Re: fixed-length row

David Garamond <lists(at)zara(dot)6(dot)isreserved(dot)com> writes:
> The MySQL manual recommends that we create a "fixed-length row" if
> possible, for speed (especially scanning speed).

> Is there a similar recommendation in PostgreSQL?

No. There are some marginal optimizations that take place if your
columns are fixed-width and not null, but I wouldn't suggest contorting
your database design to enable them to occur.

In particular, people who have taken this bait generally think that
it's a good idea to substitute char(n) for varchar(n). That is almost
inevitably a pessimization, because the extra I/O time for all those
padding blanks will surely swamp the few CPU cycles saved by using
precalculated field offsets. (Not to mention that char(n) is not really
fixed-width anyway, in Postgres or any other implementation that
supports variable-length character encodings.)

I'd be willing to speculate that the MySQL manual's advice is bad
even for MySQL, but I haven't benchmarked the case there.

regards, tom lane