Codecov Report

Merging #320 into master will increase coverage by 0.05%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #320      +/-   ##
==========================================
+ Coverage   91.17%   91.23%   +0.05%     
==========================================
  Files          18       18              
  Lines         873      878       +5     
  Branches       73       73              
==========================================
+ Hits          796      801       +5     
  Misses         77       77

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 89312f3...3dac6e7. Read the comment docs.

@mengxr @felixcheung any thoughts on this one.

This is very helpful, any updates?

Hey @rjurney,

Not sure of your status or involvement in this project but is there a chance you could push this through? 🤞

AFAICT this would unlock consistent component IDs across Connected Component executions for components that have unchanged nodes.

@followingell I like this, but my machine is presently in an unhappy state as it relates to this project as I dip my toes in Spark proper waters. What testing have you done?

@followingell I like this, but my machine is presently in an unhappy state as it relates to this project as I dip my toes in Spark proper waters. What testing have you done?

Unfortunately none. I don't have access at the moment to a machine where I could test this without issues.

Okay, give me a few days and I’ll test it.

We have a lot of pending PRs related to Connected Components. I'm going to push this one forward after merging #552

Sounds good. I like this PR.

I will try to resolve conflicts.

@architch Thanks for resolving conflicts. You can use an existing pre-commit hook to fix style (or just run ./build/sbt root/scalafmtAll)

Thanks for the contribution, @architch !

@architch I want to notify you that after some discussions we made a decision to keep as the default the behavior before your PR: #632

To use min vertex label as component label you will need to explicitly modify the parameter. You can use the config spark.graphframes.useLabelsAsComponents (set it to true) or API (useLabelsAsComponents: bool in Python and setUseLabelsAsComponents(value: Boolean) in JVM).

Sorry about it, but do a breaking change will require too much work from downstream projects. I think for users it won't be hard to set config or argument to true and have a behavior from your PR.

This seems like a happy medium :)

Thanks @SemyonSinchenko for the info. It makes sense to make it configurable.