Easier visualisation of dependency graph?

[maxalbert]

When looking at #784, I realised that calculate_dependency_graph produces a networkx graph of query dependencies but doesn't help much with actually visualising it. The docstring demonstrates how to export a .dot file, but this needs the external graphviz tool to actually visualise it (e.g. by converting to pdf).

It would be helpful to be able to draw dependency graph directly in a Jupyter notebook. Here is a useful snippet to export a dependency graph to a SVG, which can be rendered directly in the notebook:

import flowmachine
import networkx as nx
from flowmachine.features import daily_location
from flowmachine.utils import calculate_dependency_graph
from io import BytesIO
from IPython.display import SVG

flowmachine.connect(flowdb_user="flowdb", flowdb_password="flowflow", redis_password="fm_redis")

dl = daily_location(date="2016-01-01")

G = calculate_dependency_graph(dl)
A = nx.nx_agraph.to_agraph(G)
svg_str = BytesIO()
A.draw(svg_str, format="svg", prog="dot")
svg_str = svg_str.getvalue().decode("utf8")
svg = SVG(svg_str)

Would this be useful to provide as a helper in flowmachine? We probably want to avoid getting too fancy with dependency graph visualisation, but this seems like a useful addition. Thoughts?

[greenape]

Could just update the example usage with this?

[maxalbert]

Yeah, good point. If it turns out useful and we end up copying & pasting a lot we can still wrap it up in a helper function or class.