Filtering the dependencies to be considered during the vaadinPrepareFrontend task.

[jwn-css]

Is your feature request related to a problem? Please describe.

If the number of dependencies of the project is very large and the types of dependencies are very different, the execution of the vaadinPrepareFrontend task can take a relatively long time.

Describe the solution you'd like

Filtering on specific file names or extensions would be helpful, if only to reduce the number of dependencies to consider.

Describe alternatives you've considered

Alternatively, it is possible to partially exclude the task and execute it only when needed. However, it is then inevitably necessary that all participants in the project know exactly what the background of the task is and when one must execute the task and when not.

Additional context

It would be very helpful if this feature, if considered, would also be included in the version for the current LTS i.e. Vaadin 14.

[mstahv]

Hi,

Would you add some log output regarding the scanning that Vaadin does. Something like "Visited {} classes. Took {} ms."

You can also maybe use blacklist/whitelist to define what gets scanned:
https://vaadin.com/docs/v14/flow/integrations/spring/tutorial-spring-configuration#special-configuration-parameters

cheers,
matti

[mvysny]

Hi all, couple of thoughts:

• If I remember correctly, only zip-type files are considered (so no .dlls for example); on top of that only the .jar files are considered (files with the jar extension). Please see could not create Vfs.Dir from url, no matching UrlType #115 for more details.
• @mstahv I think that groupId+artifactId-based filtering would be more beneficial in this particular case. The vaadin.blacklisted-packages setting can not modify the classpath and thus the only option is to ignore certain Java packages; however vaadinPrepareFrontend task works in much earlier stages and we could squeeze even more performance by ignoring entire jars; also thinking in terms of dependency jars instead of packages makes more sense to me in the context of a Gradle plugin. Also I believe that's what the original reporter also wanted. @jwn-css what are your thoughts on this please?
• A combined include/exclude mechanism similar to Gradle Copy task could be used, to provide multitude of ways of configuration: https://docs.gradle.org/current/dsl/org.gradle.api.tasks.Copy.html#org.gradle.api.tasks.Copy:includes
• Would you add some log output regarding the scanning that Vaadin does. Something like "Visited {} classes. Took {} ms." - let me investigate on this.

[mvysny]

Would you add some log output regarding the scanning that Vaadin does. Something like "Visited {} classes. Took {} ms."

Unfortunately the class processing happens internally in the NodeTasks class (provided by the flow-server.jar) which then uses it internally to discover classes (for example see the FrontendWebComponentGenerator class for more details). Looks like it's impossible for the Gradle plugin itself to dump any performance stats; but we can ask the Flow team to add some kind of debugging support to the NodeTasks class @mstahv

That being said, the Gradle plugin dumps the set of jar files to be sent to NodeTasks, using the logger INFO level: Passing this classpath to NodeTasks.Builder: [flow-server.jar, ...]

[mvysny]

The API proposal: the build.gradle.kts snippets could look like the following:

vaadin {
  filterClasspath {
    include("com.vaadin:*")
    exclude("org.slf4j:*")
    exclude("commons-codec:commons-codec")
  }
}

The com.vaadin:flow-server will always be present and applied, regardless of the combination of the above. The API must work both with build.gradle and build.gradle.kts buildscripts (both Groovy and Kotlin).

@jwn-css what do you think?

[mstahv]

@caalador What do you think, would it be feasible (what Martin suggested)? I don't know if we know anything about artifactId or groupId in the right place 🤔

[mstahv]

@pleku checked briefly and it looks like it might be possible:

https://github.com/vaadin/spring/blob/master/vaadin-spring/src/main/java/com/vaadin/flow/spring/VaadinServletContextInitializer.java#L565
https://github.com/vaadin/spring/blob/master/vaadin-spring/src/main/java/com/vaadin/flow/spring/VaadinServletContextInitializer.java#L762

We should create a cleaner enhancement issue to the flow repository.

[mvysny]

@mstahv not sure what you mean 🤔 I believe the only place to implement a groupId/artifactId-based filtering is the Plugin code itself; at runtime the groupId/artifactId information is lost and not available. I am prototyping the solution code in the Plugin code itself, it should be ready by tomorrow 👍

[pleku]

@mvysny I thought the discussion was on how to make it happen for Spring Boot where it is very slow to scan things 😄

I'm afraid the plugin code itself is not a full solution for all of the scanning, but it is a start. It would make sense to make it work for all occasions and not have different API/implementation for different technologies. But maybe testing it first with the gradle plugin is the best way to get forward and then expand based on the experience there 👍🏻