I figured out what appears to happen with cascading delete using a
seqscan. In this case, the foreign keys in the child table are not
equally distributed. A few parent values occur often. Most parent
values do not occur at all. So the planner, faced with an unknown
generic key, takes the safe route.
What I've done is remove the FK (maybe it would be better to leave it
albeit disabled for documentation) and written my own AFTER DELETE
trigger that uses EXECUTE to delay planning until the actual value is
known. This appears to work correctly.
--
Sincerely,
Andrew Lazarus mailto:andrew@xxxxxxxxxxxx
BEGIN:VCARD
VERSION:2.1
N:Lazarus;Andrew;;;Ph.D.
FN:Andrew Lazarus, Ph.D.
EMAIL;PREF;INTERNET:andrew@xxxxxxxxxxxx
TITLE:Director of R&D
ADR;WORK:;800-366-0688;3028 Fillmore Street;San Francisco;CA;94123;USA
LABEL;WORK;ENCODING=QUOTED-PRINTABLE:800-366-0688=0D=0A3028 Fillmore S=
treet=0D=0ASan Francisco=0D=0ACA=0D=0A94123=0D=0AUSA
X-GENDER:Male
REV:18991230T080000Z
END:VCARD
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
