caiwei-ebay commented on pull request #158: URL: https://github.com/apache/maven-resolver/pull/158#issuecomment-1070824708
> @caiwei-ebay @michael-o @ibabiankou @jebeaudet IMO, this PR is good to be merged, so let's merge it. > > I personally am not quite convinced about this (BFS + skipper), I'd need more tests, and would like to see how these "scale" as well (test on project with 100, 200, 400 modules and so on). @michael-o @cstamas Actually Skipper does not help with multi module project. Maven builds module one by one, this means Maven will call [collectDependencies](https://github.com/bfs-skip/maven-resolver/blob/master/maven-resolver-impl/src/main/java/org/eclipse/aether/internal/impl/collect/DefaultDependencyCollector.java#L166) one by one for each module, and thus the cache (Args.pool, not a static one) won't be reused among the resolution of multiple modules. This means if we can reuse Args.pool (make map static & concurrent safe [here](https://github.com/bfs-skip/maven-resolver/blob/master/maven-resolver-impl/src/main/java/org/eclipse/aether/internal/impl/collect/DataPool.java#L72)), it should be faster when building multiple modules. I would like to try that in near future, maybe after @ibabiankou's PR. Skipper only helps avoid unnecessary duplicate node resolution, it does least help for maven projects developed by maven experts like Spring as they have managed dependency versions very well and exclusions are rarely used. To print dependency:tree of entire spring-boot project (with large number of modules), it would take just 20s with my windows desktop. To be frank, it is not a common case in Opensource or people using Opensource project, this is why I say this skipper solution is only helpful for large projects with complex dependencies (micro service tech is not employed, people rely on libraries developed by other team to call APIs provided by them). The tests [here](https://github.com/bfs-skip/maven-resolver/blob/master/maven-resolver-impl/src/test/java/org/eclipse/aether/internal/impl/collect/DefaultDependencyCollectorUseSkipTest.java#L83) actually demonstrate the case: same node with different exclusions, the 2nd one will be skipped as it is the loser node. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@maven.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
