Github user paul-rogers commented on a diff in the pull request:
https://github.com/apache/drill/pull/932#discussion_r142017748
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/physical/impl/spill/RecordBatchSizer.java
---
@@ -74,53 +74,52 @@
public final int estSize;
/**
- * Number of times the value here (possibly repeated) appears in
- * the record batch.
+ * Number of occurrences of the value in the batch. This is trivial
+ * for top-level scalars: it is the record count. For a top-level
+ * repeated vector, this is the number of arrays, also the record
+ * count. For a value nested inside a repeated map, it is the
+ * total number of values across all maps, and may be less than,
+ * greater than (but unlikely) same as the row count.
*/
public final int valueCount;
/**
- * The number of elements in the value vector. Consider two cases.
- * A required or nullable vector has one element per row, so the
- * <tt>entryCount</tt> is the same as the <tt>valueCount</tt> (which,
- * in turn, is the same as the row count.) But, if this vector is an
- * array, then the <tt>valueCount</tt> is the number of columns, while
- * <tt>entryCount</tt> is the total number of elements in all the
arrays
- * that make up the columns, so <tt>entryCount</tt> will be different
than
- * the <tt>valueCount</tt> (normally larger, but possibly smaller if
most
- * arrays are empty.
- * <p>
- * Finally, the column may be part of another list. In this case, the
above
- * logic still applies, but the <tt>valueCount</tt> is the number of
entries
- * in the outer array, not the row count.
+ * Total number of elements for a repeated type, or 1 if this is
+ * a non-repeated type. That is, a batch of 100 rows may have an
+ * array with 10 elements per row. In this case, the element count
+ * is 1000.
*/
- public int entryCount;
+ public final int elementCount;
--- End diff --
Good point. However, a single batch of greater than 2 GB is far more than
the sort can handle, so we'd not even get this far if the batch was this large.
Still, the point is valid, so a new commit changes batch size variables
from int to long.
---
