2009/10/1 Sam Mason <sam@xxxxxxxxxxxxx>
Ah yes, that makes total sense! I knew max wouldn't be logical in such as case, but couldn't think of the alternative. Thanks!
That's clear now.
Giving my question more thought, I believe it's pointless.
bool_or and bool_and are aggregates that work over boolean data types.
Ah yes, that makes total sense! I knew max wouldn't be logical in such as case, but couldn't think of the alternative. Thanks!
I believe it's more to do with the fact that if you add a boolean column
and then subsequently an int column then you're going to struggle to
"pack" them efficiently. PG always puts columns on the "end" so that you
can add a column in constant time (i.e. no need to rewrite the table
in some common situations). Once you start doing this then packing is
awkward and a single byte becomes much easier. Whether the value is
NULL is stored elsewhere in the row.
That's clear now.
Not sure which aspects you're referring to here, sorry.
> And does its storage as a byte affect indexing or query planning?
You've answered my question. Thanks Sam.
