Archives of the TeradataForum
Message Posted: Fri, 28 Jan 2005 @ 15:58:04 GMT
| Subj: || || Re: Record count stored in Data Dictionary |
| From: || || Dieter Noeth |
SRINI GORRELLA wrote:
| ||Stats just give the unique values not the row counts ... This is true if and only if you have a UPI on table that too with NOT NULL
| ||If you have a table with NUPI , stats on NUPI will give you just the unique values but not the row counts|| |
True if you look at "help stats" output, but the actual row count is stored within the binary stats. And "help index" always returns the
approixmate number of _rows_ in a table for the PI column[s].
| ||I don't understand what's the need to store row counts in dictionary|| |
Just to get quick info without a "select count(*)"?
MS SQL Server has that kind of info in sysindexes: approximate counts for number of rows and IIRC updated rows.
In Teradata a "select count(*) from table;" could be really fast, if it would sum up the number of rows per datablock, which is stored within
the cylinder index. But it's doing a stupid Full Table Scan instead reading every datablock :-(