Move CI tests from Travis to GitHub

[dhruvmanila]

Proposal for using pipenv:

Goal: Add Pipfile and Pipfile.lock (keeping requirements.txt) and speed up our builds

pipenv is a tool which combines the power of pip and venv. My main goal for using this is only to speed up our builds. Why? Take a look at: https://github.com/dhruvmanila/algorithms-keeper/runs/1394249008?check_suite_focus=true. The entire virtual environment directory is cached and so we just skip the installing part which saves approximately 60 seconds in every run.

What happens if we update our dependencies?

The cache works based on a key and the key will depend on one of the dependency file which in this case can be Pipfile or Pipfile.lock (lock file is preferable). When we update any dependency, we update the lock file which means the key will change and thus the cache won't be found. In this case, the virtual environment will be created and cached again, thus updating the cache.

@cclauss Please take a look at the proposal and if there are any doubts, do ask. DO NOT MERGE this PR yet.

Move CI tests to GitHub Actions

Describe your change:

• Test changes

Checklist:

• I have read CONTRIBUTING.md.
• This pull request is all my own work -- I have not plagiarized.
• I know that pull requests will not be merged if they fail the automated tests.

[cclauss]

I am not a fan of pipenv... Pining dependencies are for projects that need reliable builds on a rigidly fixed set of dependencies. This is for production software than is heavily supported and must not break.

I much prefer our dependencies to float, to always be the latest-and-greatest. Locking in pined dependencies runs contrary to that goal. If a new release of Python or one of our dependencies breaks our algorithms, I would like to know that as soon as our continuous integration is run.

So, I am OK using pipenv if we never put version numbers in Pipfile but if we need to manually add/change version numbers in that file, then I would be interested to avoid it.

[dhruvmanila]

To avoid using version numbers, we just put an asterisk symbol like so: numpy = "*" but if that is the case then I should say that we will have to manually update the lock file using the command pipenv lock or we could just avoid the lock file and get every information from the Pipfile itself.

We will just add the Pipfile file and use that to install and update all our dependencies.

Do you want to try this in a new branch?

[cclauss]

What if we continue to use requirements.txt and trust the cache if $(pip freeze) does not change.

[dhruvmanila]

There's nothing to worry about if we are using requirements.txt as we will just be using pip to install the dependencies.

If that is the case then please take a look at the workflow file in this PR which is basically a mirror of Travis configuration except for the Travis Buddy notification part which I think I can incorporate into our bot. Only if it ever gets accepted 😕

[amaank404]

Awesome 😄

[dhruvmanila]

@cclauss this seems good to go, can you take a quick look at the changes?

[cclauss]

Nice!!