Archives of the TeradataForum

Message Posted: Thu, 21 Aug 2008 @ 16:32:13 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

<-- Anonymously Posted: Thu, 21 Aug 2008 10:26 -->

All bytes will be used by the hash function, and column order does not matter as far as the hash.

But column order DOES matter as far as building the stats histogram - only the leftmost N bytes (could vary by release) are used to define the intervals. Very long / very non-unique values can prevent Teradata from collecting useful distribution statistics.

Also, note that the order of columns in the index (or stats histogram interval) currently depends NOT on the order in the index definition (or collect stats) but on the order in the table row.

	<Prev	Next>		<<First	<Prev	Next>	Last>>