WIP: Allow partially updating the track IDs

[yfukai]

This PR intends to allow partially updating the track IDs for GraphView while maintaining the overall consistency of the IDs. When users call the assign_track_id for GraphView, the track IDs for nodes outside the view are also updated to maintain the consistency.
This assumes that #150 is marged.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 83.33333% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.11%. Comparing base (58cdd0d) to head (0f4f947).

Files with missing lines	Patch %	Lines
src/tracksdata/graph/_graph_view.py	87.50%	1 Missing and 1 partial ⚠️
src/tracksdata/graph/_sql_graph.py	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #151      +/-   ##
==========================================
+ Coverage   88.03%   88.11%   +0.07%     
==========================================
  Files          49       49              
  Lines        3243     3255      +12     
  Branches      544      545       +1     
==========================================
+ Hits         2855     2868      +13     
  Misses        231      231              
+ Partials      157      156       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[JoOkuma]

Hey @yfukai,
I have seen this is a draft PR, so I'm just leaving some thoughts and not a full review.

I initially didn't implement assign_track_ids for the SQLGraph because it essentially needs the whole graph topology into memory, which is equivalent to loading it in memory as a view, which can be achieved through sql_graph.filter().subgraph().

Maybe this could be called internally by us to make SQLGraph API consistent.
filter().subgraph() copies all the content of the graph, we could have a new method for the SQLFilter that returns only the graph topology filter().rx_graph or similar.

[JoOkuma]

@yfukai, I'm leaving tomorrow for vacation, so I'll only be able to review this after September 8th

[yfukai]

@JoOkuma Hi, thanks for the comment! Yeah I was thinking something similar and I'll try to address in this PR. Enjoy your vacation!

[yfukai]

Just note for myself: I started to work on with the use cases of partially updating the track IDs in a manual correction software (usually the updates are partial).

I was planning to use the

rx_graph.filter(node_id=node_ids).subgraph().assign_track_ids()

pattern where the node_ids are the updated node IDs by manual correction, but noticed that this involves copying the node props to the filtered subgraph. Maybe I should implement something like

rx_graph.assign_track_ids(node_id=node_ids)

which make a shallow "topology" copy of the graph and the assign the IDs. I'll check if there's an option to make the shallow copy in the former option and then decide.

[yfukai]

Update: I noticed I could do

rx_graph.filter(node_id=node_ids).subgraph(node_attr_keys=[], edge_attr_keys=[]).assign_track_ids()

Maybe this is sufficient?

[yfukai]

Note for myself: Digging into the error of the benchmarking.py, I noticed what I'm implementing actually does not match with the usecase of pruning the edges from the solution... I'll probably go with the

rx_graph.assign_track_ids(node_id=node_ids)

option.

[yfukai]

I believe I messed up this PR. I'll open another one with copying a part of code from this one.