parquet-testing icon indicating copy to clipboard operation
parquet-testing copied to clipboard

Can't run ParquetMR tests in IDEs

Open asfimport opened this issue 3 years ago • 0 comments

ParquetMR contains a suite of self-tests. When one of those self-tests fails, it would be nice to be able to pull up the test in an IDE like IntelliJ. Then we can use the debugger and track down what the problem is. Unfortunately, this is impossible due to some unfortunate design choices.

If I open a test source file in IntelliJ and tell the IDE to run it, the first errors I get is a bunch of errors in org.apache.parquet.VersionTest, which evidently cannot find a bunch of symbols. If I hack VersionTest to no longer depend on those symbols, the next thing that happens is this error:


/home/theosib/packages/PARQUET-2069/parquet-mr/parquet-column/src/main/java/org/apache/parquet/io/MessageColumnIO.java:43:46
java: cannot find symbol
  symbol:   class IncrementallyUpdatedFilterPredicateBuilder
  location: package org.apache.parquet.filter2.recordlevel

Evidently, IncrementallyUpdatedFilterPredicateBuilder doesn't exist. Instead, it is auto-generated during the maven build process, but the code is stashed away somewhere where IntelliJ can't find it, and this is the end of the road.

If we want more people to be able to contribute to ParquetMR, I think we should work towards making it easier to debug.

Environment: In my case, IntelliJ IDEA. Reporter: Timothy Miller / @theosib-amazon

Note: This issue was originally created as PARQUET-2147. Please see the migration documentation for further details.

asfimport avatar May 13 '22 20:05 asfimport