I'm designing a system where I'll be making frequent updates to rows, but
some columns will change far less frequently than others. All columns will
be read with equal frequency, though probably by means of a materialized
view. Updates will happen via a stored proc.
Which makes the most sense?
1. Blindly overwrite the value for all columns, even columns that haven't
2. Compare the current value and update each column at a time, leaving
unchanged columns alone.
3. Break out my tables more, so that infrequently updated columns are in
I'm thinking that 1 and 2 are similar and less complex than 3, but might
result in more dead tuples. 2 and 3 will have to look up values before
updates can happen. I don't know if 2 has any advantage over 1. 1 is
certainly the simpliest.
pgsql-general by date
|Next:||From: Sally Ruggero||Date: 2004-10-29 19:42:10|
|Subject: Recommended Procedure for Archiving Table Data|
|Previous:||From: Eric E||Date: 2004-10-29 19:20:52|
|Subject: SELECT a value from various tables depending on a column value|