Re: Why it is using/not using index scan?

Dne 31.3.2011 19:26, Laszlo Nagy napsal(a):
> For this query:
>
> select pp.id,pp.product_id,pp.selling_site_id,pp.asin
> from product_price pp
> where
> (pp.asin is not null and pp.asin<>'')
> and (pp.upload_status_id<>1)
> and pp.selling_site_id in (8,7,35,6,9)
> and (pp.last_od < 'now'::timestamp - '1 week'::interval )
> limit 5000
>
> Query plan is:
>
> "Limit (cost=9182.41..77384.80 rows=3290 width=35)"
> " -> Bitmap Heap Scan on product_price pp (cost=9182.41..77384.80
> rows=3290 width=35)"
> " Recheck Cond: ((last_od < '2011-03-24
> 13:05:09.540025'::timestamp without time zone) AND (selling_site_id =
> ANY ('{8,7,35,6,9}'::bigint[])))"
> " Filter: ((asin IS NOT NULL) AND (asin <> ''::text) AND
> (upload_status_id <> 1))"
> " -> Bitmap Index Scan on idx_product_price_last_od_ss
> (cost=0.00..9181.59 rows=24666 width=0)"
> " Index Cond: ((last_od < '2011-03-24
> 13:05:09.540025'::timestamp without time zone) AND (selling_site_id =
> ANY ('{8,7,35,6,9}'::bigint[])))"
>
> For this query:
>
> select pp.id,pp.product_id,pp.selling_site_id,pp.asin
> from product_price pp
> where
> (pp.asin is not null and pp.asin<>'')
> and (pp.upload_status_id<>1)
> and pp.selling_site_id in (8,7,35,6,9)
> and (pp.last_od + '1 week'::interval < 'now'::timestamp )
> limit 5000
>
> Query plan is:
>
> "Limit (cost=0.00..13890.67 rows=5000 width=35)"
> " -> Seq Scan on product_price pp (cost=0.00..485889.97 rows=174898
> width=35)"
> " Filter: ((asin IS NOT NULL) AND (asin <> ''::text) AND
> (upload_status_id <> 1) AND ((last_od + '7 days'::interval) <
> '2011-03-31 13:06:17.460013'::timestamp without time zone) AND
> (selling_site_id = ANY ('{8,7,35,6,9}'::bigint[])))"
>
>
> The only difference is this: instead of (pp.last_od < 'now'::timestamp -
> '1 week'::interval ) I have used (pp.last_od + '1 week'::interval <
> 'now'::timestamp )

That's the only difference as you see it - the planner actually found
out the former query is expected to return 3290 rows while the latter
one is expected to return 174898 rows. That's the reason why the second
query is using seqscan instead of index scan - for a lot of rows, the
index scan tends to be very inefficient.

Next time post EXPLAIN ANALYZE output, as it provides data from the
actual run, so we can see if there are any issues with those estimates.

Anyway, you may try to disable sequential scans (just run 'set
enable_seqacan=off' before running the query) and you'll see if index
scan really would be better.

> First query with index scan opens in 440msec. The second query with seq
> scan opens in about 22 seconds. So the first one is about 50x faster.

Every database/planner has some weaknesses - it may be the case that an
index scan would be faster but postgresql is not able to use it in this
case for some reason.

> My concern is that we are working on a huge set of applications that use
> thousands of different queries on a database. There are programs that we
> wrote years ago. The database structure continuously changing. We are
> adding new indexes and columns, and of course we are upgrading
> PostgreSQL when a new stable version comes out. There are cases when a
> change in a table affects 500+ queries in 50+ programs. I really did not
> think that I have to be THAT CAREFUL with writing conditions in SQL. Do
> I really have to manually analyze all those queries and "correct"
> conditions like this?

You have to be that careful, and it's not specific to PostgreSQL. I'm
working with other databases and the same holds for them - SQL looks so
simple that a chimp might learn it, but only the best chimps may produce
good queries.

> If so, then at least I would like to know if there is a documentation or
> wiki page where I can learn about "how not to write conditions". I just
> figured out that I need to put constant expressions on one side of any
> comparison, if possible. But probably there are other rules I wouldn't
> think of.

I'm not aware of such official document. I've planned to write an
article "10 ways to ruin your query" but I somehow forgot about it.
Anyway it's mostly common sense, i.e. once you know how indexes work
you'll immediately see if a condition may benefit from them or not.

Tomas