 |
 |
Archives of the TeradataForum
Message Posted: Thu, 06 May 2004 @ 08:09:43 GMT
| Subj: | | Re: Collect Stats on Indexes versus Multi-Column |
| |
| From: | | Victor Sokovin |
On Wed, 5 May 2004 21:56:34 +0200, Dieter Noeth wrote:
| | There's a difference:
You can drop and recreate that index without losing statistics (similar to a single column index), if you collect on multi-column, because those
statistics are stored as a pseudo-index (with indextype 'M') in dbc.indexes. | |
I believe different algorithms for calculating the stats might be used as well. It is especially clear from the documentation if you take
the part on sample stats. This might be the difference even in the single column case. I would keep experimenting with both syntax forms.
And the two statements you mentioned are surely not equivalent if TableA != TableB but I assume you were not testing our observation
capacity.
Regards,
Victor
| |
