Documentation (since 0.11.1 at least) says :
"Note: PigStorage and TextLoader correctly read compressed files as long as they are NOT CONCATENATED FILES generated in this manner: ..."
This is not true for gz, since
- I did a test - concatenated&compress some files and processed them. The same was done with the raw files (no compression). The results were identical
Jira's https://issues.apache.org/jira/i#browse/HADOOP-4012 and
https://issues.apache.org/jira/i#browse/HADOOP-6835 says the concatenation problems were fixed in Hadoop 0.22, Hadoop 0.20 respectively for both: bz2 and gz. That said Hadoop (1 and 2) are supporting concatenated archives bz2, gz already.
Pig deals with bz2 on its own(historical reasons) which is redundant to hadoop-common. Therefore this activity should be left to hadoop-common (there is no need to be handled by Pig anymore).
The documentation needs to be updated accordingly (concatenated gz, bz2 are processing correctly with hadoop-commons). Also a remark that tar.gz and tar.bz2 are not supported would be helpful since many users are using tar.gz or tar.bz2 automatically.