Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.sai.msu.su/~megera/postgres/talks/semi-structured-postgresql-japan.pdf Äàòà èçìåíåíèÿ: Sat Sep 27 15:24:23 2014 Äàòà èíäåêñèðîâàíèÿ: Mon Apr 11 08:10:20 2016 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: m 43

Schema-less PostgreSQL
Current and Future
September, 2014, Tokyo, Japan

Oleg Bartunov, SAI MSU


Oleg Bartunov, Teodor Sigaev
· Locale support · Extendability (indexing)
· GiST (KNN), GIN, SP-GiST

· Full Text Search (FTS) · Jsonb, VODKA · Extensions:
· · · · · intarray pg_trgm ltree hstore plantuner

https://www.facebook.com/oleg.bartunov obartunov@gmail.com, teodor@sigaev.ru https://www.facebook.com/groups/postgresql/


Alexander Korotkov
· · · · · Indexed regexp search GIN compression & fast scan Fast GiST build Range types indexing Split for GiST

aekorotkov@gmail.com


Agenda
· · · · · · The problem Hstore Introduction to jsonb indexing Jsquery - Jsonb Query Language Exercises on jsonb GIN opclasses with Jsquery support VODKA access method


The problem
· The world of data and applications is changing · BIG DATA (Volume of data,Velocity of data in-out, Variety of data) · Web applications are service-oriented
· · · · Service itself can aggregate data, check consistency of data High concurrency, simple queries Simple database (key-value) is ok Eventual consistency is ok, no ACID overhead

· Application needs faster releases · NoSQL databases match all of these -- scalable, efficient, fault-tolerant, no rigid schema, ready to accept any data.


NoSQL
· Key-value databases
· Ordered k-v for ranges support

· Column family (column-oriented) stores · Big Table -- value has structure:
· column families, columns, and timestamped versions (maps-of maps-of maps)

· Document databases · Value has arbitrary structure · Graph databases -- evolution od ordered-kv



The problem
· What if application needs ACID and flexibility of NoSQL ? · Relational databases work with data with schema known in advance · One of the major compaints to relational databases is rigid schema. It's not easy to change schema online (ALTER TABLE ... ADD COLUMN...) · Application should wait for schema changing, infrequent releases · NoSQL uses json format, why not have it in relational database ?

JSON in PostgreSQL This is the challenge !


Challenge to PostgreSQL !
· Full support of semi-stuctured data in PostgreSQL
· · · · Storage Operators and functions Efficiency (fast access to storage, indexes) Integration with CORE (planner, optimiser)

· Actually, PostgreSQL is schema-less database since 2003 -- hstore, one of the most popular extension !


Google insights about hstore


Introduction to Hstore
id col1 col2 col3 col4 col5

A lot of columns key1, .... keyN




The problem:

Total number of columns may be very large Only several fields are searchable ( used in WHERE) Other columns are used only to output These columns may not known in advance



Solution


New data type (hstore), which consists of (key,value) pairs (a'la perl hash)


Introduction to Hstore
id col1 col2 col3 col4 col5

Hstore key1=>val1, key2=>val2,.....



Easy to add key=>value pair No need change schema, just change hstore.



Schema-less PostgreSQL in 2003 !


Introduction to hstore
· Hstore -- key/value binary storage
'a=>1, b=>2'::hstore
· Key, value -- strings · Get value for a key: hstore -> text · Operators with indexing support (GiST, GIN) Check for key: hstore ? text Contains: hstore @> hstore · check documentations for more · Functions for hstore manipulations (akeys, avals, skeys, svals, each,......)
(inspired by perl hash)

· Hstore provides PostgreSQL schema-less feature !
· Faster releases, no problem with schema upgrade


Hstore binary storage
HEntry array String array

Varlena header

Npairs:31

Key endpos: 31

Val endpos: 31

...

key

val

...

New version flag:1

ISNULL:1

Start First key i-th key i-th value
Oleg Bartunov, Teodor Sigaev

End HEntry[0] HEntry[i*2] HEntry[i*2 + 1]
PGConf.EU, Dublin, Nov 1, 2013

0 HEntry[i*2 - 1] HEntry[i*2]
Nested hstore with array support

Pairs are lexicographically ordered by key


Hstore limitations



Levels: Numbe Numbe Length

unlimited r of elements in array: 2^31 r of pairs in hash: 2^31 of string: 2^31 bytes 2^31 bytes = 2 GB

Oleg Bartunov, Teodor Sigaev

Nested hstore with array support

PGConf.EU, Dublin, Nov 1, 2013


History of hstore development
· May 16, 2003 -- first version of hstore


History of hstore development
· May 16, 7.3 · Dec, 05, (thanks, · May 23, · Sep, 20, 2003 - first (unpublished) version of hstore for PostgreSQL 2006 - hstore is a part of PostgreSQL 8.2 Hubert Depesz Lubaczewski!) 2007 - GIN index for hstore, PostgreSQL 8.3 2010 - Andrew Gierth improved hstore, PostgreSQL 9.0


Inverted Index

Btree

Btree


Inverted Index

QUERY: compensation accelerometers INDEX: accelerometers compensation 5,10,25,28,30,36,58,59,61,73,74 30,68 RESULT:

30


GIN improvements
· GIN in 9.4 is greatly improved
· Posting lists compression (varbyte encoding) -- smaller indexes
· 9.3: always 6 bytes (4 bytes blockNumber , 2 bytes offset): 90 bytes (0,8) (0,14) (0,17) (0,22) (0,26) (0,33) (0,34) (0,35) (0,45) (0,47)
(1,6) (1,8) (0,48) (1,3) (1,4)

· 9.4: 1-6 bytes per each item, deltas from previous item: 21 bytes
SELECT g % 10 FROM

(0,8) +6 +3 +5 +4 +7 +1 +1 +10 +2 +1 +2051 +1+2 +2 generate_series(1,10000000) g; 11Mb vs 58Mb

· Fast scan of posting lists - «rare & frequent» queries much faster
· 9.3: read posting lists for «rare» and «frequent» and join them Time(frequent & rare) ~ Time(frequent) · 9.4: start from posting list for «rare» and skip «frequent» list if no match Time(frequent & rare) ~ Time(rare)


Hstore is DEAD ? No !
· How hstore benefits by GIN improvement in 9.4 ?
GIN stands for Generalized Inverted Index, so virtually all data types, which use GIN, get benefit !
· · · · Default hstore GIN opclass considers keys and values separately Keys are «frequent», value are «rare» Contains query: hstore @> 'key=>value' improved a lot for «rare» values Index size is smaller, less io


Hstore 9.3 vs 9.4
Total: 7240858 geo records: "fcode"=>"RFSU", "point"=>"(8.85,112.53333)", "fclass"=>"U", "asciiname"=>"London Reefs", "elevation"=>NULL, "geonameid"=>"1879967", "population"=>"0" Query: SELECT count(*) FROM geo WHERE geo @> 'fcode=>STM'; gin_hstore_ops: index keys and values gin_hstore_bytea_ops = gin_hstore_ops, no collation comparison gin_hstore_hash_ops: index hash(key.value)


Hstore 9.3 vs 9.4
9.3
|-------------------------+-------+----------+-------+---------| | Name | Type | Owner | Table | Size | |-------------------------+-------+----------+-------+---------| | geo | table | postgres | | 1352 MB | | geo_hstore_bytea_ops | index | postgres | geo | 1680 MB | | geo_hstore_hash_ops_idx | index | postgres | geo | 1073 MB | |-------------------------+-------+----------+-------+---------|

9.4
|-------------------------+-------+----------+-------+---------| | Name | Type | Owner | Table | Size | |-------------------------+-------+----------+-------+---------| | geo | table | postgres | | 1352 MB | | geo_hstore_bytea_ops | index | postgres | geo | 1296 MB | | geo_hstore_hash_ops_idx | index | postgres | geo | 925 MB | |-------------------------+-------+----------+-------+---------|

CREATE OPERATOR CLASS gin_hstore_bytea_ops FOR TYPE hstore ........................................................................................ FUNCTION 1 byteacmp(bytea,bytea), ........................................................................................ STORAGE bytea; CREATE INDEX: 239 s Much faster comparison (no collation)

CREATE OPERATOR CLASS gin_hstore_ops FOR TYPE hstore ........................................................................................ FUNCTION 1 bttextcmp(text,text),, ........................................................................................ STORAGE text; CREATE INDEX: 2870 s


Hstore 9.3 vs 9.4
SUMMARY: 9.4 GIN posting list compression: indexes are smaller 9.4 GIN is smart regarding 'freq & rare' queries: time (freq & rare) ~ time (rare) instead of time (freq & rare) ~ time (freq) gin_hstore_hash_ops is good on 9.3 & 9.4 and faster default gin opclass Use gin_hstore_bytea_ops instead of default gin_hstore_ops -- much faster create index


Get hstore_ops from: from https://github.com/akorotkov/hstore_ops


Introduction to hstore
· Hstore benefits
· In provides a flexible model for storing a semi-structured data in relational database · hstore has binary storage and rich set of operators and functions, indexes

· Hstore drawbacks
· Too simple model ! Hstore key-value model doesn't supports tree-like structures as json (introduced in 2006, 3 years after hstore)

· Json -- popular and standartized (ECMA-404 The JSON Data Interchange Standard, JSON RFC-7159) · Json -- PostgreSQL 9.2, textual storage


Hstore vs Json
· hstore is faster than json even on simple data
CREATE TABLE hstore_test AS (SELECT 'a=>1, b=>2, c=>3, d=>4, e=>5'::hstore AS v FROM generate_series(1,1000000)); CREATE TABLE json_test AS (SELECT '{"a":1, "b":2, "c":3, "d":4, "e":5}'::json AS v FROM generate_series(1,1000000));

SELECT sum((v->'a')::text::int) FROM json_test; 851.012 ms SELECT sum((v->'a')::int) FROM hstore_test; 330.027 ms


Hstore vs Json
· PostgreSQL already has json since 9.2, which supports documentbased model, but
· It's slow, since it has no binary representation and needs to be parsed every time · Hstore is fast, thanks to binary representation and index support · It's possible to convert hstore to json and vice versa, but current hstore is limited to key-value

· Need hstore with document-based model. Share it's binary representation with json !


Nested hstore


Nested hstore & jsonb
· Nested hstore at PGCon-2013, Ottawa, Canada ( May 24) -- thanks Engine Yard for support !
One step forward true json data type.Nested hstore with arrays support

· Binary storage for nested data at PGCon Europe -- 2013, Dublin, Ireland (Oct 29)
Binary storage for nested data structuresand application to hstore data type

· November, 2013 -- binary storage was reworked, nested hstore and jsonb share the same storage. Andrew Dunstan joined the project. · January, 2014 - binary storage moved to core


Nested hstore & jsonb
· Feb-Mar, 2014 - Peter Geoghegan joined the project, nested hstore was cancelled in favour to jsonb (Nested hstore patch for 9.3). · Mar 23, 2014 Andrew Dunstan committed jsonb to 9.4 branch !
pgsql: Introduce jsonb, a structured format for storing json.
Introduce jsonb, a structured format for storing json. The new format accepts exactly the same data as the json type. However, it is stored in a format that does not require reparsing the orgiginal text in order to process it, making it much more suitable for indexing and other operations. Insignificant whitespace is discarded, and the order of object keys is not preserved. Neither are duplicate object keys kept - the later value for a given key is the only one stored.


Jsonb vs Json
SELECT '{"c":0, "a":2,"a":1}'::json, '{"c":0, json | jsonb -----------------------+-----------------{"c":0, "a":2,"a":1} | {"a": 1, "c": 0} (1 row) "a":2,"a":1}'::jsonb;

· · · ·

js js js js

o o o o

n: nb: nb: nb:

textual storage «as is» no whitespaces no duplicate keys, last key win keys are sorted


Jsonb vs Json
· Data · 1,252,973 Delicious bookmarks · Server · MBA, 8 GB RAM, 256 GB SSD · Test · Input performance - copy data to table · Access performance - get value by key · Search performance contains @> operator


Jsonb vs Json
· Data · 1,252,973 bookmarks from Delicious in json format ( js) · The same bookmarks in jsonb format ( jb) · The same bookmarks as text (tx)
=# \dt+ List of relations Schema | Name | Type | Owner | Size | Description --------+------+-------+----------+---------+------------public | jb | table | postgres | 1374 MB | overhead is < 4% public | js | table | postgres | 1322 MB | public | tx | table | postgres | 1322 MB |


Jsonb vs Json
· Input performance (parser) Copy data (1,252,973 rows) as text, json,jsonb
copy tt from '/path/to/test.dump'

Text: 34 s Json: 37 s Jsonb: 43 s

- as is - json validation - json validation, binary storage


Jsonb vs Json (binary storage)
· Access performance -- get value by key
· Base: · Jsonb: · Json: SELECT js FROM js; SELECT j->>'updated' FROM jb; SELECT j->>'updated' FROM js;

Base: Jsonb: Json:

0.6 s 1s 9.6 s

0.4 9

Jsonb ~ 20X faster Json


Jsonb vs Json
EXPLAIN ANALYZE SELECt count(*) FROM js WHERE js #>>'{tags,0,term}' = 'NYC'; QUERY PLAN ---------------------------------------------------------------------------Aggregate (cost=187812.38..187812.39 rows=1 width=0) (actual time=10054.602..10054.602 rows=1 loops=1) -> Seq Scan on js (cost=0.00..187796.88 rows=6201 width=0) (actual time=0.030..10054.426 rows=123 loops=1) Filter: ((js #>> '{tags,0,term}'::text[]) = 'NYC'::text) Rows Removed by Filter: 1252850 Planning time: 0.078 ms Execution runtime: 10054.635 ms (6 rows)

Json: no contains @> operator, search first array element


Jsonb vs Json (binary storage)
EXPLAIN ANALYZE SELECT count(*) FROM jb WHERE jb @> '{"tags":[{"term":"NYC"}]}'::jsonb; QUERY PLAN --------------------------------------------------------------------------------------Aggregate (cost=191521.30..191521.31 rows=1 width=0) (actual time=1263.201..1263.201 rows=1 loops=1) -> Seq Scan on jb (cost=0.00..191518.16 rows=1253 width=0) (actual time=0.007..1263.065 rows=285 loops=1) Filter: (jb @> '{"tags": [{"term": "NYC"}]}'::jsonb) Rows Removed by Filter: 1252688 Planning time: 0.065 ms Execution runtime: 1263.225 ms Execution runtime: 10054.635 ms (6 rows)

Jsonb ~ 10X faster Json


Jsonb vs Json (GIN: key && value)
CREATE INDEX gin_jb_idx ON jb USING gin(jb);
EXPLAIN ANALYZE SELECT count(*) FROM jb WHERE jb @> '{"tags":[{"term":"NYC"}]}'::jsonb; QUERY PLAN --------------------------------------------------------------------------------------Aggregate (cost=4772.72..4772.73 rows=1 width=0) (actual time=8.486..8.486 rows=1 loops=1) -> Bitmap Heap Scan on jb (cost=73.71..4769.59 rows=1253 width=0) (actual time=8.049..8.462 rows=285 loops=1) Recheck Cond: (jb @> '{"tags": [{"term": "NYC"}]}'::jsonb) Heap Blocks: exact=285 -> Bitmap Index Scan on gin_jb_idx (cost=0.00..73.40 rows=1253 width=0) (actual time=8.014..8.014 rows=285 loops=1) Index Cond: (jb @> '{"tags": [{"term": "NYC"}]}'::jsonb) Planning time: 0.115 ms Execution runtime: 8.515 ms Execution runtime: 10054.635 ms (8 rows)

Jsonb ~ 150X faster Json


Jsonb vs Json (GIN: hash path.value)
CREATE INDEX gin_jb_path_idx ON jb USING gin(jb jsonb_path_ops);
EXPLAIN ANALYZE SELECT count(*) FROM jb WHERE jb @> '{"tags":[{"term":"NYC"}]}'::jsonb; QUERY PLAN --------------------------------------------------------------------------------------Aggregate (cost=4732.72..4732.73 rows=1 width=0) (actual time=0.644..0.644 rows=1 loops=1) -> Bitmap Heap Scan on jb (cost=33.71..4729.59 rows=1253 width=0) (actual time=0.102..0.620 rows=285 loops=1) Recheck Cond: (jb @> '{"tags": [{"term": "NYC"}]}'::jsonb) Heap Blocks: exact=285 -> Bitmap Index Scan on gin_jb_path_idx (cost=0.00..33.40 rows=1253 width=0) (actual time=0.062..0.062 rows=285 loops=1) Index Cond: (jb @> '{"tags": [{"term": "NYC"}]}'::jsonb) Planning time: 0.056 ms Execution runtime: 0.668 ms Execution runtime: 10054.635 ms (8 rows)

Jsonb ~ 1800X faster Json


MongoDB 2.6.0
· Load data - ~13 min SLOW !
mongoimport --host localhost -c js --type json < delicious-rss-1250k 2014-04-08T22:47:10.014+0400 3700 1233/second ... 2014-04-08T23:00:36.050+0400 1252000 1547/second 2014-04-08T23:00:36.565+0400 check 9 1252973 2014-04-08T23:00:36.566+0400 imported 1252973 objects

Jsonb 43 s

· Search - ~ 1s (seqscan) THE SAME
db.js.find({tags: {$elemMatch:{ term: "NYC"}}}).count() 285 -- 980 ms

· Search - ~ 1ms (indexscan) Jsonb 0.7ms
db.js.ensureIndex( {"tags.term" : 1} ) db.js.find({tags: {$elemMatch:{ term: "NYC"}}}).


Summary: PostgreSQL 9.4 vs Mongo 2.6.0
· Operator contains @>
· · · · json jsonb jsonb mongo : : : : 10 8.5 0.7 1.0 s ms ms ms seqscan GIN jsonb_ops GIN jsonb_path_ops btree index 636 295 44 387 100

·Table size

· Index size
· jsonb_ops jsonb_path_ops · jsonb_path_ops (tags) · mongo (tags) mongo (tags.term)

· Text : · Json : · Jsonb : Mb (no compression, 815Mb) · mongo : Mb Mb USING gin((jb->'tags') jsonb_path_ops Mb Mb

·Input performance:
34 37 43 13 s s s m

·postgres : 1.3Gb ·mongo : 1.8Gb


Jsonb (Apr, 2014)
· Documentation
· JSON Types, JSON Functions and Operators

· There are many functionality left in nested hstore
· Can be an extension

· Need query language for jsonb
· <,>,&& ... operators for values a.b.c.d && [1,2,10] · Structural queries on paths
*.d && [1,2,10]

· Indexes !


Jsonb query
· Currently, one can search jsonb data using
· Contains operators - jsonb @> jsonb, jsonb <@ jsonb (GIN indexes)
jb @> '{"tags":[{"term":"NYC"}]}'::jsonb

Keys should be specified from root Equivalence operator -- jsonb = jsonb (GIN indexes) · Exists operators -- jsonb ? text, jsonb ?! text[], jsonb ?& text[] (GIN indexes) Only root keys supported · Operators on jsonb parts (functional indexes)
jb WHERE jb ?| '{tags,links}'

SELECT ('{"a": {"b":5}}'::jsonb -> 'a'->>'b')::int > 2; CREATE INDEX ....USING BTREE ( (jb->'a'->>'b')::int);

Very cumbersome, too many functional indexes


Jsonb query
· Need Jsonb query language
· · · · More operators on keys, values Types support Schema support (constraints on keys, values) Indexes support

· Introduce Jsquery - textual data type and @@ match operator

jsonb @@ jsquery


Jsonb query language (Jsquery)
Expr ::= path value_expr | path HINT value_expr | NOT expr | NOT HINT value_expr | NOT value_expr | path '(' expr ')' | '(' expr ')' | expr AND expr | expr OR expr value_expr ::= '=' scalar_value | IN '(' value_list ')' | '=' array | '=' '*' | '<' NUMERIC | '<' '=' NUMERIC | '>' NUMERIC | '>' '=' NUMERIC | '@' '>' array | '<' '@' array | '&' '&' array | IS ARRAY | IS NUMERIC | IS OBJECT | IS STRING | IS BOOLEAN path ::= key | path '.' key_any | NOT '.' key_any ::= '*' | '#' | '% ' | '$' | STRING ........ value_list ::= scalar_value | value_list ',' scalar_value array ::= '[' value_list ']'

key

key_any ::= key | NOT

scalar_value ::= null | STRING | true | false | NUMERIC | OBJECT .......


Jsonb query language (Jsquery)
· # - any element array
SELECT '{"a": {"b": [1,2,3]}}'::jsonb @@ 'a.b.# = 2';

path

::= key | path '.' key_any | NOT '.' key_any ::= '*' | '#' | '% ' | '$' | STRING ........

· % - any key
SELECT '{"a": {"b": [1,2,3]}}'::jsonb @@ '%.b.# = 2';

key

· * - anything
SELECT '{"a": {"b": [1,2,3]}}'::jsonb @@ '*.# = 2';

· $ - current element
select '{"a": {"b": [1,2,3]}}'::jsonb @@ 'a.b.# ($ = 2 OR $ < 3)';

key_any ::= key | NOT

· Use "double quotes" for key !
select 'a1."12222" < 111'::jsquery;


Jsonb query language (Jsquery)
· Scalar
select '{"a": {"b": [1,2,3]}}'::jsonb @@ 'a.b.# IN (1,2,5)';

· Test for key existence
select '{"a": {"b": [1,2,3]}}'::jsonb @@ 'a.b = *';

· Array overlap
select '{"a": {"b": [1,2,3]}}'::jsonb @@ 'a.b && [1,2,5]';

· Array contains
select '{"a": {"b": [1,2,3]}}'::jsonb @@ 'a.b @> [1,2]';

· Array contained

select '{"a": {"b": [1,2,3]}}'::jsonb @@ 'a.b <@ [1,2,3,4,5]';

value_expr ::= '=' scalar_value | IN '(' value_list ')' | '=' array | '=' '*' | '<' NUMERIC | '<' '=' NUMERIC | '>' NUMERIC | '>' '=' NUMERIC | '@' '>' array | '<' '@' array | '&' '&' array | IS ARRAY | IS NUMERIC | IS OBJECT | IS STRING | IS BOOLEAN


Jsonb query language (Jsquery)
· Type checking
select '{"x": true}' @@ 'x IS boolean'::jsquery, '{"x": 0.1}' @@ 'x IS numeric'::jsquery; ?column? | ?column? ----------+---------t |t select '{"a":{"a":1}}' @@ 'a IS object'::jsquery; ?column? ---------t

IS BOOLEAN IS NUMERIC IS ARRAY IS OBJECT IS STRING

select '{"a":["xxx"]}' @@ 'a IS array'::jsquery, '["xxx"]' @@ '$ IS array'::jsquery; ?column? | ?column? ----------+---------t |t


Jsonb query language (Jsquery)
· How many products are similar to "B000089778" and have
product_sales_rank in range between 10000-20000 ?

· SQL

SELECT count(*) FROM jr WHERE (jr->>'product_sales_rank')::int > 10000 and (jr->> 'product_sales_rank')::int < 20000 and ....boring stuff SELECT count(*) FROM jr WHERE jr @@ ' similar_product_ids && ["B000089778"] AND product_sales_rank( $ > 10000 AND $ < 20000)'

· Jsquery · Mongodb
db.reviews.find( { $and :[ {similar_product_ids: { $in ["B000089778"]}}, {product_sales_rank:{$gt:10000, $lt:20000}}] } ).count()


Jsonb query language (Jsquery)
explain( analyze, buffers) select count(*) from jb where jb @> '{"tags":[{"term":"NYC"}]}'::jsonb; QUERY PLAN --------------------------------------------------------------------------------------------------------------Aggregate (cost=191517.30..191517.31 rows=1 width=0) (actual time=1039.422..1039.423 rows=1 loops=1) Buffers: shared hit=97841 read=78011 -> Seq Scan on jb (cost=0.00..191514.16 rows=1253 width=0) (actual time=0.006..1039.310 rows=285 loops=1) Filter: (jb @> '{"tags": [{"term": "NYC"}]}'::jsonb) Rows Removed by Filter: 1252688 Buffers: shared hit=97841 read=78011 Planning time: 0.074 ms

Execution time: 1039.444 ms
explain( analyze,costs off) select count(*) from jb where jb @@ 'tags.#.term = "NYC"'; QUERY PLAN -------------------------------------------------------------------Aggregate (actual time=891.707..891.707 rows=1 loops=1) -> Seq Scan on jb (actual time=0.010..891.553 rows=285 loops=1) Filter: (jb @@ '"tags".#."term" = "NYC"'::jsquery) Rows Removed by Filter: 1252688

Execution time: 891.745 ms


Jsquery (indexes)
· GIN opclasses with jsquery support
· jsonb_value_path_ops -- use Bloom filtering for key matching {"a":{"b":{"c":10}}} 10.( bloom(a) or bloom(b) or bloom(c) )
· Good for key matching (wildcard support) , not good for range query

· jsonb_path_value_ops -- hash path (like jsonb_path_ops) {"a":{"b":{"c":10}}} hash(a.b.c).10
· No wildcard support, no problem with ranges
List of relations Schema | Name | Type | Owner | Table | Size | Description --------+-------------------------+-------+----------+--------------+---------+------------public | jb | table | postgres | | 1374 MB | public | jb_value_path_idx | index | postgres | jb | 306 MB | public | jb_gin_idx | index | postgres | jb | 544 MB | public | jb_path_value_idx | index | postgres | jb | 306 MB | public | jb_path_idx | index | postgres | jb | 251 MB |


Jsquery (indexes)
explain( analyze,costs off) select count(*) from jb where jb @@ 'tags.#.term = "NYC"'; QUERY PLAN ------------------------------------------------------------------------------------------------Aggregate (actual time=0.609..0.609 rows=1 loops=1) -> Bitmap Heap Scan on jb (actual time=0.115..0.580 rows=285 loops=1) Recheck Cond: (jb @@ '"tags".#."term" = "NYC"'::jsquery) Heap Blocks: exact=285 -> Bitmap Index Scan on jb_value_path_idx (actual time=0.073..0.073 rows=285 loops=1) Index Cond: (jb @@ '"tags".#."term" = "NYC"'::jsquery) Execution time: 0.634 ms (7 rows)


Jsquery (indexes)
explain( analyze,costs off) select count(*) from jb where jb @@ '*.term = "NYC"'; QUERY PLAN ------------------------------------------------------------------------------------------------Aggregate (actual time=0.688..0.688 rows=1 loops=1) -> Bitmap Heap Scan on jb (actual time=0.145..0.660 rows=285 loops=1) Recheck Cond: (jb @@ '*."term" = "NYC"'::jsquery) Heap Blocks: exact=285 -> Bitmap Index Scan on jb_value_path_idx (actual time=0.113..0.113 rows=285 loops=1) Index Cond: (jb @@ '*."term" = "NYC"'::jsquery) Execution time: 0.716 ms (7 rows)


Citus dataset
· 3023162 reviews from Citus 1998-2000 years · 1573 MB

{ " " " " " " " " customer_id": "AE22YDHSBFYIP", product_category": "Business & Investing", product_group": "Book", product_id": "1551803542", product_sales_rank": 11611, product_subcategory": "General", product_title": "Start and Run a Coffee Bar (Start & Run a)", review_date": { "$date": 31363200000 }, "review_helpful_votes": 0, "review_rating": 5, "review_votes": 10, "similar_product_ids": [ "0471136174", "0910627312", "047112138X", "0786883561", "0201570483" ] }


Jsquery (indexes)
explain (analyze, costs off) select count(*) from jr where jr @@ ' similar_product_ids && ["B000089778"]'; QUERY PLAN -----------------------------------------------------------------------------------------------Aggregate (actual time=0.359..0.359 rows=1 loops=1) -> Bitmap Heap Scan on jr (actual time=0.084..0.337 rows=185 loops=1) Recheck Cond: (jr @@ '"similar_product_ids" && ["B000089778"]'::jsquery) Heap Blocks: exact=107 -> Bitmap Index Scan on jr_path_value_idx (actual time=0.057..0.057 rows=185 loops=1) Index Cond: (jr @@ '"similar_product_ids" && ["B000089778"]'::jsquery) Execution time: 0.394 ms (7 rows)


Jsquery (indexes)
· No statistics, no planning :(
Not selective, better not use index! explain (analyze, costs off) select count(*) from jr where jr @@ ' similar_product_ids && ["B000089778"] AND product_sales_rank( $ > 10000 AND $ < 20000)'; QUERY PLAN -------------------------------------------------------------------------------------------------------------------------------------Aggregate (actual time=126.149..126.149 rows=1 loops=1) -> Bitmap Heap Scan on jr (actual time=126.057..126.143 rows=45 loops=1) Recheck Cond: (jr @@ '("similar_product_ids" && ["B000089778"] & "product_sales_rank"($ > 10000 & $ < 20000))'::jsquery) Heap Blocks: exact=45 -> Bitmap Index Scan on jr_path_value_idx (actual time=126.029..126.029 rows=45 loops=1) Index Cond: (jr @@ '("similar_product_ids" && ["B000089778"] & "product_sales_rank"($ > 10000 & $ < 20000))'::jsquery) Execution time: 129.309 ms !!! No statistics (7 rows)


MongoDB 2.6.0
db.reviews.find( { $and :[ {similar_product_ids: { $in:["B000089778"]}}, {product_sales_rank:{$gt:10000, $lt:20000}}] } ) .explain() { "n" : 45, .................... "millis" : 7, "indexBounds" : { "similar_product_ids" : [ index size = 400 MB just for similar_product_ids !!! [ "B000089778", "B000089778" ] ] }, }


Jsquery (indexes)
· If we rewrite query and use planner
explain (analyze,costs off) select count(*) from jr where jr @@ ' similar_product_ids && ["B000089778"]' and (jr->>'product_sales_rank')::int>10000 and (jr->>'product_sales_rank')::int<20000; ----------------------------------------------------------------------------------------------------------------------------------------Aggregate (actual time=0.479..0.479 rows=1 loops=1) -> Bitmap Heap Scan on jr (actual time=0.079..0.472 rows=45 loops=1) Recheck Cond: (jr @@ '"similar_product_ids" && ["B000089778"]'::jsquery) Filter: ((((jr ->> 'product_sales_rank'::text))::integer > 10000) AND (((jr ->> 'product_sales_rank'::text))::integer < 20000)) Rows Removed by Filter: 140 Heap Blocks: exact=107 -> Bitmap Index Scan on jr_path_value_idx (actual time=0.041..0.041 rows=185 loops=1) Index Cond: (jr @@ '"similar_product_ids" && ["B000089778"]'::jsquery) Execution time: 0.506 ms Potentially, quer y could be faster Mongo ! (9 rows)


Jsquery (optimiser) -- NEW !
·

Jsquery now has built-in optimiser for simple queries.
explain (analyze, costs off) select count(*) from jr where jr @@ 'similar_product_ids && ["B000089778"] AND product_sales_rank( $ > 10000 AND $ < 20000)' -----------------------------------------------------------------------------------------------------------------------------------------Aggregate (actual time=0.422..0.422 rows=1 loops=1) -> Bitmap Heap Scan on jr (actual time=0.099..0.416 rows=45 loops=1) Recheck Cond: (jr @@ '("similar_product_ids" && ["B000089778"] AND "product_sales_rank"($ > 10000 AND $ < 20000))'::jsquery) Rows Removed by Index Recheck: 140 Heap Blocks: exact=107 -> Bitmap Index Scan on jr_path_value_idx (actual time=0.060..0.060 rows=185 loops=1) Index Cond: (jr @@ '("similar_product_ids" && ["B000089778"] AND "product_sales_rank"($ > 10000 AND $ < 20000))'::jsquery) Execution time: 0.480 ms vs 7 ms MongoDB !


Jsquery (optimiser) -- NEW !
·

Jsquery now has built-in optimiser for simple queries. Analyze query tree and push non-selective parts to recheck (like filter)
Selectivity classes: 1) Equality (x = c) 2) Range (c1 < x < c2) 3) Inequality (c > c1) 4) Is (x is type) 5) Any (x = *)
SELECT gin_debug_query_path_value('similar_product_ids && ["B000089778"] AND product_sales_rank( $ > 10000 AND $ < 20000)'); gin_debug_query_path_value ------------------------------------------------similar_product_ids.# = "B000089778" , entry 0 +


Jsquery (optimiser) -- NEW !
·

Jsquery optimiser pushes non-selective operators to recheck
explain (analyze, costs off) select count(*) from jr where jr @@ 'similar_product_ids && ["B000089778"] AND product_sales_rank( $ > 10000 AND $ < 20000)' -----------------------------------------------------------------------------------------------------------------------------------------Aggregate (actual time=0.422..0.422 rows=1 loops=1) -> Bitmap Heap Scan on jr (actual time=0.099..0.416 rows=45 loops=1) Recheck Cond: (jr @@ '("similar_product_ids" && ["B000089778"] AND "product_sales_rank"($ > 10000 AND $ < 20000))'::jsquery) Rows Removed by Index Recheck: 140 Heap Blocks: exact=107 -> Bitmap Index Scan on jr_path_value_idx (actual time=0.060..0.060 rows=185 loops=1) Index Cond: (jr @@ '("similar_product_ids" && ["B000089778"] AND "product_sales_rank"($ > 10000 AND $ < 20000))'::jsquery) Execution time: 0.480 ms


Jsquery (HINTING) -- NEW !
·

Jsquery now has HINTING ( if you don't like optimiser)!
explain (analyze, costs off) select count(*) from jr where jr @@ 'product_sales_rank > 10000' ---------------------------------------------------------------------------------------------------------Aggregate (actual time=2507.410..2507.410 rows=1 loops=1) -> Bitmap Heap Scan on jr (actual time=1118.814..2352.286 rows=2373140 loops=1) Recheck Cond: (jr @@ '"product_sales_rank" > 10000'::jsquery) Heap Blocks: exact=201209 -> Bitmap Index Scan on jr_path_value_idx (actual time=1052.483..1052.48 rows=2373140 loops=1) Index Cond: (jr @@ '"product_sales_rank" > 10000'::jsquery) Execution time: 2524.951 ms

· Better not to use index -- HINT /* --noindex */
explain (analyze, costs off) select count(*) from jr where jr @@ 'product_sales_rank /*-- noindex */ > 10000'; ---------------------------------------------------------------------------------Aggregate (actual time=1376.262..1376.262 rows=1 loops=1) -> Seq Scan on jr (actual time=0.013..1222.123 rows=2373140 loops=1) Filter: (jr @@ '"product_sales_rank" /*-- noindex */ > 10000'::jsquery) Rows Removed by Filter: 650022 Execution time: 1376.284 ms


Contrib/jsquery
· Jsquery index support is quite efficient ( 0.5 ms vs Mongo 7 ms ! ) · Future direction
· Make jsquery planner friendly · Need statistics for jsonb

· Availability
· Jsquery + opclasses are available as extensions · Grab it from https://github.com/akorotkov/jsquery (branch master) , we need your feedback ! · We will release it after PostgreSQL 9.4 release · Need real sample data and queries !


PostgreSQL 9.4+ Open-source Relational database Strong support of json


Better indexing ...
· GIN is a proven and effective index access method · Need indexing for jsonb with operations on paths (no hash!) and values
· B-tree in entry tree is not good - length limit, no prefix compression
List of relations Schema | Name | Type | Owner | Table | Size | Description --------+-----------------------------+-------+----------+---------------+---------+------------public | jb | table | postgres | | 1374 MB | public | jb_uniq_paths | table | postgres | | 912 MB | public | jb_uniq_paths_btree_idx | index | postgres | jb_uniq_paths | 885 MB |text_pattern_ops public | jb_uniq_paths_spgist_idx | index | postgres | jb_uniq_paths | 598 MB |now much less !


Better indexing ...
· Provide interface to change hardcoded B-tree in Entry tree
· Use spgist opclass for storing paths and values as is (strings hashed in values)

· We may go further - provide interface to change hardcoded B-tree in posting tree
· GIS aware full text search !

· New index access method

CREATE INDEX ... USING VODKA


GIN History
· Introduced at PostgreSQL Anniversary Meeting in Toronto, Jul 7-8, 2006 by Oleg Bartunov and Teodor Sigaev


GIN History
· Introduced at PostgreSQL Anniversary Meeting in Toronto, Jul 7-8, 2006 by Oleg Bartunov and Teodor Sigaev · Supported by JFG Networks (France) · «Gin stands for Generalized Inverted iNdex and should be considered as a genie, not a drink.» · Alexander Korotkov, Heikki Linnakangas have joined GIN++ development in 2013


GIN History
· From GIN Readme, posted in -hackers, 2006-04-26
TODO ---Nearest future: * Opclasses for all types (no programming, just many catalog changes). Distant future: * Replace B-tree of entries to something like GiST (VODKA ! 2014) * Add multicolumn support * Optimize insert operations (background index insertion)


GIN index structure for jsonb
{ }, { } "product_group": "Book", "product_sales_rank": 15000 "product_group": "Music", "product_sales_rank": 25000


Vodka index structure for jsonb
{ }, { } "product_group": "Book", "product_sales_rank": 15000 "product_group": "Music", "product_sales_rank": 25000


CREATE INDEX ... USING VODKA
· Delicious bookmarks, mostly text data
set maintenance_work_mem = '1GB'; List of relations Schema | Name | Type | Owner | Table | Size | Description --------+--------------------+-------+----------+-------+---------+------------public | jb | table | postgres | | 1374 MB | 1252973 rows public | jb_value_path_idx | index | postgres | jb | 306 MB | 98769.096 public | jb_gin_idx | index | postgres | jb | 544 MB | 129860.859 public | jb_path_value_idx | index | postgres | jb | 306 MB | 100560.313 public | jb_path_idx | index | postgres | jb | 251 MB | 68880.320 public | jb_vodka_idx | index | postgres | jb | 409 MB | 185362.865 public | jb_vodka_idx5 | index | postgres | jb | 325 MB | 174627.234 new spgist (6 rows)


CREATE INDEX ... USING VODKA
select count(*) from jb where jb @@ 'tags.#.term = "NYC"'; ------------------------------------------------------------------------------------------Aggregate (actual time=0.423..0.423 rows=1 loops=1) -> Bitmap Heap Scan on jb (actual time=0.146..0.404 rows=285 loops=1) Recheck Cond: (jb @@ '"tags".#."term" = "NYC"'::jsquery) Heap Blocks: exact=285 -> Bitmap Index Scan on jb_vodka_idx (actual time=0.108..0.108 rows=285 loops=1) Index Cond: (jb @@ '"tags".#."term" = "NYC"'::jsquery)

Execution time: 0.456 ms (0.634 ms, GIN jsonb_value_path_ops)
select count(*) from jb where jb @@ '*.term = "NYC"'; ------------------------------------------------------------------------------------------Aggregate (actual time=0.495..0.495 rows=1 loops=1) -> Bitmap Heap Scan on jb (actual time=0.245..0.474 rows=285 loops=1) Recheck Cond: (jb @@ '*."term" = "NYC"'::jsquery) Heap Blocks: exact=285 -> Bitmap Index Scan on jb_vodka_idx (actual time=0.214..0.214 rows=285 loops=1) Index Cond: (jb @@ '*."term" = "NYC"'::jsquery)

Execution time: 0.526 ms (0.716 ms, GIN jsonb_path_value_ops)


CREATE INDEX ... USING VODKA
· CITUS data, text and numeric
set maintenance_work_mem = '1GB'; List of relations Schema | Name | Type | Owner | Table | Size | Description --------+--------------------+-------+----------+-------+---------+------------public | jr | table | postgres | | 1573 MB | 3023162 rows public | jr_value_path_idx | index | postgres | jr | 196 MB | 79180.120 public | jr_gin_idx | index | postgres | jr | 235 MB | 111814.929 public | jr_path_value_idx | index | postgres | jr | 196 MB | 73369.713 public | jr_path_idx | index | postgres | jr | 180 MB | 48981.307 public | jr_vodka_idx3 | index | postgres | jr | 240 MB | 155714.777 public | jr_vodka_idx4 | index | postgres | jr | 211 MB | 169440.130 new spgist (6 rows)


CREATE INDEX ... USING VODKA
explain (analyze, costs off) select count(*) from jr where jr @@ ' similar_product_ids && ["B000089778"]'; QUERY PLAN ------------------------------------------------------------------------------------------Aggregate (actual time=0.200..0.200 rows=1 loops=1) -> Bitmap Heap Scan on jr (actual time=0.090..0.183 rows=185 loops=1) Recheck Cond: (jr @@ '"similar_product_ids" && ["B000089778"]'::jsquery) Heap Blocks: exact=107 -> Bitmap Index Scan on jr_vodka_idx (actual time=0.077..0.077 rows=185 loops=1) Index Cond: (jr @@ '"similar_product_ids" && ["B000089778"]'::jsquery)

Execution time: 0.237 ms (0.394 ms, GIN jsonb_path_value_idx)
(7 rows)


There are can be different flavors of Vodka


Spaghetti indexing ...

Find twirled spaghetti


Spaghetti indexing ...

R-tree fails here -- bounding box of each separate spaghetti is the same


Spaghetti indexing ...

R-tree fails here -- bounding box of each separate spaghetti is the same


Ottawa downtown: York and George streets


Idea: Use multiple boxes


Rtree Vodka


Summary
· contrib/jsquery for 9.4
· Jsquery - Jsonb Query Language · Two GIN opclasses with jsquery support · Grab it from https://github.com/akorotkov/jsquery (branch master)

· Prototype of VODKA access method · Plans for improving indexing infrastructure · This work was supported by


Another view on VODKA
· VODKA CONNECTING INDEXES
· composite index, which combines different access methods · Nested search trees


!



VODKA Optimized Dendriform Keys Array