You already know that pip installs packages in one of two environments:
- The system-wide environment
- The user environment
When importing a module or package, the Python interpreter looks for it first in the user environment, then in the system-wide environment. This sequence allows us to use the exact versions of the libraries and Python programs we need.
But even two environments aren't enough when a programmer starts working on several projects because projects can have different dependencies.
And there are even more complex cases. Different projects may depend on one library but require different versions of it. That'll mean a version conflict.
A Python developer needs some mechanism to keep different projects in isolated sandboxes. There is a mechanism precisely for this. It's called virtual environments.
Designing virtual environments
Each virtual environment is a directory.
Its contents are structurally similar to the system-wide environment — the subdirectories we named and populated accordingly. Let's look at an example:
tree env
env
├── bin
│ ├── activate
│ …
│ ├── pip
│ …
│ └── python3 -> …/python3
├── lib
│ └── python3.6
│ └── site-packages
│ ├── pip
│ │ ├── …
│ …
│ └── setuptools-40.6.2.dist-info
│ ├── …
…