Does not build with pip-compile because of dependency resolution with six

[riordan]

Hi there! 👋 This feels like a successor to #678.

I'm trying to use pip-compile to create my own deterministic installs that include pydatalab. Unfortunately this fails due to the way that six is currently being pinned to 1.1.0:

    # six needs to be pinned to 1.10.0 to work-around an apache_beam bug:
    # https://stackoverflow.com/questions/46300173/import-apache-beam-metaclass-conflict
    'six==1.10.0',

When I try to run pip-compile with datalab as its only requirement on Python 3.6.8, I get this:

➜  compile-test cat <<EOF > example.txt
datalab
EOF
➜  compile-test pip-compile example.txt -o requirements.txt
Could not find a version that matches six<2dev,==1.10.0,>=1.10.0,>=1.11.0,>=1.5,>=1.6.1,>=1.9
Tried: 0.9.0, 0.9.1, 0.9.2, 1.0.0, 1.1.0, 1.2.0, 1.3.0, 1.4.0, 1.4.1, 1.5.0, 1.5.0, 1.5.1, 1.5.1, 1.5.2, 1.5.2, 1.6.0, 1.6.0, 1.6.1, 1.6.1, 1.7.0, 1.7.0, 1.7.1, 1.7.1, 1.7.2, 1.7.2, 1.7.3, 1.7.3, 1.8.0, 1.8.0, 1.9.0, 1.9.0, 1.10.0, 1.10.0, 1.11.0, 1.11.0, 1.12.0, 1.12.0
Skipped pre-versions: 1.0b1
There are incompatible versions in the resolved dependencies.

This looks like the minimum base versions of pydatalab's other dependencies have higher six requirements than pydatalab. I can't tell if the beam issue referenced is resolved, but if so, it would be fantastic to free this from six==1.1.0.

[jblogin]

My team uses a popular package management tool called pipenv, and it fails with the same issue noted above.

You can reproduce by creating a fresh pipenv project with

pipenv install datalab==1.1.4

More specifically, the issue is with the jsonschema subdependency. If you try to install the latest appropriate version of that (3.0.1), you'll find that jsonschema requires six >= 1.11.0