Unfortunately we had to remove a redirection of stdin to stderr for performance reasons. So the rule, which was always in place, is that user code should not use stdin or stdout, which is reserved for communication with hadoop. stderr should be used for messages, diagnostics etc The randomforest function with trace on writes to stdout, breaking the java data parser (which expects typedbytes)
jwills
commented
11 years ago
Unfortunately we had to remove a redirection of stdin to stderr for performance reasons. So the rule, which was always in place, is that user code should not use stdin or stdout, which is reserved for communication with hadoop. stderr should be used for messages, diagnostics etc The randomforest function with trace on writes to stdout, breaking the java data parser (which expects typedbytes)