any interest in geopandas/matplotlib plotting option?

[everettsp]

Would this type of plotting function be a useful addition to swmmio? If so, I can polish it up and submit a PR. Would also make it easy to export to GeoJSON, which I see hasn't been implemented yet.

[bemcdonnell]

@everettsp, I think it would fit with SWMMIO the best. Have a look at the viz library in there. My suggestion is to make it fairly modular since we might move all viz out of that tool later

[aerispaha]

@everettsp, I like this idea too. Keep in mind though that there already is similar functionality provided in the swmmio.graphics module. With that you can get an visual representation of the model in a jupyter notebook, exported as PNG, or as an interactive map in an HTML file.

import swmmio

# create a swmmio Model object with a coordinate reference system 
model = swmmio.Model('path.inp', crs='EPSG:4326')

# create an image of the model 
swmmio.draw_model(model)

# visualize the model in an interactive map 
swmmio.create_map(model, filename="my-cool-model-map.html")

That said, the draw_model function uses Pillow for graphics, and I think it would be a lot more sensible to use matplotlib, or matplotlib via Geopandas. I also really would love to be able to have the interactive maps be compatible with Jupyter notebooks, similar to folium and/or how the Geopandas explore() method works. This is something that's been vexing me for a while.

We can also access the wealth of Geopandas / matplotlib plotting utilities by tapping into the swmmio.elements API, which aggregates multiple sections from the INP / RPT into Dataframe and Geodataframe objects. With that, we can quickly visualize models with the familiar matplotlib API. For example:

# grab an example model 
from swmmio.examples import jersey

# do some matplotlib via geopandas
ax = jersey.links.geodataframe.plot(linewidth=jersey.links.dataframe['Geom1']*3, capstyle='round')
jersey.nodes.geodataframe.plot('MaxDepth', ax=ax, markersize='InvertElev', zorder=2)
jersey.subcatchments.geodataframe.plot('PercImperv', ax=ax, alpha=0.5, zorder=1)

Similarly, we can export model elements as GeoJSON from this API:

jersey.nodes.geodataframe.to_json()

'{"type": "FeatureCollection", "features": [{"id": "J3", "type": "Feature", "properties": {"InvertElev": 6.547, "MaxDepth": 15.0, ...

Having said that, I still feel like it would be nice to make it easier to quickly visualize SWMM models, as you suggest. I like the idea of adding a plot method on the swmmio.Model object that returns a matplotlib object, and/or a folium-like interactive map that renders in Jupyter. My feeling is that we could approach this by leveraging the swmmio.elements module as described above (e.g. model.links.geodataframe.plot(), model.nodes.geodataframe.plot(), etc.)

What do you think of that? Overall, I definitely support you submitting a PR that helps move swmmio in this direction!

[everettsp]

Thanks for the detailed answer - some really good points. Everything said, I think a geopandas-based plotting and explore options would be useful features.

Currently I'm fetching some 'default' plotting args in a YAML file, e.g.:

nodes:
  junctions:
    zorder: 20
    color: blue
    linewidth: 2
  outfalls:
    zorder: 22
    color: red
    linewidth: 2

Simply iterate over that and plot with plot() or explore(). One annoying thing I ran into is that it'd be nice to have the type of each entry in swmmio.elements (i.e., whether a node is an outfall, junction, etc.), which allows a different style to be applied to each subtype of element. Currently, I'm just running this, but it's a bit inelegant:

from swmmio.defs import COMPOSITE_OBJECTS

def get_node_types(model:swmmio.Model) -> dict[str:str]:
    node_types = {}
    for sec in COMPOSITE_OBJECTS['nodes']["inp_sections"]:
        node_types.update({x:sec for x in getattr(model.inp, sec.lower()).index.tolist()})
    return node_types

Is there currently a nicer way of fetching this info, short of adding a new 'subtype' column to the swmmio.elements geodataframes?

Anyway, explore makes for a nice UI. If I'm feeling ambitious, I might tidy up some interactive timeseries results plotting tools, which I can eventually share as well.