Provide shaded jar with dgraph4j's dependencies #239
Conversation
|
@ryanfoxtyler Is this something Dgraph can provide? |
| tasks.named('shadowJar', ShadowJar) { | ||
| enableRelocation true | ||
| relocationPrefix "io.dgraph.dgraph4j.shaded" | ||
| relocate 'google', 'io.dgraph.dgraph4j.shaded.google' |
There was a problem hiding this comment.
@mangalaman93 looks like only those classes that're in the google package are being relocated to a new namespace. So by shading only specific dependencies and not the classes in the main dgraph4j package, we should be able to maintain compatibility with existing projects as well as have a smaller jar file-size, as already noted on Discuss.
Also, should we finalize dgrap4j-shaded as the official name for the shaded version for release ?
There was a problem hiding this comment.
The ShadowJar relocates all dependencies, but for some reason it does not automatically relocate the google package. Only with this line it also includes that package.
Classes in the io.dgraph.dgraph4j are not relocated, only dependencies.
I'd recommend to use the classifier (e.g. shaded):
implementation "io.dgraph:dgraph4j-all:24.0.0:shaded"
rather than giving it a new archive name / artifact name:
implementation "io.dgraph:dgraph4j-shaded:24.0.0"
as this better describes of what jar this is the shaded variant.
|
Thank you for the PR @EnricoMi |
Using the Dgraph4j client in an application that is bound to its own Netty dependency version like Apache Spark creates conflicts that can only be resolved by shading Dgraph4j's dependencies.
Please release the shaded jar together with your other artifacts.