by searching the PATH environment variable. The PATH environment variable is a string that contains directories to look at when searching for executables that can be executed. This means you can make an executable/.exe available to use in cmd by adding the folder that it's located in to the PATH environment variable.
To look at your PATH environment variable, execute:
To modify your PATH environment variable, look at my tutorial on How To Setup Python's PIP. In this post, I show you how to add a path to the PATH environment variable.
When I'm in a virtual environment and execute
where python, I get the following output:
I now have two "python" executables I can execute, but only the top one will be used (we only need to execute one of them).
The terminal has identified these paths in this order by looking at my PATH environment variable and searching each folder listed in it for a "python" executable. Each path is searched based on its position in the PATH environment variable - paths at the start get searched first.
When not using
where, the terminal will typically find the first executable required (first path in the list) and then stop searching as it has what it needs. This means when there are many executables with the same name in the folders to be searched, the executable found first when searching the folders in order will be executed.
There are a couple of common solutions that we can use to fix this mixing of distributions issue. Both these solutions use the knowledge above about how Windows finds the executables we are wanting to execute.
This is my most recommended as you can be sure everything is where you expect it.
Virtual environments can solve the issue we're looking at because it sets up a folder structure for its own dependencies and executables and then adds the "Scripts" path of this environment to the beginning of the PATH environment variable so that the executables within will be found instead of other executables outside the environment.
There are many tutorials on setting up virtual environments in Python and there are many different modules you can use; but here are some basics using the Python 3 built-in venv:
To set up a virtual environment, open a terminal and execute the following:
python -m venv venv_folder
This will ask Python's venv module to create a virtual environment in
venv_folder. You can substitute
venv_folder for a different folder name if you wish, the convention is to use a folder named
venv but I have used
venv_folder in this tutorial to better identify it as a folder.
Before activating the virtual environment, executing
pip will still use the globally setup distribution(s). To use the virtual environment executables, you need to activate it.
While beside the
venv_folder folder (or whatever you named it) in cmd, execute:
If your folder was named
venv/Scripts/activate(or substitute for a different name).
Now when you execute
where python and
where pip, you will see that executables from this virtual environment are being used. This is because the Scripts folder in the virtual environment has been added to the start of the PATH variable; you can see this by executing
echo %PATH% again.
To deactivate the virtual environment, go back to the same folder in cmd beside
venv_folder and execute:
where python and
where pip again, you will see that the original setup has been restored.
Also, since these methods only modify environment variables in the session, closing the terminal session will throw away the activation. You can easily reactivate the virtual environment by following the instructions above again.
Now that you know how to set up, activate and deactivate a virtual environment, you can use this to be sure what
pip you are executing.
To install dependencies, use the normal
pip install xyz and to run a Python script, use the normal
If you don't want to deal with virtual environments, you could just set your PATH environment variable to make sure you get the correct
pip each time.
To do this, first identify the root of the Python distribution you want to use. If you already have access to the
python executable or IDLE you want to use, execute the following in it:
import os import sys print (os.path.dirname(sys.executable)) print (os.path.dirname(sys.executable) + '\Scripts\\')
This will print out two paths; one which is the root folder for the Python distribution that is being used and the other being its Scripts folder.
Following my tutorial on How To Setup Python's PIP, add these two paths to the beginning/top of the PATH environment variable.
Now when you execute
where python and
pip.exe should be found in these two folders you added first.
When making these changes to the PATH environment variable, you can also identify folders from other Python distribution that you don't actively use and remove them from the PATH environment variable. Then to use these other distributions of Python, absolutely reference the executable in cmd, e.g.