Anaconda SciPy Dev: Part I (macOS)
This video shows how to install the Anaconda distribution of Python and perform an in-place build of SciPy from source on macOS.
Note that this effectively creates two different versions of SciPy - one is installed with Anaconda ("pre-built version"), and the other is built from the source code ("source-built version"). When you `import scipy` from a console in Spyder, which version is actually imported? That depends on the working directory of the Python interpreter. Unless the interpreter's working directory is the root directory of the source-built version, the pre-built version will get imported. However, by adding the root directory of the source-built version to the Python path in the Spyder IDE, we ensure that the source-built version gets imported regardless of the interpreter's working directory.
Note regarding build tests: The results "1 passed, 582 deselected, 17.13 seconds" were a little wonky. Normally there are thousands of tests to perform, and these take several minutes to complete. However, the master branch of my fork of the SciPy repo was not up to date, and it so happened that the test suite for this out-of-date version of the source code was in a strange state. Consequently, I got strange test results, even though this version of SciPy was perfectly usable. This is unlikely to happen if you follow this video in one session and have never created your own fork of the SciPy repo before, because when you fork (for the first time) your fork will contain the most up-to-date code. I, on the other hand, had created my fork long ago. Clicking the "Fork" button on GitHub did not create a new fork, it simply took me to my fork, which was out of date.
To ensure that you have the most recent code before building:
1) git remote add upstream git://github.com/scipy/scipy.git
2) git fetch upstream
3) git checkout master
4) git rebase upstream/master