added traversal algorithm to nx_parallel

[RohitP2005]

I added traversal algorithm with dfs and bfs which is _dispatchable . Kindly suggest further corrections

[RohitP2005]

@stefanv can u help with what to be done next

[dschult]

This is a lot of new functions!
But I can't tell if it is doing anything that networkx is not already doing.
For example the breadth_first_search.py functions look copied straight fro networkx. That code is not parallel. I think the idea is to split up the "embarrasing parallelizable" loop and do the iterations of that loop on different processors.

Maybe just focus on one function and get that working... Then you can do the others.
To get an idea of how this works look at the other functions already in nx-parallel. For example, betweenness_centrality.

[RohitP2005]

This is a lot of new functions! But I can't tell if it is doing anything that networkx is not already doing. For example the breadth_first_search.py functions look copied straight fro networkx. That code is not parallel. I think the idea is to split up the "embarrasing parallelizable" loop and do the iterations of that loop on different processors.

Maybe just focus on one function and get that working... Then you can do the others. To get an idea of how this works look at the other functions already in nx-parallel. For example, betweenness_centrality.

Sorry about that , i will look into it asap , thanks for the help

[RohitP2005]

@dschult
Thanks so much for your detailed feedback! You're absolutely right—it’s better to focus on implementing and testing one function thoroughly before moving on to others.

I’ve removed the DFS algorithm for now and added a BFS algorithm that is parallelized for improved performance.

Here’s an example of how to use the new parallel_bfs function:

# Create a sample graph
G = nx.erdos_renyi_graph(100, 0.05)

# Perform parallel BFS
bfs_result = parallel_bfs(G, source=0)
print("BFS Traversal Order:", bfs_result)

[Schefflera-Arboricola]

Thanks @RohitP2005 for the PR! Could you please follow this checklist for adding a new algorithm to the project - https://github.com/networkx/nx-parallel/blob/main/CONTRIBUTING.md#general-guidelines-on-adding-a-new-algorithm and you can also look at the existing algorithms in the repository to get a better idea! If any of these points in the checklist are not clear then please let us know. Thanks!

[RohitP2005]

@Schefflera-Arboricola Yeah sure , i will look into it.
I will make sure the new algorithm added meets your expectations

[Schefflera-Arboricola]

Thanks for all the work @RohitP2005 !

I haven't looked at the core implementation in this PR too closely.. but since BFS is not an embarrassingly parallel algorithm, I don't think its parallel implementation could look this simple 😅. If you are new to parallel graph algorithms then I'd suggest you consider adding an embarrassingly parallel algorithm instead (see - #82). BFS for starters might be not be great.

Thanks, and sorry for the delayed reviews :)

[Schefflera-Arboricola]

n_jobs is included in the configurations. so no need to include it as a parameter here. (see - https://github.com/networkx/nx-parallel/blob/main/Config.md)

[Schefflera-Arboricola]

the names of the functions in the main networkx project and in this nx-parallel project are usually kept same. (see - https://github.com/networkx/nx-parallel/blob/main/CONTRIBUTING.md#general-guidelines-on-adding-a-new-algorithm)

[Schefflera-Arboricola]

Suggested change

from .depth_first_search import *

[Schefflera-Arboricola]

you don't need to document the parameters that are already documented in the main networkx repo :)