Update comment to better reflect what we're doing

Author: ash211

core/src/main/scala/org/apache/spark/rdd/RDD.scala

@@ -1064,9 +1064,8 @@ abstract class RDD[T: ClassTag](
       // greater than totalParts because we actually cap it at totalParts in runJob.
       var numPartsToTry = 1
       if (partsScanned > 0) {
-        // If we didn't find any rows after the first iteration, just try all partitions next.
-        // Otherwise, interpolate the number of partitions we need to try, but overestimate it
-        // by 50%.
+        // If we didn't find any rows after the previous iteration, double and retry.  Otherwise,
+        // interpolate the number of partitions we need to try, but overestimate it by 50%.
         if (buf.size == 0) {
           numPartsToTry = partsScanned * 2
         } else {

python/pyspark/rdd.py

@@ -1031,9 +1031,9 @@ def take(self, num):
             # we actually cap it at totalParts in runJob.
             numPartsToTry = 1
             if partsScanned > 0:
-                # If we didn't find any rows after the first iteration, just
-                # try all partitions next. Otherwise, interpolate the number
-                # of partitions we need to try, but overestimate it by 50%.
+                # If we didn't find any rows after the previous iteration,
+                # double and retry.  Otherwise, interpolate the number of
+                # partitions we need to try, but overestimate it by 50%.
                 if len(items) == 0:
                     numPartsToTry = partsScanned * 2
                 else: