Refactor the DG #25
Refactor the DG #25
Conversation
|
Have you done some real-world benchmarks? |
|
It seems to keep most things the same, but makes rubygems/bundler#3803 over an order of magnitude faster |
| end | ||
| attr_accessor :incoming_edges | ||
| # graph.edges.select { |e| e.destination.equal?(self) } | ||
| # end |
|
In general, I can believe this is faster on huuuuuuge graphs because it looks like this creates less sets and has less object churn? But my random guesses seem less important than the benchmarks showing that this is faster for the cases we care about without slowing down our other benchmark cases. |
|
OK, will ship master as a 0.3.1, and if I get a 👍 on this, will add some changelog entries and make this 0.4.0. I'm actually very happy with these changes, as it makes the dependency graph way simpler conceptually |
|
👍 from me. |
|
It looks good to me 👍 Would be nice if you could include some before/after numbers in the CHANGELOG. |
|
@alloy the numbers are basically the same in all of the benchmarks I have, but it sped up that linked to issue by 25x |
|
That seems like a very impressive number to include imo :) |
Does this look any faster?