Contributor and repository graph analysis #175
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
View / edit / reply to this conversation on ReviewNB JamesKunstle commented on 2022-10-31T16:43:21Z Line #20. issue_contrib.columns =['Repo ID', 'Git', 'Issue Authors', 'Issue ID'] Purely aesthetic request: Name columns in the 'Select' statement for an easy 1:1 mental mapping between names in the table. Applies to other queries. Just a little nicer to read but doesn't impact correctness. oindrillac commented on 2022-11-07T15:35:52Z I agree with this. Will rename this, cleaner than having spaces in the names. oindrillac commented on 2022-11-14T17:32:36Z Modified the column names @JamesKunstle |
|
View / edit / reply to this conversation on ReviewNB JamesKunstle commented on 2022-10-31T16:43:22Z Line #8. ca.cntrb_id, Something tricky: commits have both Authors and Committers who are not necessarily the same person. It certainly isn't incorrect to use only the committer as this data-point, but perhaps there's something interesting to be used later by considering both the 'Committer' and the 'Author' because we'd pay attention to both contributor's work. oindrillac commented on 2022-11-07T15:36:50Z That's a very good point. Will see how we can extend this work to include the Authors as well. oindrillac commented on 2022-11-14T18:05:54Z Created an issue to address this https://github.com//issues/177 |
|
View / edit / reply to this conversation on ReviewNB JamesKunstle commented on 2022-10-31T16:43:23Z From here to the bottom of this notebook, I can imagine the steps that are being taken (e.g. above, I think that's a sparse edge matrix between commit creators and repos) but I would really appreciate some in-depth annotation of your process. What I'm seeing throughout this notebook is phenomenal work but I don't want to misinterpret your results by injecting an incorrect assumption.
essentially: would love more annotation on phases of working on this network representation. oindrillac commented on 2022-11-07T15:46:28Z Yes absolutely. Intend to add more explanation, examples and samples with this https://github.com//issues/174 but even at this phase, going to add some in-depth comments/annotation to improve readability oindrillac commented on 2022-11-14T18:06:19Z Added some explanation for these transformations and the graphs. |
|
View / edit / reply to this conversation on ReviewNB JamesKunstle commented on 2022-10-31T16:43:24Z this is such an awesome view! very cool to see the inter-connectedness of the 704 centroid w.r.t the relatively sparse connectivity of the other centroids. oindrillac commented on 2022-11-07T15:57:26Z Right! That makes me think, it would be cool to dynamic plotly like graphs from these where we can hover over these to see metadata on these nodes etc cdolfi commented on 2022-11-11T16:27:17Z Is the coloring here the same as above? blue for repo and yellow for contributor? |
|
View / edit / reply to this conversation on ReviewNB JamesKunstle commented on 2022-10-31T16:43:25Z Is this a fully-connected graph? e.g. # edges = (#nodes * (#nodes - 1)) I think. I'd assume that it would be- the weightedness of the edges would be interesting to see -- where are the largest contributor population intersection between projects? oindrillac commented on 2022-11-07T15:55:54Z No this is not a complete (fully-connected) graph. All nodes do not have a direct connection to each node. But each node has atleast one connection. Similar to a social media platform where the requirement to be on the network would be to have atleast one connection. For example - see project node '27034' and '27032', they are not connected as they do not have any shared contributors.
To your previous comment, along with details on how we got to this graph, I will also add some comments on how this graph can be interpreted. oindrillac commented on 2022-11-07T15:58:42Z Would be cool to have dynamic graph here too, so one can hover over the edges to see "which contributors" constitute these shared contributors. Would be so nice to be able to drill down into those views. cdolfi commented on 2022-11-11T16:29:25Z For the edge lengths here what do they specifically correlate to? oindrillac commented on 2022-11-14T18:07:24Z @cdolfi the edge lengths here correlate to the number of shared contributions amongst them. Added some explanations for the graph below for more clarity. |
|
Overall: I want to re-review this thoroughly with more annotation so I can follow your process very closely. It's so so cool to see these results as they are though, the interconnectedness is fascinating to see. Amazing work ya'll. |
|
I agree with this. Will rename this, cleaner than having spaces in the names. View entire conversation on ReviewNB |
|
That's a very good point. Will see how we can extend this work to include the Authors as well. View entire conversation on ReviewNB |
|
Yes absolutely. Intend to add more explanation, examples and samples with this https://github.com//issues/174 but even at this phase, going to add some in-depth comments/annotation to improve readability View entire conversation on ReviewNB |
|
No this is not a complete (fully-connected) graph. All nodes do not have a direct connection to each node. But each node has atleast one connection. Imagine a social media platform where there is a requirement that, to be on the network you need to have atleast one connection, this would be similar to such a graph. For example - see project node '27034' and '27032', they are not connected as they do not have any shared contributors.
To your previous comment, along with details on how we got to this graph, I will also add some comments on how this graph can be interpreted. View entire conversation on ReviewNB |
|
Right! That makes me think, it would be cool to dynamic plotly like graphs from these where we can hover over these to see metadata on these nodes etc View entire conversation on ReviewNB |
|
Would be cool to have dynamic graph here too, so one can hover over the edges to see "which contributors" constitute these shared contributors. Would be so nice to be able to drill down into those views. View entire conversation on ReviewNB |
|
Is the coloring here the same as above? blue for repo and yellow for contributor? View entire conversation on ReviewNB |
|
For the edge lengths here what do they specifically correlate to? View entire conversation on ReviewNB |
Co-authored-by: Hema Veeradhi <hveeradh@redhat.com>
oindrillac
force-pushed
the
graph-analysis
branch
from
7b992da to
0e9c7b5
Compare
|
Modified the column names @JamesKunstle View entire conversation on ReviewNB |
|
Created an issue to address this https://github.com//issues/177 View entire conversation on ReviewNB |
|
Added some explanation for these transformations and the graphs. View entire conversation on ReviewNB |
|
@cdolfi the edge lengths here correlate to the number of shared contributions amongst them. Added some explanations for the graph below for more clarity. View entire conversation on ReviewNB |
|
@cdolfi @JamesKunstle added some more annotations and explanations to the notebook. Please take a look |
Yes, thats correct. Added legends to each graph for clarity |
|
Just want to bump this up. Comments above were addressed and this is ready for another look. |
|
big lgtm from me |
In this PR, we created initial graph representations of existing open source GitHub repositories falling under a certain category using NetworkX.
related #168
We created 2 type of graph representations from the same data: