I may be wrong here, but when I see github and apache pig, it says that there are 8 contributors, and when I see github and look at apache spark it says there are more than 1000 contributors. And if the above is true I ask myself, why not shift to SPARK by learning it?
I also started with map reduce JAVA nightmarish coding, and then HIVE and then Pig and then realised the best time spent is the time used for actually solving data problems than programming problems. I know consultants who have ended up convincing their clients that its better to write JAVA programs than use SPARK SQL and then have spent close to 2.5 years not being able to deliver anything that works where the actual project was just a single SPARK SQL.
Personally I prefer to learn and adapt and transfer existing code to a platform that gives me the maximum business benefit with least headaches.