Dependencies and the solver

​

What are dependencies?

When you install conda packages, you’re rarely installing just one package. Most of the programs and libraries contained within these packages rely on the programs and libraries in other packages to function properly. These other packages are called dependencies.

​

What is the solver?

One of the main features of conda is its ability to manage software packages and their dependencies. Dependency management can often become complicated, especially when a package has many dependencies. Conda uses something called its dependency solver, an algorithm that determines which versions of which packages need to be installed to satisfy (or solve) all the dependencies without version conflicts.

​

How conda manages dependencies

In order to organize the dependency needs of a given package, conda requires that all conda packages ship with the following:

• Package Specifications: Each package in conda comes with a list of the required dependencies it needs to function
• Version Requirements: Dependencies often specify version constraints, such as:
  • numpy >=1.18 (NumPy version 1.18 or later)
  • python =3.8.* (Any version of Python 3.8)
  • scipy <1.5.0 (SciPy versions older than 1.5.0)

When installing a package, conda’s solver looks at the package specification for the primary package and for each one of its dependencies (and their dependencies and so on) to determine which package versions can work together.

For more information on conda’s package specification rules, see Package specification in the official conda documentation.

For advice on fixing solver errors, see Managing solver errors.