ANTLR version mismatch in newest release (v1.11.0)
ingomueller-net opened this issue · 8 comments
When I run any query using the newest release, I get the following output:
ANTLR Tool version 4.8 used for code generation does not match the current runtime version 4.7
I have tried running on Spark 3 (using Docker) and Spark 2 (both using Docker and natively).
BTW, until now, I have not found any actual problem with this.
Thanks for catching this. This is a strange message. The repo does have ANTLR-4.7-generated files. I will investigate.
This is fixed. The issue is that Spark 3.1 needs ANTLR 4.8. From the next release on, we will provide a specific jar for Spark 3.1.
Which commit is this? Version 1.16 on Spark 2.4.8 (still) produces the following output:
ANTLR Tool version 4.8 used for code generation does not match the current runtime version 4.7ANTLR Runtime version 4.8 used for parser compilation does not match the current runtime version 4.7ANTLR Tool version 4.8 used for code generation does not match the current runtime version 4.7ANTLR Runtime version 4.8 used for parser compilation does not match the current runtime version 4.72
On Spark 3, version 1.16 works fine. For 1.15, Spark 3 produces a warning and Spark 2 doesn't. Maybe Spark 2 and Spark 3 need different versions? (Though it might not be worth spending too much time on Spark 2.)
I have recompiled all the jars. Only the -for-spark-2 jar should be used with Spark 2.4.8. And -for-spark-3.0 should be used with Spark 3.0.*.
Version 1.15 is known to produce a warning for Spark 3.1.
The current releases now provide a jar for each Spark minor version and the ANTLR version is set appropriately.
@ingomueller-net do you agree if we close this issue?