Secara periodik, tugas nggoleki data sing gegandhengan nggunakake sekumpulan tombol muncul. nganti entuk jumlah rekaman sing dibutuhake.
Conto paling "urip nyata" kanggo nampilake 20 masalah paling tuwa, kadhaptar ing dhaptar karyawan (contone, ing siji divisi). Kanggo macem-macem "dashboard" manajemen kanthi ringkesan ringkes babagan wilayah kerja, topik sing padha dibutuhake asring.
Ing artikel iki, kita bakal ndeleng implementasine ing PostgreSQL saka solusi "naif" kanggo masalah kasebut, algoritma "pinter" lan rumit banget. "loop" ing SQL kanthi kondisi metu saka data sing ditemokake, sing bisa migunani kanggo pangembangan umum lan digunakake ing kasus liyane sing padha.
Ayo dadi njupuk set data test saka
CREATE INDEX ON task(owner_id, task_date, id);
-- Π° ΡΡΠ°ΡΡΠΉ - ΡΠ΄Π°Π»ΠΈΠΌ
DROP INDEX task_owner_id_task_date_idx;
Kaya sing dirungokake, kaya sing ditulis
Pisanan, ayo gawe sketsa versi panjalukan sing paling gampang, ngliwati ID para pemain.
SELECT
*
FROM
task
WHERE
owner_id = ANY('{1,2,4,8,16,32,64,128,256,512}'::integer[])
ORDER BY
task_date, id
LIMIT 20;
A little sedih - kita mung dhawuh 20 cathetan, nanging Index Scan bali menyang kita 960 larik, sing banjur uga kudu diurutake ... Coba maca kurang.
unnest + ARRAY
Wawasan pisanan sing bakal mbantu kita yaiku yen kita butuh mung 20 diurutake cathetan, banjur mung maca ora luwih saka 20 diurutake ing urutan padha kanggo saben kuncine. apik, indeks cocok (id_pemilik, tanggal_tugas, id) kita duwe.
Ayo nggunakake mekanisme sing padha kanggo ngekstrak lan "nyebar menyang kolom" cathetan tabel integral, ing ARRAY()
:
WITH T AS (
SELECT
unnest(ARRAY(
SELECT
t
FROM
task t
WHERE
owner_id = unnest
ORDER BY
task_date, id
LIMIT 20 -- ΠΎΠ³ΡΠ°Π½ΠΈΡΠΈΠ²Π°Π΅ΠΌ ΡΡΡ...
)) r
FROM
unnest('{1,2,4,8,16,32,64,128,256,512}'::integer[])
)
SELECT
(r).*
FROM
T
ORDER BY
(r).task_date, (r).id
LIMIT 20; -- ... ΠΈ ΡΡΡ - ΡΠΎΠΆΠ΅
Oh, wis luwih apik! 40% luwih cepet lan 4.5 kaping kurang data Aku kudu maca.
Materialisasi rekaman tabel liwat CTEAyo kula tarik manungsa waΓ© kanggo kasunyatan sing ing sawetara kasus Usaha kanggo langsung nggarap lapangan rekaman sawise nggoleki ing subquery, tanpa "bungkus" ing CTE, bisa nyebabake "multiply" InitPlan sebanding karo jumlah kolom sing padha:
SELECT
((
SELECT
t
FROM
task t
WHERE
owner_id = 1
ORDER BY
task_date, id
LIMIT 1
).*);
Result (cost=4.77..4.78 rows=1 width=16) (actual time=0.063..0.063 rows=1 loops=1)
Buffers: shared hit=16
InitPlan 1 (returns $0)
-> Limit (cost=0.42..1.19 rows=1 width=48) (actual time=0.031..0.032 rows=1 loops=1)
Buffers: shared hit=4
-> Index Scan using task_owner_id_task_date_id_idx on task t (cost=0.42..387.57 rows=500 width=48) (actual time=0.030..0.030 rows=1 loops=1)
Index Cond: (owner_id = 1)
Buffers: shared hit=4
InitPlan 2 (returns $1)
-> Limit (cost=0.42..1.19 rows=1 width=48) (actual time=0.008..0.009 rows=1 loops=1)
Buffers: shared hit=4
-> Index Scan using task_owner_id_task_date_id_idx on task t_1 (cost=0.42..387.57 rows=500 width=48) (actual time=0.008..0.008 rows=1 loops=1)
Index Cond: (owner_id = 1)
Buffers: shared hit=4
InitPlan 3 (returns $2)
-> Limit (cost=0.42..1.19 rows=1 width=48) (actual time=0.008..0.008 rows=1 loops=1)
Buffers: shared hit=4
-> Index Scan using task_owner_id_task_date_id_idx on task t_2 (cost=0.42..387.57 rows=500 width=48) (actual time=0.008..0.008 rows=1 loops=1)
Index Cond: (owner_id = 1)
Buffers: shared hit=4"
InitPlan 4 (returns $3)
-> Limit (cost=0.42..1.19 rows=1 width=48) (actual time=0.009..0.009 rows=1 loops=1)
Buffers: shared hit=4
-> Index Scan using task_owner_id_task_date_id_idx on task t_3 (cost=0.42..387.57 rows=500 width=48) (actual time=0.009..0.009 rows=1 loops=1)
Index Cond: (owner_id = 1)
Buffers: shared hit=4
Rekaman sing padha "digoleki" kaping 4 ... Nganti PostgreSQL 11, prilaku iki kedadeyan kanthi rutin, lan solusi kasebut yaiku "bungkus" ing CTE, sing minangka watesan mutlak kanggo pangoptimal ing versi kasebut.
Akumulator rekursif
Ing versi sadurunge, total kita maca 200 larik marga saka dibutuhake 20. Ora 960, nanging malah kurang - iku bisa?
Ayo padha nyoba nggunakake kawruh sing kita butuhake gunggunge 20 cathetan. Yaiku, kita bakal ngulang maca data mung nganti tekan jumlah sing dibutuhake.
Langkah 1: Daftar Miwiti
Temenan, dhaptar "target" saka 20 cathetan kudu diwiwiti karo cathetan "pisanan" kanggo salah sawijining kunci owner_id. Mulane, pisanan kita bakal nemokake kuwi "pisanan banget" kanggo saben tombol lan ditambahake menyang dhaptar, ngurutake ing urutan sing dikarepake - (task_date, id).
Langkah 2: Temokake entri "sabanjure".
Saiki yen kita njupuk entri pisanan saka dhaftar kita lan miwiti "langkah" luwih ing sadawane indeks ngreksa tombol owner_id, banjur kabeh cathetan ketemu persis sing sabanjurΓ© ing pilihan asil. Mesthi, mung nganti kita nyabrang tombol bokong entri kapindho ing dhaftar.
Yen ternyata kita "nyabrang" rekaman kapindho, banjur entri pungkasan sing diwaca kudu ditambahake menyang dhaptar tinimbang sing pisanan (karo owner_id padha), sawise iku maneh ngurutake dhaftar maneh.
Yaiku, kita mesthi ngerteni manawa dhaptar ora duwe luwih saka siji entri kanggo saben tombol (yen entri entek lan ora "nyebrang", mula entri pisanan saka dhaptar bakal ilang lan ora ana sing ditambahake. ), lan padha tansah diurutake ing urutan munggah tombol aplikasi (task_date, id).
Langkah 3: Filter lan "ngembangake" cathetan
Ing sawetara larik saka pilihan rekursif kita, sawetara cathetan rv
sing diduplikasi - pisanan kita temokake kayata "nyebrang tapel wates entri 2nd dhaftar", lan banjur sulih minangka 1st saka dhaftar. Dadi kedadeyan pisanan kudu disaring.
Pitakonan pungkasan sing nggegirisi
WITH RECURSIVE T AS (
-- #1 : Π·Π°Π½ΠΎΡΠΈΠΌ Π² ΡΠΏΠΈΡΠΎΠΊ "ΠΏΠ΅ΡΠ²ΡΠ΅" Π·Π°ΠΏΠΈΡΠΈ ΠΏΠΎ ΠΊΠ°ΠΆΠ΄ΠΎΠΌΡ ΠΈΠ· ΠΊΠ»ΡΡΠ΅ΠΉ Π½Π°Π±ΠΎΡΠ°
WITH wrap AS ( -- "ΠΌΠ°ΡΠ΅ΡΠΈΠ°Π»ΠΈΠ·ΡΠ΅ΠΌ" record'Ρ, ΡΡΠΎΠ±Ρ ΠΎΠ±ΡΠ°ΡΠ΅Π½ΠΈΠ΅ ΠΊ ΠΏΠΎΠ»ΡΠΌ Π½Π΅ Π²ΡΠ·ΡΠ²Π°Π»ΠΎ ΡΠΌΠ½ΠΎΠΆΠ΅Π½ΠΈΡ InitPlan/SubPlan
WITH T AS (
SELECT
(
SELECT
r
FROM
task r
WHERE
owner_id = unnest
ORDER BY
task_date, id
LIMIT 1
) r
FROM
unnest('{1,2,4,8,16,32,64,128,256,512}'::integer[])
)
SELECT
array_agg(r ORDER BY (r).task_date, (r).id) list -- ΡΠΎΡΡΠΈΡΡΠ΅ΠΌ ΡΠΏΠΈΡΠΎΠΊ Π² Π½ΡΠΆΠ½ΠΎΠΌ ΠΏΠΎΡΡΠ΄ΠΊΠ΅
FROM
T
)
SELECT
list
, list[1] rv
, FALSE not_cross
, 0 size
FROM
wrap
UNION ALL
-- #2 : Π²ΡΡΠΈΡΡΠ²Π°Π΅ΠΌ Π·Π°ΠΏΠΈΡΠΈ 1-Π³ΠΎ ΠΏΠΎ ΠΏΠΎΡΡΠ΄ΠΊΡ ΠΊΠ»ΡΡΠ°, ΠΏΠΎΠΊΠ° Π½Π΅ ΠΏΠ΅ΡΠ΅ΡΠ°Π³Π½Π΅ΠΌ ΡΠ΅ΡΠ΅Π· Π·Π°ΠΏΠΈΡΡ 2-Π³ΠΎ
SELECT
CASE
-- Π΅ΡΠ»ΠΈ Π½ΠΈΡΠ΅Π³ΠΎ Π½Π΅ Π½Π°ΠΉΠ΄Π΅Π½ΠΎ Π΄Π»Ρ ΠΊΠ»ΡΡΠ° 1-ΠΉ Π·Π°ΠΏΠΈΡΠΈ
WHEN X._r IS NOT DISTINCT FROM NULL THEN
T.list[2:] -- ΡΠ±ΠΈΡΠ°Π΅ΠΌ Π΅Π΅ ΠΈΠ· ΡΠΏΠΈΡΠΊΠ°
-- Π΅ΡΠ»ΠΈ ΠΌΡ ΠΠ ΠΏΠ΅ΡΠ΅ΡΠ΅ΠΊΠ»ΠΈ ΠΏΡΠΈΠΊΠ»Π°Π΄Π½ΠΎΠΉ ΠΊΠ»ΡΡ 2-ΠΉ Π·Π°ΠΏΠΈΡΠΈ
WHEN X.not_cross THEN
T.list -- ΠΏΡΠΎΡΡΠΎ ΠΏΡΠΎΡΡΠ³ΠΈΠ²Π°Π΅ΠΌ ΡΠΎΡ ΠΆΠ΅ ΡΠΏΠΈΡΠΎΠΊ Π±Π΅Π· ΠΌΠΎΠ΄ΠΈΡΠΈΠΊΠ°ΡΠΈΠΉ
-- Π΅ΡΠ»ΠΈ Π² ΡΠΏΠΈΡΠΊΠ΅ ΡΠΆΠ΅ Π½Π΅Ρ 2-ΠΉ Π·Π°ΠΏΠΈΡΠΈ
WHEN T.list[2] IS NULL THEN
-- ΠΏΡΠΎΡΡΠΎ Π²ΠΎΠ·Π²ΡΠ°ΡΠ°Π΅ΠΌ ΠΏΡΡΡΠΎΠΉ ΡΠΏΠΈΡΠΎΠΊ
'{}'
-- ΠΏΠ΅ΡΠ΅ΡΠΎΡΡΠΈΡΠΎΠ²ΡΠ²Π°Π΅ΠΌ ΡΠ»ΠΎΠ²Π°ΡΡ, ΡΠ±ΠΈΡΠ°Ρ 1-Ρ Π·Π°ΠΏΠΈΡΡ ΠΈ Π΄ΠΎΠ±Π°Π²Π»ΡΡ ΠΏΠΎΡΠ»Π΅Π΄Π½ΡΡ ΠΈΠ· Π½Π°ΠΉΠ΄Π΅Π½Π½ΡΡ
ELSE (
SELECT
coalesce(T.list[2] || array_agg(r ORDER BY (r).task_date, (r).id), '{}')
FROM
unnest(T.list[3:] || X._r) r
)
END
, X._r
, X.not_cross
, T.size + X.not_cross::integer
FROM
T
, LATERAL(
WITH wrap AS ( -- "ΠΌΠ°ΡΠ΅ΡΠΈΠ°Π»ΠΈΠ·ΡΠ΅ΠΌ" record
SELECT
CASE
-- Π΅ΡΠ»ΠΈ Π²ΡΠ΅-ΡΠ°ΠΊΠΈ "ΠΏΠ΅ΡΠ΅ΡΠ°Π³Π½ΡΠ»ΠΈ" ΡΠ΅ΡΠ΅Π· 2-Ρ Π·Π°ΠΏΠΈΡΡ
WHEN NOT T.not_cross
-- ΡΠΎ Π½ΡΠΆΠ½Π°Ρ Π·Π°ΠΏΠΈΡΡ - ΠΏΠ΅ΡΠ²Π°Ρ ΠΈΠ· ΡΠΏΠΏΠΈΡΠΊΠ°
THEN T.list[1]
ELSE ( -- Π΅ΡΠ»ΠΈ Π½Π΅ ΠΏΠ΅ΡΠ΅ΡΠ΅ΠΊΠ»ΠΈ, ΡΠΎ ΠΊΠ»ΡΡ ΠΎΡΡΠ°Π»ΡΡ ΠΊΠ°ΠΊ Π² ΠΏΡΠ΅Π΄ΡΠ΄ΡΡΠ΅ΠΉ Π·Π°ΠΏΠΈΡΠΈ - ΠΎΡΡΠ°Π»ΠΊΠΈΠ²Π°Π΅ΠΌΡΡ ΠΎΡ Π½Π΅Π΅
SELECT
_r
FROM
task _r
WHERE
owner_id = (rv).owner_id AND
(task_date, id) > ((rv).task_date, (rv).id)
ORDER BY
task_date, id
LIMIT 1
)
END _r
)
SELECT
_r
, CASE
-- Π΅ΡΠ»ΠΈ 2-ΠΉ Π·Π°ΠΏΠΈΡΠΈ ΡΠΆΠ΅ Π½Π΅Ρ Π² ΡΠΏΠΈΡΠΊΠ΅, Π½ΠΎ ΠΌΡ Ρ
ΠΎΡΡ ΡΡΠΎ-ΡΠΎ Π½Π°ΡΠ»ΠΈ
WHEN list[2] IS NULL AND _r IS DISTINCT FROM NULL THEN
TRUE
ELSE -- Π½ΠΈΡΠ΅Π³ΠΎ Π½Π΅ Π½Π°ΡΠ»ΠΈ ΠΈΠ»ΠΈ "ΠΏΠ΅ΡΠ΅ΡΠ°Π³Π½ΡΠ»ΠΈ"
coalesce(((_r).task_date, (_r).id) < ((list[2]).task_date, (list[2]).id), FALSE)
END not_cross
FROM
wrap
) X
WHERE
T.size < 20 AND -- ΠΎΠ³ΡΠ°Π½ΠΈΡΠΈΠ²Π°Π΅ΠΌ ΡΡΡ ΠΊΠΎΠ»ΠΈΡΠ΅ΡΡΠ²ΠΎ
T.list IS DISTINCT FROM '{}' -- ΠΈΠ»ΠΈ ΠΏΠΎΠΊΠ° ΡΠΏΠΈΡΠΎΠΊ Π½Π΅ ΠΊΠΎΠ½ΡΠΈΠ»ΡΡ
)
-- #3 : "ΡΠ°Π·Π²ΠΎΡΠ°ΡΠΈΠ²Π°Π΅ΠΌ" Π·Π°ΠΏΠΈΡΠΈ - ΠΏΠΎΡΡΠ΄ΠΎΠΊ Π³Π°ΡΠ°Π½ΡΠΈΡΠΎΠ²Π°Π½ ΠΏΠΎ ΠΏΠΎΡΡΡΠΎΠ΅Π½ΠΈΡ
SELECT
(rv).*
FROM
T
WHERE
not_cross; -- Π±Π΅ΡΠ΅ΠΌ ΡΠΎΠ»ΡΠΊΠΎ "Π½Π΅ΠΏΠ΅ΡΠ΅ΡΠ΅ΠΊΠ°ΡΡΠΈΠ΅" Π·Π°ΠΏΠΈΡΠΈ
Mangkono, kita perdagangan 50% saka data diwaca kanggo 20% wektu eksekusi. Sing, yen sampeyan duwe alasan kanggo pracaya maca bisa njupuk wektu dawa (contone, data asring ora ing cache, lan sampeyan kudu pindhah menyang disk kanggo iku), banjur ing cara iki sampeyan bisa gumantung kurang ing maca. .
Ing kasus apa wae, wektu eksekusi dadi luwih apik tinimbang ing pilihan pisanan "naif". Nanging endi saka 3 pilihan iki kanggo sampeyan.
Source: www.habr.com