Linux下Conda装旧版Matplotlib遇KeyError及PrimerProspector安装求助

Let’s work through this problem step by step—you’re running Ubuntu Server 16.04.3 LTS with Conda 4.4.8, set up a Python 2.6 environment to handle PrimerProspector’s development version dependencies, added the conda-forge channel, but keep hitting a KeyError when trying to install an older Matplotlib. Here’s how to resolve this:

1. Clear Conda Cache & Fix Channel Priorities #

The KeyError usually comes from corrupted package indexes or misordered channels. Let’s reset that first:

• Wipe all cached packages and stale index data to eliminate outdated references:

  conda clean --all -y
  

• Configure channels to prioritize conda-forge (it maintains more legacy packages for old Python versions than the default channel), and enforce strict priority:

  conda config --add channels conda-forge
  conda config --add channels defaults
  conda config --set channel_priority strict
  

  Note: Python 2.6 is end-of-life, so conda-forge’s support is limited to older build sets. Add the cf201901 label (the last conda-forge release line supporting Python 2.6) to access compatible packages:

  conda config --add channels conda-forge/label/cf201901
  

2. Install a Matplotlib Version Compatible with Python 2.6 #

The final Matplotlib version that supports Python 2.6 is 1.5.3. Install it explicitly to avoid version mismatches:

• First try via Conda (once your channels are fixed):

  conda install matplotlib=1.5.3 -y
  

• If you still get a KeyError, switch to pip—Conda’s index may no longer host the exact build for your system/Python combination:

  pip install matplotlib==1.5.3
  

3. Lock in Compatible PyCogent & Numpy Versions #

Python 2.6 requires specific dependency versions to work correctly:

• For Numpy: Install the last version supporting Python 2.6 (1.11.3):

  conda install numpy=1.11.3 -y
  

• For PyCogent: Use version 1.9 (the final release that supports Python 2.6):

  pip install pycogent==1.9
  

4. Upgrade Conda to a Python 2.6-Supported Version #

Your current Conda 4.4.8 is outdated and may have bugs handling legacy Python 2.6 environments. Upgrade to the last Conda version that supports Python 2.6 (4.6.14):

conda install conda=4.6.14 -y

This will improve compatibility with older package indexes and reduce unexpected errors like KeyError.

Why This Works #

The KeyError happens when Conda can’t find a package build that matches your Python version, OS, and architecture. By cleaning the cache, prioritizing legacy-friendly channels, and explicitly specifying versions known to support Python 2.6, you eliminate the mismatches causing the error. Falling back to pip is a reliable backup when Conda’s index no longer hosts the required builds.

内容的提问来源于stack exchange，提问作者FM Kerckhof