Hi, all:
I've been experimenting with the use of Conda to install Pyomo and other Python package that can be leveraged by Pyomo.
For example, if Conda is installed, then you can run the following command to install Pyomo:
This installs Pyomo from the conda-forge channel, which is where I've been setting up Conda packages. Similarly, the following command installs a variety of other Python packages that Pyomo relies on (pyyaml, numpy, openpyxl, etc):
Before advertising this capability to end-users, I'd like to get some input on its use by developers. Can you guys try out this installation mechanism and give me some feedback?
Conda is installed with the Anaconda Python distribution, but not with other Python distros. I've had occasionally difficulties installing and updating Conda with 'pip'. I'm curious if anyone else has similar problems.
FYI, the pyomo.extras Conda package is particularly useful for end-users. Currently, users can use pip to do a similar installation:
pip install pyomo.extras
But this is MUCH more fragile than Conda. If any package fails to install, then pip terminates. And PyPi doesn't ensure that packages are available the same way that Conda does. Hence, pip is inherently more fragile.
My expectation is that we'll be able to advertise a single installation command for users, after which many Pyomo extensions will work. For example, after installing
the NEOS solvers work, YAML files work, you can load data from Excel files, etc.
--Bill