Better notebook support for visualizing DAGs

[skrawcz]

Is your feature request related to a problem? Please describe.
If you are using Hamilton in a jupyter notebook, we should enable creating a visualization easily when you request a DAG visualized.

Describe the solution you'd like
In simplest terms, doing something like dask does for visualizing its DAGs in a notebook would be sweet.

Requirements:

1. Able to display in a notebook.
2. Able to handle really large graphs -- or some way to help people wrangle them to display.

e.g.

dr = driver.Driver(...)
dr.visualize_execution(...)
# this creates an image that is displayed

Thought - the solution here might be as simple as returning the graphviz object...

Perhaps having a default print out of a driver object could also be useful?

dr = driver.Driver(...)
dr
# this prints out some information about the DAG created -- perhaps even a rough visualization, much like dask does?

Describe alternatives you've considered
You save the image, and then load it -- I don't believe this will be the best experience though:

from IPython.display import Image
...
dr.visualize_execution(..., 'my_dag.png', ...)
...
Image('my_dag.png')
# this would then display

Additional context
For ideas/inspiration:

• https://docs.dask.org/en/stable/graphviz.html?highlight=visualize
• Graphviz docs

[elijahbenizzy]

On my phone but I've prototyped some examples

Non-interactive

Interactive

These run after the hello world view in the notebook. Planning to add examples.

[skrawcz]

Nice -- but is there a way we just return the graphviz/networkx object?

[elijahbenizzy]

Nice -- but is there a way we just return the graphviz/networkx object?

Currently it doesn't but I think we should change it to-- makes the above code much easier to work with.

[skrawcz]

Implemented in #286