Add should_run functionality

[akshitasure12]

This PR handles #77.

• I ran the following as a test:

NETWORKX_BACKEND_PRIORITY_ALGOS=parallel NETWORKX_FALLBACK_TO_NX=False python3 test_should_run.py

Contents of test_should_run.py :

import networkx as nx
import logging

nxl = logging.getLogger("networkx")
nxl.addHandler(logging.StreamHandler())
nxl.setLevel(logging.DEBUG)

G = nx.empty_graph(5)
print(nx.number_of_isolates(G))

Output:

Call to 'number_of_isolates' has inputs from {'networkx'} backends, and will try to use backends in the following order: ['parallel', 'networkx']
Backend 'parallel' shouldn't run `number_of_isolates` with arguments: (G=<networkx.classes.graph.Graph object at 0x1055fb6e0>), because: Fast algorithm; not worth converting
Trying next backend: 'networkx'
Call to 'isolates' has inputs from {'networkx'} backends, and will try to use backends in the following order: ['parallel', 'networkx']
Backend 'parallel' does not implement 'isolates'
Trying next backend: 'networkx'
5

[Schefflera-Arboricola]

Consider adding some unit tests for should_run-- creating a centralised test-- like test_get_chunks-- might be good here.

[Schefflera-Arboricola]

Thanks for this PR @akshitasure12 !

Left a few comments/suggestions below. LMKWYT.

Also, could we add should_run for more functions* with similar should run messages and have a single should_run-- instead of having same should_run for several algorithms? LMK if that's possible. Thanks!

*some shortest paths functions also doesn't show much speed ups-- refer timing folder for more

[akshitasure12]

Also, could we add should_run for more functions* with similar should run messages and have a single should_run-- instead of having same should_run for several algorithms? LMK if that's possible. Thanks!

I’ve added should_run as an argument to _configure_if_nx_active() and moved the various should_run helpers for different graph conditions (like number of nodes) into a separate file named should_run_policies.py(open to renaming if you have suggestions). This eliminates the need to create individual _ functions for different algorithms. Let me know what you think!

I tested it on test_should_run.py:

import networkx as nx
import logging

nxl = logging.getLogger("networkx")
nxl.addHandler(logging.StreamHandler())
nxl.setLevel(logging.DEBUG)

G = nx.empty_graph(5)
print(nx.number_of_isolates(G))

Output:

Call to 'number_of_isolates' has inputs from {'networkx'} backends, and will try to use backends in the following order: ['parallel', 'networkx']
Backend 'parallel' shouldn't run `number_of_isolates` with arguments: (G=<networkx.classes.graph.Graph object at 0x103f9d6a0>), because: Fast algorithm; skip parallel execution
Trying next backend: 'networkx'
Call to 'isolates' has inputs from {'networkx'} backends, and will try to use backends in the following order: ['parallel', 'networkx']
Backend 'parallel' does not implement 'isolates'
Trying next backend: 'networkx'
5

[Schefflera-Arboricola]

review comments from meeting:
- including should_run for if n_jobs=None or 1 or 0 for all nx-parallel algorithms-- default should run?
- Adding should_run to more algorithms-- based on performance heatmaps' trends

[akshitasure12]

@dschult The should_run functionality has been implemented, and the tests are passing. It’s ready for review-- looking forward to the feedback.
Thank you :)

[dschult]

This looks good!
I have one suggestion to remove the word "custom" in one place (to match the other removals). If leaving it was intentional then leave it. :)