Add a 2020 version of the JDBC post #7
Conversation
| classpath = os.path.join(os.getcwd(), "all-jar/target/drill-odbc-0.2-SNAPSHOT-jar-with-dependencies.jar") | ||
| jpype.startJVM(jpype.getDefaultJVMPath(), f"-Djava.class.path={classpath}") |
There was a problem hiding this comment.
This is still using a fairly old pattern. Directly setting up the classpath is still acceptable, but has the risk that if you need more than once class element you need to use the system separator which is different between windows and linux.
I would recommend the the code be changed to
classpath = os.path.join(os.getcwd(), "all-jar/target/drill-odbc-0.2-SNAPSHOT-jar-with-dependencies.jar")
jpype.startJVM(classpath=classpath)
The classpath argument can either be a string or a list of strings. Wildcards are accepted. This will automatically handle the separator if more than one is provided.
The other option is to use addClassPath. You can call addClassPath either before or after the startJVM statement which is great if you already have the JVM open and want to then add the library to it after. There may still be some interactions on this as I haven't tested it extensively with jdbc.
There was a problem hiding this comment.
addClassPath didn't work after the JVM was started. Then the driver couldn't be found, I'll open an issue with a reproducible example.
| java.lang.NoClassDefFoundError: java.lang.NoClassDefFoundError: oadd/org/apache/drill/exec/store/StoragePluginRegistry | ||
| ``` |
There was a problem hiding this comment.
JDBC has some strange behaviors as they do stuff with reflection that causes loading problems. I haven't seen this one yet though I suspect it would be resolved with a JClass statement before the connection is established. I would recommend putting in a ticket so that this one doesn't get lost.
There was a problem hiding this comment.
Adding a JClass statement didn't help, I'll also open a reproducible issue.
FYI @Thrameos , @mlondschien