Hello Jeremy,
Comments, notes?
One oddity about pg_relation_size and pg_table_size is that they can be
easily blocked by user activity. In fact it happens to us often in
reporting environments and we have instead written different versions of
them that avoid the lock contention and still give "close enough" results.
This blocking could result in quite unexpected behavior, that someone uses
your proposed command and it never returns. Has that been considered as a
reality at least to be documented?
ISTM that it does not change anything wrt the current behavior because of
the prudent lazy approach: the sorting is *only* performed when the size
is already available in one of the printed column.
Maybe the more general question could be "is there a caveat somewhere that
when doing \d.+ a user may have issues with locks because of the size
computations?".
--
Fabien.
