Add a load_graph attribute to BasePartitioner

[csebastiao]

Add an attribute load_graph=None where we can load a networkX graph (or a graphml) into the BasePartitioner. As of now, can only either download it straight from OSM, or need to already have the graph inside the Config.GRAPH_DIR folder with the name {city_name}.graphml who need to have already some metrics computed on.

Necessary for any graph not coming from a simple Nominatim call.

[csebastiao]

Since the graph might not "work" as you could pass any networkX graph, need to test if the graph is following the necessary format to use BasePartitioner.

[cbueth]

Thanks @csebastiao for creating this issue. As it has come back up now we will reactivate it, adding the ability to import a graph independently of the automatic downloading. Has there been any updates from your side, @csebastiao?
When importing there are preprocessing steps that would need to be done to the graph and we might save a prepared copy of it. This way the expensive and static preprocessing (i.e. population distribution) is independent of the partitioning where a user might like to change parameters and experiment with the outcomes.
If someone might like to work on it, please let us know here in a comment. I'm happy to help with the design decisions and guidance.

[juanfonsecaLS1]

I have just created a draft PR #128 with an alternative solution for this. A separate functions compiles the minimum preprocessing steps to prepare the graph, which can be then used through the graph argument of the BasePartioner. Here is an example of the use.

import osmnx as ox
import networkx as nx
import superblockify as sb

# Manual download of a road network graph
point_city = (48.8577441884716, 2.338998102824722)
distance = 2500
G_raw = ox.graph_from_point(center_point=point_city, dist=distance, network_type='drive')

# Preprocessing function 
G_processed = sb.preprocess_graph(G_raw)


# Running the partionner on the graph 
part = sb.ResidentialPartitioner(
    name="Paris_test",
    city_name="Paris",
    unit="time",
    graph=G_processed,
)

part.run(
    calculate_metrics=True,
    make_plots=True,
    replace_max_speeds=False,
)

I have not included any tests yet, but I will do after we review this initial iteration. Let me know your thoughts on this @cbueth @csebastiao

[juanfonsecaLS1]

@mszell

[cbueth]

Thanks for working on this issue, creating the PR. You can find my comments on the pull request itself.