"compile only" PigServer/PigUnit requires params substitution. To be more specific, to compile a java app "javac" don't need to know the internals of the app.
"registering dependencies" Same case with dependencies as well. By copy-pasting & hard-coding libraries into the script introduces new set of problems during version upgrade of any library. In java world, no one wants to hardcode jar names inside a java file. Hope, this clarifies my requirement. On Sun, Apr 27, 2014 at 4:29 AM, Jay Vyas <jayunit...@gmail.com> wrote: > Ah, I guess from the "pig maven integration" subject heading i assumed you > wanted tight maven integration of your pig scripts. In any case, here is > what i was suggesting. > > 1) How can we just *compile (compile only)* pig scripts? > > See > https://github.com/apache/bigtop/blob/master/bigtop-bigpetstore/src/main/java/org/apache/bigtop/bigpetstore/etl/PigCSVCleaner.java. > That class will run a pig script for by creating an instance of the > "PigServer" object. This allows you to craft your own pig tests in java > that can validate your data flow and pig script in a java app. > Alternatively, use PigUnit to test your pig script. Again, although these > do more than compilation, the effect is essentially the same: They > validate your script without running in a real cluster. > > 2) Also, what is the clean way to handle library dependencies of a Pig > script? > > Sorry, about this, I thought by the subject "pig maven integration" you > were referring to the various dependencies you need in a maven project to > run pig code with real API calls. For that, you can paste the pom > dependencies . > > If you mean external libraries, just use the pig "register" command in your > pig script.
