Spark's log4j wrapper is already used in a few places for e.g. logging warnings, and we should fold our logging into that infrastructure, where we currently have a tiny homegrown progress helper.
The former supports a lot more customizability (setting levels per package/class-name, lazily evaluating expressions iff they are being logged at a level that is displayable, etc.).
Spark's log4j wrapper is already used in a few places for e.g. logging warnings, and we should fold our logging into that infrastructure, where we currently have a tiny homegrown
progresshelper.The former supports a lot more customizability (setting levels per package/class-name, lazily evaluating expressions iff they are being logged at a level that is displayable, etc.).