Managing solver errors

Although conda’s solver is designed to install packages and their dependencies without creating version conflicts, it can still happen. This tutorial will go over possible solver conflict scenarios and ways to fix them.

​

Direct dependency conflict

Imagine you have an environment that you know will require Python 3.9, but you also want to install package-a into that environment. You decide to specify version 2.0 of package-a.

The only problem is, package-a 2.0 requires Python 3.10 or later (>=3.10) and doesn’t have any backwards compatibility built into its version requirements.

If you try to install python and package-a into an environment with these exact requirements:

conda create --name solver-error-environment python=3.9 package-a=2.0

You will get an error like the following:

LibMambaUnsatisfiableError: Encountered problems while solving:
  - package package-a-2.0-py310 requires python >=3.10, 
    but none of the providers can be installed

This error tells you that package-a requires python greater than or equal to version 3.10, but something is blocking that version from being installed.

​

Solving the version conflict

There are several strategies you can use to resolve a dependency version conflict.

​

Create different environments

The safest approach is to create separate conda environments for conflicting packages. Based on the example above, ask yourself “If I need Python 3.9 in this environment, can I put package-a in a different environment?”

# Create an environment for Python 3.9
conda create --name python-env python=3.9

# Create a separate environment for package-a
conda create --name second-env package-a=2.0

​

Only pin packages that require it

If specifying (or pinning) a version for package-a along with python is leading to solver errors, try only giving conda the name package-a and no version number. This will give conda’s solver more options to evaluate.

# Create an environment with only Python pinned
conda create --name python-env python=3.9 package-a

Since you haven’t specified a version number for package-a, conda will attempt to find the latest possible version that works with Python 3.9.

​

Specify package ranges

Maybe you do want some control over package-a’s version. In that case, try specifying a version range for package-a.

# Create an environment with only Python pinned
conda create --name python-env python=3.9 "package-a<2.0"

Since you’ve specified a version less than 2.0 for package-a, conda will attempt to find a version older than 2.0 that works with Python 3.9.

​

Transitive dependency conflict

Because packages often depend on several other packages, which in turn often depend on other packages, another conflict type you will likely see is a transitive dependency conflict. Transitive dependencies occur when one package depends on another, which in turn depends on a third package. For example:

In the above scenario, Package A depends on Package B and Package B depends on Package C. This makes Package C a transitive dependency of Package A.

Let’s say you want to install Package A and another package, Package D, into the same environment. Package D requires version 3.5 of Package C. For the version of Package A that you want to install, its dependency, Package B, requires version 4.0 or higher of Package C. This is an example of a transitive dependency conflict. Package A doesn’t conflict with Package D directly, but one of its dependencies does.

Try one of the strategies suggested for solving the direct dependency conflict. Transitive dependency conflicts can be more complex to solve because the conflicting package or packages might not be directly connected to the package you’re trying to install.