Bryan,
Here is what has consistently worked, for cantera with the Matlabi interface installed (I'm mixing actual code and pseudocode, but you get the idea):
For Mac OS:
- Download anaconda with python 2.7 and 64-bit
- install scons: conda install -c conda-forge scons=2.5.1
- install cantera: conda install -c cantera cantera
- exit the shell
- open shell, invoke bash and make sure that the python path now points to anaconda
- clone cantera git
- scons build python_toolbox=full, system_sundials=n, matlab_toolbox=y, matlab_path=/Applications/MATLAB_20XX
- scons test
- scons install
- open matlab
- add path('/usr/local/something_matlab_toolbox_that_cantera_outputs_at_the_end_of_install')
- setenv('PYTHON_CMD', '~/username/anaconda2/bin/python2.7')
- setenv('PYTHONPATH', '~/usr/local/something_python_that_cantera_outputs_at_the_end_of_install')
Save the last three steps to an m-file and run every time you open matlab. Or lookup instruction to place in startup file online
For Windows, again, with the MATLAB interface functionality:
- Download anaconda with python 2.7 and 32 or 64 bit depending on OS
- open the conda shell that is now in Start
- install cantera: conda install -c cantera cantera
- open My Computer -> yada yada -> and set PYTHON_CMD to point to anaconda's python executable
- Now download the *.msi file from the cantera documentation page
- Run the cantera msi installer
- open matlab and do the three steps : add path, set python_cmd and pythonpath like in MAC OS
Save the last three steps to an m-file and run every time you open matlab. Or lookup instruction to place in startup file online
For Ubuntu:
I have not tried but will bet that equivalent steps involving Anaconda will work
You may think that some steps are redundant (e.g. in Windows why do the anaconda install and the msi install, the msi should suffice) but trust me, this does not work on all systems. I am not fully sure why, but I think it has to do with Dell's Win 10 installation. For assembled PC's and manually installing Windows via a CD/image the cantera installation instructions seem to work. There is also an issue if your user name is "John Doe" vs. "JohnDoe". The above steps work in all cases
When I said users vs. developers, I meant to say install the most stable release of cantera vs. compile from source. In Mac OS, you have no choice, but I think the distinction allows us to create OS-specific instructions for how to compile cantera. This dichotomy already exists in the instructions currently on the web-page, so I am not sure I understand why you think you will need two sets of instructions. I am in fact suggesting simplifying the steps. Why bother with *.whl files, downloading a very specific version of numpy, etc. Conda takes care of all this, as you too seem to agree with
And finally, I suggest doing away with Homebrew. It has been updated now and if someone has an existing installation of homebrew, you have to invoke sudo to fix it. Very dangerous. Not sure if a new Mac out of the box will need to do the same, but again, why bother. With Conda, you can completely do away with homebrew
If you think any of the above is good enough for updating the documentation, let me konw and I'll submit a PR. The reason I did not submit a PR is if you philosophically disagree on these steps, then I can save my effort
Thanks,
Santosh