How do we want to handle continuous integration checks (Docker, Travis, etc.)

[rgerkin]

What do we want to handle in Travis-CI? Testing of the Docker containers? Testing of the code from inside the containers? Testing of the code from a regular build? Which code? The sophisticated parallel computationally-intensive stuff with lots of requirements, or just the simple testing examples?

[rgerkin]

Current plan:
The .travis.yml file will install conda, install the package itself, and then run the unit tests. Docker will not be used for this kind of commit-by-commit testing of NeuronUnit. The Docker stacks will be maintained independently at scidash/docker-stacks.
The requirements.txt file will consist only of a single line (".") which points to setup.py, which is where all the dependencies will be described.
All installation should be done via pip install --allow-dependency-links . or pip install -e --allow-dependency-links .

[rgerkin]

Now with pip 18.1, we can just pip install -e . and it will handle the PEP 508 URL dependencies that we need. No need for Docker stacks right now; the Docker images can be test separately.