Add support for automatic dict and set attributes in module io

[ncotie]

Proposing some extra code support in module io for decoding stringified lists, dicts and sets automatically when running load_graphml(), allowing users to avoid having to specify ast.literal_eval as dtype for customized (non-OSM) node and/or edge attributes.

The proposed code expands existing code from _convert_edge_attr_types() that handles stringified lists as edge attributes.

This is linked to Issue #1067.

Changes are made in _convert_node_attr_types() and _convert_edge_attr_types(), adding the following to both:

    # check for stringified dicts or sets: both can be converted
    # to python objects using ast.literal_eval()
    if value.startswith("{") and value.endswith("}"):
        with contextlib.suppress(SyntaxError, ValueError):
            data[attr] = ast.literal_eval(value)

Tested with toy networks as in the issue description, for list, set and dict node and edge attributes.

[gboeing]

We previously only ran literal_eval on edge attributes. Is there a use case for node attributes? OSMnx does not produce any node attributes of type list, dict, or set.

[ncotie]

I can describe my usage, if it makes sense to you. It's clearly all custom attributes, as I mentioned in the issue description, so not necessarily OSMnx scope.

I'm doing 15-minute city metrics, based on OSMnx with a lot of my own code on top. I fetch a walking network via OSMnx, then integrate amenity POI nodes into it, adding linking edges with lengths. The network nodes, apart from the amenities, are considered to be "residential locations", without getting into filtering details. Processing then obtains, among other things, shortest path distances, per residential node, to each amenity node. These are stored per node into a dict of dicts attribute keyed by amenity category and amenity node.
Accessible amenities are counted per residential node and edge, also, which allows for producing isochrone buffered geometries, for example.

It's true that I could have managed these outside the graph object using dicts, but I had wanted to make the work graph-oriented, and I believe it simplifies things, avoiding the need to keep separate objects consistent when doing things like truncating, etc, etc.

[gboeing]

CI tests are failing. You can check the failure here and correct it manually, or see the tests documentation.

[gboeing]

Tests are failing. Please lint and test your code locally before pushing. See the tests folder.

[ncotie]

Geoff, apologies but the push earlier today surprised me, I didn't know that when I pushed to my own fork that it would update here immediately... Impressive integration in git...
As noted in the last merge, I created a local env from /tests/env-ci.yml, and looked to see how to test it. Called format.sh and lint-test.sh, which made two formatting changes automatically, but I saw that by coincidence, and I can't find any other results from the local test, so I can't tell if it passes properly now or not.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.83%. Comparing base (dfa7226) to head (3ff4730).

Additional details and impacted files

@@           Coverage Diff           @@
##               io    #1075   +/-   ##
=======================================
  Coverage   97.83%   97.83%           
=======================================
  Files          27       27           
  Lines        2493     2497    +4     
=======================================
+ Hits         2439     2443    +4     
  Misses         54       54           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ncotie]

Geoff, tell me please if I should be doing anything specific with the PR now. I'm assuming the "review" is for you to do and that there's nothing needed from me for now. The "1 change requested" tagged just above is I think already commented in the discussion earlier, and I don't believe I need to take any action on it. If that's not correct, please let me know.

[ncotie]

Thanks Geoff!