Remove scalalogging-slf4j from dependencies#297
Remove scalalogging-slf4j from dependencies#297felixcheung merged 2 commits intographframes:masterfrom
Conversation
Codecov Report
@@ Coverage Diff @@
## master #297 +/- ##
==========================================
- Coverage 90.9% 90.83% -0.08%
==========================================
Files 18 19 +1
Lines 803 818 +15
Branches 69 69
==========================================
+ Hits 730 743 +13
- Misses 73 75 +2
Continue to review full report at Codecov.
|
d76c37d to
b95df05
Compare
b95df05 to
dc8e3ad
Compare
build.sbt
Outdated
| // e.g. spDependencies += "databricks/spark-avro:0.1" | ||
|
|
||
| libraryDependencies += "org.scalatest" %% "scalatest" % defaultScalaTestVer % "test" | ||
| libraryDependencies += "org.slf4j" % "slf4j-api" % "1.7.21" |
There was a problem hiding this comment.
is there a reason for .21?
spark is running with 1.7.16
https://github.com/apache/spark/blob/9b6baeb7b9f73e1a38581f481ea7232db712deb8/pom.xml#L119
There was a problem hiding this comment.
@felixcheung thanks, I totally missed that part, which is checking which version Spark is running. I've just picked up the version which I've always used
Use slf4j-api 1.7.16
|
I don't know why Travis build failed. It failed 2 times one the same commit for different reasons
I will close and reopen this PR one more time to trigger Travis build to run |
|
Okay, it passed |
|
thanks! |
|
Great to see this change. Unfortunately shading scala-logging 2.x didn't work for us, I presume because 2.x hardcodes package paths in in macros / quasi-quotes. It would be great to see a new release of graphframes with this change. Is there a snapshots repository? |
|
@tmoschou @felixcheung perhaps it can be cherry-picked to 0.6.x release branch? |
|
I don't think we would do a 0.6.1 release? |
* Remove scalalogging-slf4j from dependencies * Update build.sbt Use slf4j-api 1.7.16
In this PR I've removed scala-logging-slf4j in favor of bare slf4j-api.
GraphFrames depends on scala-logging v.2.1.2 which is incompatible with later scala-logging 3.x. Thus if other dependencies introduce scala-logging 3.x you get
java.lang.IncompatibleClassChangeError. So you have to use shading to overcome this issue.Simply upgrading scala-logging to 3.x will break Scala 2.10 cross-compilation for GraphFrames because 3.x is not available for Scala 2.10. So maybe it's better to remove scala-logging dependency in favor of bare slf4j-api, just like in Spark (see this commit in Spark)