|
|
Archives of the TeradataForum
Message Posted: Wed, 15 Apr 2009 @ 15:22:48 GMT
Subj: | | Re: Uniquely identify indexes |
|
From: | | Anomy Anom |
<-- Anonymously Posted: Wednesday, April 15, 2009 11:04 -->
Depends on what you mean. DBC.Indices has a row for each index column, so if you are looking for unique (DatabaseName, TableName, IndexNumber)
you could simply qualify on ColumnPosition=1.
You'd probably also need to filter on IndexType. Not all "index types" are actually "indexes", e.g. 'M' for Multi-Column, non-Index stats; the
list of valid values depends on your release. See the Data Dictionary manual.
Note that in Teradata, indexes need not have names; if you omit a name when you create the index, name will be NULL, not a system-generated
value. And if an index does have a name, the scope is only for that table, i.e. you can use exactly the same index name for two entirely different
tables. So if you want to do something with this list like generate COLLECT STATS statements, you may need to do a recursive query or multiple
self-joins to string together the list of column names that define the index.
| |