Nicht ganz frische PostgreSQL-Versionen (mindestens bis 9.6 inklusive)
optimieren »NOT IN«-mit-Subquery nicht automatisch und müssen daher
für jede Zeile des äußeren Selects einen linearen Scan auf die
Subquery-Tabelle machen.
Deutlich effektiver ist das in diesem Fall äquivalente »NOT
EXISTS«-mit-Subquery.
Beispiel: Namen aller Kunden, für die noch keine Rechnung geschrieben
wurde. Falsch mit »NOT IN«:
SELECT name
FROM customer
WHERE id NOT IN (
SELECT customer_id
FROM ar
);
Besser und semantisch äquivalent:
SELECT name
FROM customer
WHERE NOT EXISTS (
SELECT customer_id
FROM ar
WHERE customer_id = customer.id
);
Geschwindigkeitssteigerung ist auch ein Euphemismus. Vor der Änderung
war der Selftest bei der LINET-Produktivdatenbank nicht in der Lage,
seine Tests innerhalb von drei Tagen auszuführen. Nach der Änderung
dauert der Selftest weniger als eine Minute.
Neure PostgreSQL-Versionen (z.B. v11) erkennen dieses Pattern
automatisch.
my $query = qq|
SELECT purpose from bank_transactions
WHERE cleared is true
my $query = qq|
SELECT purpose from bank_transactions
WHERE cleared is true
- AND id not in (SELECT bank_transaction_id from reconciliation_links)
+ AND NOT EXISTS (SELECT bank_transaction_id from reconciliation_links WHERE bank_transaction_id = bank_transactions.id)
AND transdate >= ? AND transdate <= ?|;
my $bt_cleared_no_link = selectall_hashref_query($::form, $self->dbh, $query, $self->fromdate, $self->todate);
AND transdate >= ? AND transdate <= ?|;
my $bt_cleared_no_link = selectall_hashref_query($::form, $self->dbh, $query, $self->fromdate, $self->todate);
my $query = qq|
SELECT purpose from bank_transactions
WHERE invoice_amount <> 0
my $query = qq|
SELECT purpose from bank_transactions
WHERE invoice_amount <> 0
- AND id not in (SELECT bank_transaction_id from bank_transaction_acc_trans)
+ AND NOT EXISTS (SELECT bank_transaction_id FROM bank_transaction_acc_trans WHERE bank_transaction_id = bank_transactions.id)
AND itime > (SELECT min(itime) from bank_transaction_acc_trans)
AND transdate >= ? AND transdate <= ?|;
AND itime > (SELECT min(itime) from bank_transaction_acc_trans)
AND transdate >= ? AND transdate <= ?|;
SELECT purpose from bank_transactions
WHERE id in
(SELECT bank_transaction_id from bank_transaction_acc_trans
SELECT purpose from bank_transactions
WHERE id in
(SELECT bank_transaction_id from bank_transaction_acc_trans
- where acc_trans_id NOT IN (select acc_trans_id from acc_trans)
+ WHERE NOT EXISTS (SELECT acc_trans.acc_trans_id FROM acc_trans WHERE acc_trans.acc_trans_id = bank_transaction_acc_trans.acc_trans_id)
AND transdate >= ? AND transdate <= ?)|;
my $bt_assigned_no_acc_trans = selectall_hashref_query($::form, $self->dbh, $query, $self->fromdate, $self->todate);
AND transdate >= ? AND transdate <= ?)|;
my $bt_assigned_no_acc_trans = selectall_hashref_query($::form, $self->dbh, $query, $self->fromdate, $self->todate);