Archives of the TeradataForum

Message Posted: Wed, 05 Nov 2003 @ 18:48:42 GMT

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

While I would agree that it is understandable, given the indeterminate nature of NULL that the unique values would be set as it is....checking the number of rows, in addition to the unique values before deciding which table to redistribute would seem to be a reasonable thing to do. I would agree that perhaps the PI choice might be better made but this would seem to be hole that the optimizer/others could fall into under less graphic conditions than those illustrated in this particular example. For example, what if you had a composite PI in which one of the value was NULL only 50% of the time? That still might be enough to skew the decision making for how the query should be executed in a potentially bad direction, where factoring the row counts information into the decision making process ,in addition to the unique values data, might point to a better choice. It would definitely be interesting to hear someone knowledgeable from NCR weigh in on this topic!.....Neil

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