We should accept this patch for no reason other than it boosts our coverall number significantly!
Coverage increased (+10.67%) when pulling 15dc1a7 on ericzundel:zundel/ivy-imports-map-jars-once into 3697878 on pantsbuild:master.
I was wondering why. I guess there is a lot of ivy code we don't have unit tests for, but 10% sounds like an outsized increase?
In IvyImports, make a single call to ivy to map all jar files.
Review Request #1364 — Created Nov. 19, 2014 and discarded
|dturner-tw, ity, jsirois, patricklaw|
If you have many targets with jars to map in, this is a significant performance increase.
Added a new unit test.
Ran this in our internal repo where we have a lot of uses of java_protobuf_library() using imports=.
CI running at: https://travis-ci.org/pantsbuild/pants/builds/41531811
Incorporated feedback from David Turner.
Revision 2 (+89 -7)