question about the count method in spark dataset?

Question

I was reading the book 'spark definitive guide' It has an example like below.

val myRange = spark.range(1000).toDF("number")
val divisBy2 = myRange.where("number % 2 = 0")
divisBy2.count()

Below is the description for the three lines of code.

we started a Spark job that runs our filter transformation (a narrow
transformation), then an aggregation (a wide transformation) that performs the counts on a per
partition basis, and then a collect, which brings our result to a native object in the respective
language

I know the count is an action not a transformation, since it return an actual value and I can not call 'explain' on the return value of count.

But I was wondering why the count will cause the wide transformation, how can I know the execution plan of this count in tis case since I can not invoke the 'explain' after count

Thanks.

updated:

This image is the spark ui screenshot, I take it from databricks notebook, I said there is a shuffle write and read operation, does it mean that there is a wide transformation?

Answer 1

Here is the execution plan:

== Physical Plan ==
*(2) HashAggregate(keys=[], functions=[count(1)], output=[count#7L])
+- Exchange SinglePartition
   +- *(1) HashAggregate(keys=[], functions=[partial_count(1)], output=[count#10L])
      +- *(1) Project
         +- *(1) Filter ((id#0L % 2) = 0)
            +- *(1) Range (0, 1000, step=1, splits=8)

What we can see here:

1. Counting made inside each partition
2. All partitions merged into the single one
3. Final counting made