gistfile1.txt

update, I have a test case here: https://gist.github.com/Chandler/7766963

I'm running it with multiple files as the input and cascading.hadoop.hfs.combine.files=true

When I run it against stock elephant bird it fails with the original error
"com.twitter.elephantbird.mapred.input.DeprecatedLzoTextInputFormat cannot be cast to org.apache.hadoop.mapred.FileInputFormat"

https://github.com/kevinweil/elephant-bird/pull/359
When I run it with Dmitriy's elephant bird patch it fails with "DeprecatedInputFormatWrapper can not support RecordReaders that don't return same key & value objects. current reader class : class com.twitter.elephantbird.mapreduce.input.LzoLineRecordReader"

I addressed the RecordReaders issue here: https://github.com/kevinweil/elephant-bird/pull/360

When I run the test against both mine and Dmitriy's patches the job succeeds!

But something is up with the combine files setting.

input is two identical lzo text files that contain 3 newline separated words
hello
hello
tiger

output with cascading.hadoop.hfs.combine.files=false:
tiger 2
hello 4

output with cascading.hadoop.hfs.combine.files=true
1 hello 3
tiger 2