How to limit decimal values to 2 digits before applying agg function?

You can use bround:

val df = Seq(2.346, 1.549).toDF("A")
df.select(bround(df("A"), 2)).show
+------------+
|bround(A, 2)|
+------------+
|        2.35|
|        1.55|
+------------+


df.agg(sum(bround(df("A"), 2)).as("appSum")).show
+------------------+
|            appSum|
+------------------+
|3.9000000000000004|
+------------------+
                                          ^
df.agg(sum(df("A")).as("exactSum")).show
+--------+
|exactSum|
+--------+
|   3.895|
+--------+

The above solution does work for spark 2.0 version however for folks like me who are still using 1.5.*+ versions below is something that will work.(I used round function as suggested by @Psidom):

val df = Seq(2.346, 1.549).toDF("A")
df.select(bround(df("A"), 2)).show
+------------+
|bround(A, 2)|
+------------+
|        2.35|
|        1.55|
+------------+

val total=df.agg(sum(round(df.col(colName),2)).cast("double")).first.getDouble(0)
total: Double = 3.90