Hi,

Is there a better / faster way of doing a table count?

Curently using:

select count(*)
from table

Thanks!

Depends on what you need it for. If a good estimate is sufficient (and provided statistics are current) you can use

Thanks!

I've wondered this myself. You could replace * with a constant, something like 'count(1)', or maybe just select an indexed column rather than *, i.e. count(indexed_column).

I tend to use count(*) or count(1), depending on which way the wind is blowing.

I've read articles that say that any one of the 3 concepts is the best but nothing to date has convinced me.

Maybe someone in this forum could enlighten us?

If you get the card (row count) form syscat.tables, it is only accurate as of the last time runstats command was executed on that table.

When running a count(*) opn a table with no WHERE clause (or if a WHERE clause is used, all predicates are in the index), DB2 will read an index rather than the table to get the count of rows (in most situations).