Slow triggers checking primary key constraints with postgresql - postgresql

I am trying to check primary key constraints along the inheritance tree in postgresql using a plpythonu trigger (check_pk). I'm using (PostgreSQL) 9.4.5.
My question is why does the insert or update take at least 50 ms, when the execution of the trigger itself takes at most 5 ms? Am I doing something wrong? Is there a way to improve this?
The code
(git#github.com:collorg/oopg.git)
The tables:
I'm testing with three tables parent, childb and childc (see bellow for the trigger):
create table parent(
a text primary key
);
create trigger check_pk
before insert or update on parent
for each row execute procedure check_pk();
create table childb(
b text,
primary key(a, b)
) inherits(parent);
create trigger check_pk
before insert or update on childb
for each row execute procedure check_pk();
create table childc(
c text,
primary key(a, c)
) inherits(parent);
create trigger check_pk
before insert or update on childc
for each row execute procedure check_pk();
The tests:
insert into parent (a) values ('a') -- OK
insert into childb (a, b) values ('a', 'a') -- FAILS
insert into childb (a, b) values ('b', 'a') -- OK
insert into parent (a) values ('b') -- FAILS
insert into parent (a) values ('b') -- FAILS
insert into childc (a, c) values ('b', 'a') -- FAILS
insert into childc (a, c) values ('c', 'a') -- OK
select * from parent -- a, b, c
update parent set a = 'b' -- FAILS
update childb set a = 'c' -- FAILS
update childb set a = 'd' -- OK
Here is an excerpt of the postgresql logs (I've set log_min_duration_statement to 10 ms in postgresql.conf):
======== get_pk_fields(59959)
check_pk_oid: SELECT a FROM public.parent WHERE a = 'c' limit 1
CLEF DUPLIQUEE
check_pk_oid duration: 0:00:00.003948
check_pk duration: 0:00:00.004504
2015-12-10 08:53:16 CET LOG: durée : 71.940 ms, instruction : update parent set a = 'c'
The update takes 71.940 ms when the execution of the check_pk trigger takes 4.5 ms.
The trigger:
create language plpythonu;
--
--
--
CREATE FUNCTION check_pk()
RETURNS trigger
AS $$
from datetime import datetime
from sys import stderr
begin = datetime.now()
oid = TD['relid']
GD['td'] = TD
ok = plpy.execute(
"SELECT check_pk_oid({})".format(oid))[0]['check_pk_oid']
stderr.write("check_pk duration: {}\n".format(datetime.now() - begin))
if not ok:
return 'SKIP'
$$ LANGUAGE plpythonu;
--
--
--
CREATE FUNCTION check_pk_oid(integer)
returns boolean
AS $$
"""Return False if the key is found in any of the parents."""
from datetime import datetime
from sys import stderr
from psycopg2.extensions import adapt
begin = datetime.now()
oid = args[0]
stderr.write("{} check_pk_oid({})\n".format(8*'=', oid))
TD = GD['td']
stderr.write("GD['td'] = {}\n".format(TD))
parent_oid = plpy.execute(
"SELECT get_inhparent('{}')".format(oid))[0]['get_inhparent']
stderr.write("oid du parent {}\n".format(parent_oid))
if parent_oid:
# recurse on parent_oid
query = ("SELECT check_pk_oid({})".format(parent_oid))
stderr.write("check uid request: {}\n".format(query))
return plpy.execute(query)[0]['check_pk_oid']
# Get the FQTN and the field names of the primary key
pk_infos = plpy.execute(
"SELECT get_pk_fields({})".format(oid))[0]['get_pk_fields']
fqtn, pk_fieldnames = pk_infos[0], pk_infos[1:]
if not pk_fieldnames:
stderr.write(
"check_pk_oid duration ok 1: {}\n".format(datetime.now() - begin))
return True
# Clause for the SELECT request
fields = []
clause = []
for field in pk_fieldnames:
fields.append(field)
if TD['new'][field] == 0:
valeur = 0
else:
valeur = TD['new'][field] or ""
valeur = adapt(valeur)
clause.append("{} = {}".format(field, str(valeur)))
# construction de la requête d''extraction
req = "SELECT {} FROM {} WHERE {} limit 1".format(
', '.join(fields), fqtn, ' and '.join(clause))
stderr.write("check_pk_oid: {}\n".format(req))
if len(plpy.execute(req)) == 1:
stderr.write("CLEF DUPLIQUEE\n")
stderr.write("check_pk_oid duration: {}\n".format(datetime.now() - begin))
return False
stderr.write("check_pk_oid duration ok 2: {}\n".format(datetime.now() - begin))
return True
$$ LANGUAGE plpythonu;
--
--
--
CREATE FUNCTION get_inhparent(integer)
RETURNS integer
AS $$
from sys import stderr
relid = args[0]
stderr.write("{} get_inhparent({})\n".format(8*'=', relid))
query = (
"SELECT inhparent FROM pg_catalog.pg_inherits WHERE inhrelid = {}".format(
relid))
stderr.write('get_inhparent: {}\n'.format(query))
rec = plpy.execute(query)
try:
return rec[0]['inhparent']
except:
return 0
$$ LANGUAGE plpythonu;
--
--
--
CREATE FUNCTION get_pk_fields(oid)
RETURNS varchar[]
AS $$
"""
Return the field names in the primary key
"""
from sys import stderr
oid = args[0]
stderr.write("{} get_pk_fields({})\n".format(8*'=', oid))
# rec_st : record contenant schemaname et relname
rec_st = plpy.execute(
"""SELECT schemaname, relname
FROM pg_catalog.pg_stat_all_tables
WHERE relid = {}""".format(oid))
schemaname = rec_st[0]['schemaname']
relname = rec_st[0]['relname']
l_fieldnames = plpy.execute(
"""
SELECT
a.attrelid AS tableid,
c.relkind AS tablekind,
n.nspname::varchar AS schemaname,
c.relname::varchar AS relationname,
array_agg(distinct i.inhparent) as parent,
array_agg(a.attname::varchar) AS fieldnames,
array_agg(a.attnum) as attnums,
array_agg(a.attislocal) AS local,
cn_pk.contype AS pkey
FROM
pg_class c -- table
LEFT JOIN pg_namespace n ON
c.relname = '{}' and
n.oid = c.relnamespace and
n.nspname = '{}'
LEFT JOIN pg_inherits i ON
i.inhrelid = c.oid
LEFT JOIN pg_attribute a ON
a.attrelid = c.oid
JOIN pg_type pt ON
a.atttypid = pt.oid
-- LEFT JOIN pg_constraint cn_uniq ON
-- cn_uniq.contype = 'u' AND
-- cn_uniq.conrelid = a.attrelid AND
-- a.attnum = ANY( cn_uniq.conkey )
JOIN pg_constraint cn_pk ON
cn_pk.contype = 'p' AND
cn_pk.conrelid = a.attrelid AND
a.attnum = ANY( cn_pk.conkey )
WHERE
n.nspname <> 'pg_catalog'::name AND
n.nspname <> 'information_schema'::name AND
( c.relkind = 'r'::"char" )
GROUP BY
a.attrelid,
c.relkind,
n.nspname,
c.relname,
cn_pk.contype""".format(relname, schemaname))[0]['fieldnames']
fqtn = "{}.{}".format(schemaname, relname)
return [fqtn] + l_fieldnames
fieldnames = ','.join(l_fieldnames)
resultat = fqtn + ":" + fieldnames
stderr.write("{}\n".format(resultat))
return resultat
$$ LANGUAGE plpythonu;

I've got my answer. plpython is great for prototyping but it comes with a cost. If I use this trigger which basically does nothing, the insert time is around 30 ms... So I guess I'll have to code that in C if I want better performances.
create language plpythonu;
CREATE OR REPLACE FUNCTION slow()
RETURNS trigger
AS $$
pass
$$ LANGUAGE plpythonu;
Note: the code posted here is obsolete and buggy (no multiple inheritance, ...) but it continues to evolve on https://github.com/collorg/oopg.

Related

Select columns that are primary key or foreign key only

Assuming that I have the following table
CREATE TABLE my_table (
a TEXT NULL,
id TEXT PRIMARY KEY NOT NULL,
c TEXT NULL,
d TEXT REFERENCES other_table_1(id) NOT NULL,
e TEXT REFERENCES other_table_2(id) NOT NULL
);
I want to perform a select statement that only select important column that is primary key and foreign key only.
SELECT (...?) FROM my_table
Expected output columns only id, d, e
What is the best non-hacky way I can achieve this?
You can create a function that build the sql statement for you and after you can execute the result.
CREATE OR REPLACE FUNCTION build_select(_tbl regclass)
RETURNS text AS
$func$
SELECT format('SELECT %s FROM %s'
, string_agg(quote_ident(important_column), ', ')
, $1)
FROM (SELECT a.attname as important_column
FROM pg_index i
JOIN pg_attribute a ON a.attrelid = i.indrelid
AND a.attnum = ANY(i.indkey)
WHERE i.indrelid = $1
AND i.indisprimary
UNION
SELECT ta.attname AS important_column
FROM (
SELECT conname, conrelid, confrelid,
unnest(conkey) AS conkey, unnest(confkey) AS confkey
FROM pg_constraint
WHERE conrelid = $1
) sub
JOIN pg_attribute AS ta ON ta.attrelid = conrelid AND ta.attnum = conkey
JOIN pg_attribute AS fa ON fa.attrelid = confrelid AND fa.attnum = confkey) my_sub_query
$func$ LANGUAGE sql;
With this function, if you do:
SELECT build_select('my-schema.my-table')
Will return you:
SELECT id, d, e FROM my-schema.my-table
And you can execute this one

Table bloat on Postgres

I have a small (~200GB) data warehouse running on Postgres 9.5.15 on AWS RDS instance.
For robustness, I'm inserting new data into analytical schema (result of ELT) as follows:
insert new slice
remove the old slice using delete command
vacuum
I know Postgres soft deletes tuples when you execute delete or update commands. This is not a concern giving the table sizes. The problem is that the dead tuples are not removed with explicit vacuum on (3) or regular autovacuum. So, if the pipeline is executed many times, I end up with HUGE table bloat that affects table performance a lot, not mentioning extra storage wasted.
Moreover, when I started investigating I found out that even system tables have this issue:
schemaname | relname | n_live_tup | n_dead_tup | ratio%
pg_catalog | pg_attribute | 46081 | 8339587 | 18097
pg_catalog | pg_depend | 27375 | 2490507 | 9097
pg_catalog | pg_statistic20094 | 1208474 | 6013
That might make general performance of the instance worse in ways I can't even imagine. When I try to do VACUUM FULL VERBOSE pg_catalog.pg_attribute it gives me this:
"pg_attribute": found 0 removable, 8387117 nonremovable row versions in 152494 pages
I have read the articles like "3 reasons of table bloat" but neither does not apply (I'm not doing replication, I don't have hanging transactions, etc.). I can use something like pg_repack to get rid of the bloat at some schedule but I'd like to understand the reason why it happens. Also I don't want to repack system tables for sure.
My only hypothesis is that vacuum requires all the dead tuples to fit in memory that is limited to maintenance_work_mem setting (127MB for our instance) and we need to increase that but I need a side opinion first.
I ended up writing my own functions to repack the data and running them on schedule:
-- repack an individual table
CREATE OR REPLACE FUNCTION admin.repack_table(text)
RETURNS text
AS $$
DECLARE SQL text;
BEGIN
SELECT
'CREATE TEMP TABLE t1 (LIKE '||$1||');'||chr(10)||
'INSERT INTO t1 SELECT * FROM '||$1||';'||chr(10)||
'TRUNCATE TABLE '||$1||';'||chr(10)||
'INSERT INTO '||$1||' SELECT * FROM t1;'||chr(10)||
'DROP TABLE t1;'||chr(10)||
'ANALYZE '||$1||';'
INTO SQL;
EXECUTE SQL;
RETURN $1;
END;
$$ LANGUAGE plpgsql;
-- repack all tables in certain schema (with an optional threshold for N of dead tuples)
CREATE OR REPLACE FUNCTION admin.repack_schema(text,int default 5000)
RETURNS table (table_name text)
AS $$
DECLARE SQL text;
BEGIN
RETURN QUERY (
with
schema as (select $1)
select admin.repack_table(t.table_schema||'.'||t.table_name)
from information_schema.tables t
where t.table_schema=(select * from schema)
and t.table_name in (
select relname
from pg_stat_all_tables
where schemaname=(select * from schema)
and n_dead_tup>$2
and n_live_tup<1000000 -- avoid repacking too large tables
)
);
END;
$$ LANGUAGE plpgsql;
Bloat Score Query
The following SQL query will examine each table in the public schema and identify dead rows (tuples) that are wasting disk space.
SELECT schemaname || '.' || relname as tblnam,
n_dead_tup,
(n_dead_tup::float / n_live_tup::float) * 100 as pfrag
FROM pg_stat_user_tables
WHERE schemaname = 'public' and n_dead_tup > 0 and n_live_tup > 0 order by pfrag desc;
If this query returns a high percentage ( pfrag ) of dead tuples, the VACUUM command may be used to reclaim space.
7 Considered to be high
From wiki.postgres.org
SELECT
current_database(), schemaname, tablename, /*reltuples::bigint, relpages::bigint, otta,*/
ROUND((CASE WHEN otta=0 THEN 0.0 ELSE sml.relpages::float/otta END)::numeric,1) AS tbloat,
CASE WHEN relpages < otta THEN 0 ELSE bs*(sml.relpages-otta)::BIGINT END AS wastedbytes,
iname, /*ituples::bigint, ipages::bigint, iotta,*/
ROUND((CASE WHEN iotta=0 OR ipages=0 THEN 0.0 ELSE ipages::float/iotta END)::numeric,1) AS ibloat,
CASE WHEN ipages < iotta THEN 0 ELSE bs*(ipages-iotta) END AS wastedibytes
FROM (
SELECT
schemaname, tablename, cc.reltuples, cc.relpages, bs,
CEIL((cc.reltuples*((datahdr+ma-
(CASE WHEN datahdr%ma=0 THEN ma ELSE datahdr%ma END))+nullhdr2+4))/(bs-20::float)) AS otta,
COALESCE(c2.relname,'?') AS iname, COALESCE(c2.reltuples,0) AS ituples, COALESCE(c2.relpages,0) AS ipages,
COALESCE(CEIL((c2.reltuples*(datahdr-12))/(bs-20::float)),0) AS iotta -- very rough approximation, assumes all cols
FROM (
SELECT
ma,bs,schemaname,tablename,
(datawidth+(hdr+ma-(case when hdr%ma=0 THEN ma ELSE hdr%ma END)))::numeric AS datahdr,
(maxfracsum*(nullhdr+ma-(case when nullhdr%ma=0 THEN ma ELSE nullhdr%ma END))) AS nullhdr2
FROM (
SELECT
schemaname, tablename, hdr, ma, bs,
SUM((1-null_frac)*avg_width) AS datawidth,
MAX(null_frac) AS maxfracsum,
hdr+(
SELECT 1+count(*)/8
FROM pg_stats s2
WHERE null_frac<>0 AND s2.schemaname = s.schemaname AND s2.tablename = s.tablename
) AS nullhdr
FROM pg_stats s, (
SELECT
(SELECT current_setting('block_size')::numeric) AS bs,
CASE WHEN substring(v,12,3) IN ('8.0','8.1','8.2') THEN 27 ELSE 23 END AS hdr,
CASE WHEN v ~ 'mingw32' THEN 8 ELSE 4 END AS ma
FROM (SELECT version() AS v) AS foo
) AS constants
GROUP BY 1,2,3,4,5
) AS foo
) AS rs
JOIN pg_class cc ON cc.relname = rs.tablename
JOIN pg_namespace nn ON cc.relnamespace = nn.oid AND nn.nspname = rs.schemaname AND nn.nspname <> 'information_schema'
LEFT JOIN pg_index i ON indrelid = cc.oid
LEFT JOIN pg_class c2 ON c2.oid = i.indexrelid
) AS sml
ORDER BY wastedbytes DESC
Human readable wasted disk space
WITH constants AS (
SELECT current_setting('block_size')::numeric AS bs, 23 AS hdr, 4 AS ma
), bloat_info AS (
SELECT
ma,bs,schemaname,tablename,
(datawidth+(hdr+ma-(case when hdr%ma=0 THEN ma ELSE hdr%ma END)))::numeric AS datahdr,
(maxfracsum*(nullhdr+ma-(case when nullhdr%ma=0 THEN ma ELSE nullhdr%ma END))) AS nullhdr2
FROM (
SELECT
schemaname, tablename, hdr, ma, bs,
SUM((1-null_frac)*avg_width) AS datawidth,
MAX(null_frac) AS maxfracsum,
hdr+(
SELECT 1+count(*)/8
FROM pg_stats s2
WHERE null_frac<>0 AND s2.schemaname = s.schemaname AND s2.tablename = s.tablename
) AS nullhdr
FROM pg_stats s, constants
GROUP BY 1,2,3,4,5
) AS foo
), table_bloat AS (
SELECT
schemaname, tablename, cc.relpages, bs,
CEIL((cc.reltuples*((datahdr+ma-
(CASE WHEN datahdr%ma=0 THEN ma ELSE datahdr%ma END))+nullhdr2+4))/(bs-20::float)) AS otta
FROM bloat_info
JOIN pg_class cc ON cc.relname = bloat_info.tablename
JOIN pg_namespace nn ON cc.relnamespace = nn.oid AND nn.nspname = bloat_info.schemaname AND nn.nspname <> 'information_schema'
), index_bloat AS (
SELECT
schemaname, tablename, bs,
COALESCE(c2.relname,'?') AS iname, COALESCE(c2.reltuples,0) AS ituples, COALESCE(c2.relpages,0) AS ipages,
COALESCE(CEIL((c2.reltuples*(datahdr-12))/(bs-20::float)),0) AS iotta -- very rough approximation, assumes all cols
FROM bloat_info
JOIN pg_class cc ON cc.relname = bloat_info.tablename
JOIN pg_namespace nn ON cc.relnamespace = nn.oid AND nn.nspname = bloat_info.schemaname AND nn.nspname <> 'information_schema'
JOIN pg_index i ON indrelid = cc.oid
JOIN pg_class c2 ON c2.oid = i.indexrelid
)
SELECT
type, schemaname, object_name, bloat, pg_size_pretty(raw_waste) as waste
FROM
(SELECT
'table' as type,
schemaname,
tablename as object_name,
ROUND(CASE WHEN otta=0 THEN 0.0 ELSE table_bloat.relpages/otta::numeric END,1) AS bloat,
CASE WHEN relpages < otta THEN '0' ELSE (bs*(table_bloat.relpages-otta)::bigint)::bigint END AS raw_waste
FROM
table_bloat
UNION
SELECT
'index' as type,
schemaname,
tablename || '::' || iname as object_name,
ROUND(CASE WHEN iotta=0 OR ipages=0 THEN 0.0 ELSE ipages/iotta::numeric END,1) AS bloat,
CASE WHEN ipages < iotta THEN '0' ELSE (bs*(ipages-iotta))::bigint END AS raw_waste
FROM
index_bloat) bloat_summary
ORDER BY raw_waste DESC, bloat DESC

I am getting Dollar sign unterminated

I want to create a function like below which inserts data as per the input given. But I keep on getting an error about undetermined dollar sign.
CREATE OR REPLACE FUNCTION test_generate
(
ref REFCURSOR,
_id INTEGER
)
RETURNS refcursor AS $$
DECLARE
BEGIN
DROP TABLE IF EXISTS test_1;
CREATE TEMP TABLE test_1
(
id int,
request_id int,
code text
);
IF _id IS NULL THEN
INSERT INTO test_1
SELECT
rd.id,
r.id,
rd.code
FROM
test_2 r
INNER JOIN
raw_table rd
ON
rd.test_2_id = r.id
LEFT JOIN
observe_test o
ON
o.raw_table_id = rd.id
WHERE o.id IS NULL
AND COALESCE(rd.processed, 0) = 0;
ELSE
INSERT INTO test_1
SELECT
rd.id,
r.id,
rd.code
FROM
test_2 r
INNER JOIN
raw_table rd
ON rd.test_2_id = r.id
WHERE r.id = _id;
END IF;
DROP TABLE IF EXISTS tmp_test_2_error;
CREATE TEMP TABLE tmp_test_2_error
(
raw_table_id int,
test_2_id int,
error text,
record_num int
);
INSERT INTO tmp_test_2_error
(
raw_table_id,
test_2_id,
error,
record_num
)
SELECT DISTINCT
test_1.id,
test_1.test_2_id,
'Error found ' || test_1.code,
0
FROM
test_1
WHERE 1 = 1
AND data_origin.id IS NULL;
INSERT INTO tmp_test_2_error
SELECT DISTINCT
test_1.id,
test_1.test_2_id,
'Error found ' || test_1.code,
0
FROM
test_1
INNER JOIN
data_origin
ON
data_origin.code = test_1.code
WHERE dop.id IS NULL;
DROP table IF EXISTS test_latest;
CREATE TEMP TABLE test_latest AS SELECT * FROM observe_test WHERE 1 = 2;
INSERT INTO test_latest
(
raw_table_id,
series_id,
timestamp
)
SELECT
test_1.id,
ds.id AS series_id,
now()
FROM
test_1
INNER JOIN data_origin ON data_origin.code = test_1.code
LEFT JOIN
observe_test o ON o.raw_table_id = test_1.id
WHERE o.id IS NULL;
CREATE TABLE latest_observe_test as Select * from test_latest where 1=0;
INSERT INTO latest_observe_test
(
raw_table_id,
series_id,
timestamp,
time
)
SELECT
t.id,
ds.id AS series_id,
now(),
t.time
FROM
test_latest t
WHERE t.series_id IS DISTINCT FROM observe_test.series_id;
DELETE FROM test_2_error re
USING t
WHERE t.test_2_id = re.test_2_id;
INSERT INTO test_2_error (test_2_id, error, record_num)
SELECT DISTINCT test_2_id, error, record_num FROM tmp_test_2_error ORDER BY error;
UPDATE raw_table AS rd1
SET processed = case WHEN tre.raw_table_id IS null THEN 2 ELSE 1 END
FROM test_1 tr
LEFT JOIN
tmp_test_2_error tre ON tre.raw_table_id = tr.id
WHERE rd1.id = tr.id;
OPEN ref FOR
SELECT 1;
RETURN ref;
OPEN ref for
SELECT o.* from observe_test o
;
RETURN ref;
OPEN ref FOR
SELECT
rd.id,
ds.id AS series_id,
now() AS timestamp,
rd.time
FROM test_2 r
INNER JOIN raw_table rd ON rd.test_2_id = r.id
INNER JOIN data_origin ON data_origin.code = rd.code
WHERE o.id IS NULL AND r.id = _id;
RETURN ref;
END;
$$ LANGUAGE plpgsql VOLATILE COST 100;
I am not able to run this procedure.
Can you please help me where I have done wrong?
I am using squirrel and face the same question as you.
until I found that:
-- Note that if you want to create the function under Squirrel SQL,
-- you must go to Sessions->Session Properties
-- then SQL tab and change the Statement Separator from ';' to something else
-- (for intance //). Otherwise Squirrel SQL sends one piece to the server
-- that stops at the first encountered ';', and the server cannot make
-- sense of it. With the separator changed as suggested, you type everything
-- as above and end with
-- ...
-- end;
-- $$ language plpgsql
-- //
--
-- You can then restore the default separator, or use the new one for
-- all queries ...
--

PostgreSQL: Pass value from an array and do union

WITH temp AS (select * from t1 where c1 = 'string1')
select 'string1' as col1, t2.col2, temp.col3 from t2 inner join temp on t2.c2 = temp.c2 where t2.some_col like 'string1%'
union
WITH temp AS (select * from t1 where c1 = 'string2')
select 'string2' as col1, t2.col2, temp.col3 from t2 inner join temp on t2.c2 = temp.c2 where t2.some_col like 'string2%'
...
Above is just an example of a PostgreSQL query I am trying to run. It's a union of two completely similar queries. They only use different values for matching string1 and string2.
I have about 20 such queries that I want to do a union on. They only differ by the variable I want to use for comparison such as string1
How can I use such array of values ['string1', 'string2', 'string3', .., 'string20'], run a query on each variable from this array and union them?
What about a old fashioned plpgsql?
CREATE OR REPLACE FUNCTION get_all_foo(arr varchar[]) RETURNS TABLE (col1 TEXT, col2 TEXT, col3 TEXT) AS
$BODY$
DECLARE
arr_value varchar;
generated_query varchar := '';
array_index smallint := 1;
array_length smallint;
BEGIN
array_length := array_length(arr, 1);
FOREACH arr_value IN ARRAY arr LOOP
generated_query := generated_query || format(' (WITH temp AS (select * from t1 where c1 = %L) '
'select %L as col1, t2.col2, temp.col3 from t2 inner join temp on t2.c2 = temp.c2 where t2.some_col like ''%s%%'')', arr_value, arr_value, arr_value);
IF array_index < array_length THEN
generated_query := generated_query || ' UNION ';
END IF;
array_index := array_index+1;
END LOOP;
RAISE DEBUG 'Generated query: %', generated_query;
RETURN QUERY EXECUTE generated_query;
END
$BODY$
LANGUAGE plpgsql;
--Uncomment to see generated query
--SET client_min_messages = DEBUG;
SELECT * FROM get_all_foo(array['string1', 'string2', 'string3', 'string4', 'string5']);
select c1 as col1, t2.col2, temp.col3
from
(select col2, c2 from t2 where
some_col like 'string1%' or some_col like 'string2%' or <other strings in the similar fashion>) t2
inner join
(select c1,c2,col3 from t1 where c1 in ('string1', 'string2', <other strings in the similar fashion>)) temp
on t2.c2 = temp.c2;
WITH temp AS (
select *
from t1
where c1 = any(array['string1','string2','string3']))
select distinct
temp.c1 as col1, t2.col2, temp.col3
from t2 inner join
temp on (t2.c2 = temp.c2 and t2.some_col like temp.c1||'%')

How to generate the "create table" sql statement for an existing table in postgreSQL

I have created a table in postgreSQL. I want to look at the SQL statement used to create the table but cannot figure it out.
How do I get the create table SQL statement for an existing table in Postgres via commandline or SQL statement?
pg_dump -t 'schema-name.table-name' --schema-only database-name
More info - in the manual.
(NOTICE - this solution is not working with PostgreSQL v12+)
My solution is to log in to the postgres db using psql with the -E option as follows:
psql -E -U username -d database
In psql, run the following commands to see the sql that postgres uses to generate
the describe table statement:
-- List all tables in the schema (my example schema name is public)
\dt public.*
-- Choose a table name from above
-- For create table of one public.tablename
\d+ public.tablename
Based on the sql echoed out after running these describe commands, I was able to put together
the following plpgsql function:
CREATE OR REPLACE FUNCTION generate_create_table_statement(p_table_name varchar)
RETURNS text AS
$BODY$
DECLARE
v_table_ddl text;
column_record record;
BEGIN
FOR column_record IN
SELECT
b.nspname as schema_name,
b.relname as table_name,
a.attname as column_name,
pg_catalog.format_type(a.atttypid, a.atttypmod) as column_type,
CASE WHEN
(SELECT substring(pg_catalog.pg_get_expr(d.adbin, d.adrelid) for 128)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = a.attrelid AND d.adnum = a.attnum AND a.atthasdef) IS NOT NULL THEN
'DEFAULT '|| (SELECT substring(pg_catalog.pg_get_expr(d.adbin, d.adrelid) for 128)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = a.attrelid AND d.adnum = a.attnum AND a.atthasdef)
ELSE
''
END as column_default_value,
CASE WHEN a.attnotnull = true THEN
'NOT NULL'
ELSE
'NULL'
END as column_not_null,
a.attnum as attnum,
e.max_attnum as max_attnum
FROM
pg_catalog.pg_attribute a
INNER JOIN
(SELECT c.oid,
n.nspname,
c.relname
FROM pg_catalog.pg_class c
LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE c.relname ~ ('^('||p_table_name||')$')
AND pg_catalog.pg_table_is_visible(c.oid)
ORDER BY 2, 3) b
ON a.attrelid = b.oid
INNER JOIN
(SELECT
a.attrelid,
max(a.attnum) as max_attnum
FROM pg_catalog.pg_attribute a
WHERE a.attnum > 0
AND NOT a.attisdropped
GROUP BY a.attrelid) e
ON a.attrelid=e.attrelid
WHERE a.attnum > 0
AND NOT a.attisdropped
ORDER BY a.attnum
LOOP
IF column_record.attnum = 1 THEN
v_table_ddl:='CREATE TABLE '||column_record.schema_name||'.'||column_record.table_name||' (';
ELSE
v_table_ddl:=v_table_ddl||',';
END IF;
IF column_record.attnum <= column_record.max_attnum THEN
v_table_ddl:=v_table_ddl||chr(10)||
' '||column_record.column_name||' '||column_record.column_type||' '||column_record.column_default_value||' '||column_record.column_not_null;
END IF;
END LOOP;
v_table_ddl:=v_table_ddl||');';
RETURN v_table_ddl;
END;
$BODY$
LANGUAGE 'plpgsql' COST 100.0 SECURITY INVOKER;
Here is the function usage:
SELECT generate_create_table_statement('tablename');
And here is the drop statement if you don't want this function to persist permanently:
DROP FUNCTION generate_create_table_statement(p_table_name varchar);
Generate the create table statement for a table in postgresql from linux commandline:
Create a table for a demo:
CREATE TABLE your_table(
thekey integer NOT NULL,
ticker character varying(10) NOT NULL,
date_val date,
open_val numeric(10,4) NOT NULL
);
pg_dump manual, can output the table create psql statement:
pg_dump -U your_user your_database -t your_table --schema-only
Which prints:
-- pre-requisite database and table configuration omitted
CREATE TABLE your_table (
thekey integer NOT NULL,
ticker character varying(10) NOT NULL,
date_val date,
open_val numeric(10,4) NOT NULL
);
-- post-requisite database and table configuration omitted
Explanation:
pg_dump helps us get information about the database itself. -U stands for username. My pgadmin user has no password set, so I don't have to put in a password. The -t option means specify for one table. --schema-only means print only data about the table, and not the data in the table.
pg_dump is elite C code that tries to play nicely with the evolving sql standards, and takes care of the thousand details that arise between postgresql's query language, and its representation on a disk. If you want to roll your own "psql disk to create statement" arrangement, ye be dragons: https://doxygen.postgresql.org/pg__dump_8c_source.html
Another option to get around pg_dump is to save the table-create SQL statement when you create the table. Keep it somewhere safe and retrieve it when you need it.
Or get the table name, column name and datatype information from postgresql with SQL:
CREATE TABLE your_table( thekey integer NOT NULL,
ticker character varying(10) NOT NULL,
date_val date,
open_val numeric(10,4) NOT NULL
);
SELECT table_name, column_name, data_type
FROM information_schema.columns
WHERE table_name = 'your_table';
Which prints:
┌────────────┬─────────────┬───────────────────┐
│ table_name │ column_name │ data_type │
├────────────┼─────────────┼───────────────────┤
│ your_table │ thekey │ integer │
│ your_table │ ticker │ character varying │
│ your_table │ date_val │ date │
│ your_table │ open_val │ numeric │
└────────────┴─────────────┴───────────────────┘
If you want to find the create statement for a table without using pg_dump, This query might work for you (change 'tablename' with whatever your table is called):
SELECT
'CREATE TABLE ' || relname || E'\n(\n' ||
array_to_string(
array_agg(
' ' || column_name || ' ' || type || ' '|| not_null
)
, E',\n'
) || E'\n);\n'
from
(
SELECT
c.relname, a.attname AS column_name,
pg_catalog.format_type(a.atttypid, a.atttypmod) as type,
case
when a.attnotnull
then 'NOT NULL'
else 'NULL'
END as not_null
FROM pg_class c,
pg_attribute a,
pg_type t
WHERE c.relname = 'tablename'
AND a.attnum > 0
AND a.attrelid = c.oid
AND a.atttypid = t.oid
ORDER BY a.attnum
) as tabledefinition
group by relname;
when called directly from psql, it is usefult to do:
\pset linestyle old-ascii
Also, the function generate_create_table_statement in this thread works very well.
Dean Toader Just excellent!
I'd modify your code a little, to show all constraints in the table and to make possible to use regexp mask in table name.
CREATE OR REPLACE FUNCTION public.generate_create_table_statement(p_table_name character varying)
RETURNS SETOF text AS
$BODY$
DECLARE
v_table_ddl text;
column_record record;
table_rec record;
constraint_rec record;
firstrec boolean;
BEGIN
FOR table_rec IN
SELECT c.relname FROM pg_catalog.pg_class c
LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE relkind = 'r'
AND relname~ ('^('||p_table_name||')$')
AND n.nspname <> 'pg_catalog'
AND n.nspname <> 'information_schema'
AND n.nspname !~ '^pg_toast'
AND pg_catalog.pg_table_is_visible(c.oid)
ORDER BY c.relname
LOOP
FOR column_record IN
SELECT
b.nspname as schema_name,
b.relname as table_name,
a.attname as column_name,
pg_catalog.format_type(a.atttypid, a.atttypmod) as column_type,
CASE WHEN
(SELECT substring(pg_catalog.pg_get_expr(d.adbin, d.adrelid) for 128)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = a.attrelid AND d.adnum = a.attnum AND a.atthasdef) IS NOT NULL THEN
'DEFAULT '|| (SELECT substring(pg_catalog.pg_get_expr(d.adbin, d.adrelid) for 128)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = a.attrelid AND d.adnum = a.attnum AND a.atthasdef)
ELSE
''
END as column_default_value,
CASE WHEN a.attnotnull = true THEN
'NOT NULL'
ELSE
'NULL'
END as column_not_null,
a.attnum as attnum,
e.max_attnum as max_attnum
FROM
pg_catalog.pg_attribute a
INNER JOIN
(SELECT c.oid,
n.nspname,
c.relname
FROM pg_catalog.pg_class c
LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE c.relname = table_rec.relname
AND pg_catalog.pg_table_is_visible(c.oid)
ORDER BY 2, 3) b
ON a.attrelid = b.oid
INNER JOIN
(SELECT
a.attrelid,
max(a.attnum) as max_attnum
FROM pg_catalog.pg_attribute a
WHERE a.attnum > 0
AND NOT a.attisdropped
GROUP BY a.attrelid) e
ON a.attrelid=e.attrelid
WHERE a.attnum > 0
AND NOT a.attisdropped
ORDER BY a.attnum
LOOP
IF column_record.attnum = 1 THEN
v_table_ddl:='CREATE TABLE '||column_record.schema_name||'.'||column_record.table_name||' (';
ELSE
v_table_ddl:=v_table_ddl||',';
END IF;
IF column_record.attnum <= column_record.max_attnum THEN
v_table_ddl:=v_table_ddl||chr(10)||
' '||column_record.column_name||' '||column_record.column_type||' '||column_record.column_default_value||' '||column_record.column_not_null;
END IF;
END LOOP;
firstrec := TRUE;
FOR constraint_rec IN
SELECT conname, pg_get_constraintdef(c.oid) as constrainddef
FROM pg_constraint c
WHERE conrelid=(
SELECT attrelid FROM pg_attribute
WHERE attrelid = (
SELECT oid FROM pg_class WHERE relname = table_rec.relname
) AND attname='tableoid'
)
LOOP
v_table_ddl:=v_table_ddl||','||chr(10);
v_table_ddl:=v_table_ddl||'CONSTRAINT '||constraint_rec.conname;
v_table_ddl:=v_table_ddl||chr(10)||' '||constraint_rec.constrainddef;
firstrec := FALSE;
END LOOP;
v_table_ddl:=v_table_ddl||');';
RETURN NEXT v_table_ddl;
END LOOP;
END;
$BODY$
LANGUAGE plpgsql VOLATILE
COST 100;
ALTER FUNCTION public.generate_create_table_statement(character varying)
OWNER TO postgres;
Now you can, for example, make the following query
SELECT * FROM generate_create_table_statement('.*');
which results like this:
CREATE TABLE public.answer (
id integer DEFAULT nextval('answer_id_seq'::regclass) NOT NULL,
questionid integer NOT NULL,
title character varying NOT NULL,
defaultvalue character varying NULL,
valuetype integer NOT NULL,
isdefault boolean NULL,
minval double precision NULL,
maxval double precision NULL,
followminmax integer DEFAULT 0 NOT NULL,
CONSTRAINT answer_pkey
PRIMARY KEY (id),
CONSTRAINT answer_questionid_fkey
FOREIGN KEY (questionid) REFERENCES question(id) ON UPDATE RESTRICT ON DELETE RESTRICT,
CONSTRAINT answer_valuetype_fkey
FOREIGN KEY (valuetype) REFERENCES answervaluetype(id) ON UPDATE RESTRICT ON DELETE RESTRICT);
for each user table.
The easiest method I can think of is to install pgAdmin 3 (found here) and use it to view your database. It will automatically generate a query that will create the table in question.
If you want to do this for various tables at once, you meed to use the -t switch multiple times (took me a while to figure out why comma separated list wasn't working). Also, can be useful to send results to an outfile or pipe to a postgres server on another machine
pg_dump -t table1 -t table2 database_name --schema-only > dump.sql
pg_dump -t table1 -t table2 database_name --schema-only | psql -h server_name database_name
Here is another solution to the old question. There have been many excellent answers to this question over the years and my attempt borrows heavily from them.
I used Andrey Lebedenko's solution as a starting point because its output was already very close to my requirements.
Features:
following common practice I have moved the foreign key constraints outside the table definition. They are now included as ALTER TABLE statements at the bottom. The reason is that a foreign key can also link to a column of the same table. In that fringe case the constraint can only be created after the table creation is completed. The create table statement would throw an error otherwise.
The layout and indenting looks nicer now (at least to my eye)
Drop command (commented out) in the header of the definition
The solution is offered here as a plpgsql function. The algorithm does however not use any procedural language. The function just wraps one single query that can be used in a pure sql context as well.
removed redundant subqueries
Identifiers are now quoted if they are identical to reserved postgresql language elements
replaced the string concatenation operator || with the appropriate string functions to improve performance, security and readability of the code.
Note: the || operator produces NULL if one of the combined strings is NULL. It should only be used when that is the desired behaviour. (check out the
usage in the code below for an example)
CREATE OR REPLACE FUNCTION public.wmv_get_table_definition (
p_schema_name character varying,
p_table_name character varying
)
RETURNS SETOF TEXT
AS $BODY$
BEGIN
RETURN query
WITH table_rec AS (
SELECT
c.relname, n.nspname, c.oid
FROM
pg_catalog.pg_class c
LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE
relkind = 'r'
AND n.nspname = p_schema_name
AND c.relname LIKE p_table_name
ORDER BY
c.relname
),
col_rec AS (
SELECT
a.attname AS colname,
pg_catalog.format_type(a.atttypid, a.atttypmod) AS coltype,
a.attrelid AS oid,
' DEFAULT ' || (
SELECT
pg_catalog.pg_get_expr(d.adbin, d.adrelid)
FROM
pg_catalog.pg_attrdef d
WHERE
d.adrelid = a.attrelid
AND d.adnum = a.attnum
AND a.atthasdef) AS column_default_value,
CASE WHEN a.attnotnull = TRUE THEN
'NOT NULL'
ELSE
'NULL'
END AS column_not_null,
a.attnum AS attnum
FROM
pg_catalog.pg_attribute a
WHERE
a.attnum > 0
AND NOT a.attisdropped
ORDER BY
a.attnum
),
con_rec AS (
SELECT
conrelid::regclass::text AS relname,
n.nspname,
conname,
pg_get_constraintdef(c.oid) AS condef,
contype,
conrelid AS oid
FROM
pg_constraint c
JOIN pg_namespace n ON n.oid = c.connamespace
),
glue AS (
SELECT
format( E'-- %1$I.%2$I definition\n\n-- Drop table\n\n-- DROP TABLE IF EXISTS %1$I.%2$I\n\nCREATE TABLE %1$I.%2$I (\n', table_rec.nspname, table_rec.relname) AS top,
format( E'\n);\n\n\n-- adempiere.wmv_ghgaudit foreign keys\n\n', table_rec.nspname, table_rec.relname) AS bottom,
oid
FROM
table_rec
),
cols AS (
SELECT
string_agg(format(' %I %s%s %s', colname, coltype, column_default_value, column_not_null), E',\n') AS lines,
oid
FROM
col_rec
GROUP BY
oid
),
constrnt AS (
SELECT
string_agg(format(' CONSTRAINT %s %s', con_rec.conname, con_rec.condef), E',\n') AS lines,
oid
FROM
con_rec
WHERE
contype <> 'f'
GROUP BY
oid
),
frnkey AS (
SELECT
string_agg(format('ALTER TABLE %I.%I ADD CONSTRAINT %s %s', nspname, relname, conname, condef), E';\n') AS lines,
oid
FROM
con_rec
WHERE
contype = 'f'
GROUP BY
oid
)
SELECT
concat(glue.top, cols.lines, E',\n', constrnt.lines, glue.bottom, frnkey.lines, ';')
FROM
glue
JOIN cols ON cols.oid = glue.oid
LEFT JOIN constrnt ON constrnt.oid = glue.oid
LEFT JOIN frnkey ON frnkey.oid = glue.oid;
END;
$BODY$
LANGUAGE plpgsql;
Even more modification based on response from #vkkeeper. Added possibility to query table from the specific schema.
CREATE OR REPLACE FUNCTION public.describe_table(p_schema_name character varying, p_table_name character varying)
RETURNS SETOF text AS
$BODY$
DECLARE
v_table_ddl text;
column_record record;
table_rec record;
constraint_rec record;
firstrec boolean;
BEGIN
FOR table_rec IN
SELECT c.relname, c.oid FROM pg_catalog.pg_class c
LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE relkind = 'r'
AND n.nspname = p_schema_name
AND relname~ ('^('||p_table_name||')$')
ORDER BY c.relname
LOOP
FOR column_record IN
SELECT
b.nspname as schema_name,
b.relname as table_name,
a.attname as column_name,
pg_catalog.format_type(a.atttypid, a.atttypmod) as column_type,
CASE WHEN
(SELECT substring(pg_catalog.pg_get_expr(d.adbin, d.adrelid) for 128)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = a.attrelid AND d.adnum = a.attnum AND a.atthasdef) IS NOT NULL THEN
'DEFAULT '|| (SELECT substring(pg_catalog.pg_get_expr(d.adbin, d.adrelid) for 128)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = a.attrelid AND d.adnum = a.attnum AND a.atthasdef)
ELSE
''
END as column_default_value,
CASE WHEN a.attnotnull = true THEN
'NOT NULL'
ELSE
'NULL'
END as column_not_null,
a.attnum as attnum,
e.max_attnum as max_attnum
FROM
pg_catalog.pg_attribute a
INNER JOIN
(SELECT c.oid,
n.nspname,
c.relname
FROM pg_catalog.pg_class c
LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE c.oid = table_rec.oid
ORDER BY 2, 3) b
ON a.attrelid = b.oid
INNER JOIN
(SELECT
a.attrelid,
max(a.attnum) as max_attnum
FROM pg_catalog.pg_attribute a
WHERE a.attnum > 0
AND NOT a.attisdropped
GROUP BY a.attrelid) e
ON a.attrelid=e.attrelid
WHERE a.attnum > 0
AND NOT a.attisdropped
ORDER BY a.attnum
LOOP
IF column_record.attnum = 1 THEN
v_table_ddl:='CREATE TABLE '||column_record.schema_name||'.'||column_record.table_name||' (';
ELSE
v_table_ddl:=v_table_ddl||',';
END IF;
IF column_record.attnum <= column_record.max_attnum THEN
v_table_ddl:=v_table_ddl||chr(10)||
' '||column_record.column_name||' '||column_record.column_type||' '||column_record.column_default_value||' '||column_record.column_not_null;
END IF;
END LOOP;
firstrec := TRUE;
FOR constraint_rec IN
SELECT conname, pg_get_constraintdef(c.oid) as constrainddef
FROM pg_constraint c
WHERE conrelid=(
SELECT attrelid FROM pg_attribute
WHERE attrelid = (
SELECT oid FROM pg_class WHERE relname = table_rec.relname
AND relnamespace = (SELECT ns.oid FROM pg_namespace ns WHERE ns.nspname = p_schema_name)
) AND attname='tableoid'
)
LOOP
v_table_ddl:=v_table_ddl||','||chr(10);
v_table_ddl:=v_table_ddl||'CONSTRAINT '||constraint_rec.conname;
v_table_ddl:=v_table_ddl||chr(10)||' '||constraint_rec.constrainddef;
firstrec := FALSE;
END LOOP;
v_table_ddl:=v_table_ddl||');';
RETURN NEXT v_table_ddl;
END LOOP;
END;
$BODY$
LANGUAGE plpgsql VOLATILE
COST 100;
Here is a single statement that will generate the DDL for a single table in a specified schema, including constraints.
SELECT 'CREATE TABLE ' || pn.nspname || '.' || pc.relname || E'(\n' ||
string_agg(pa.attname || ' ' || pg_catalog.format_type(pa.atttypid, pa.atttypmod) || coalesce(' DEFAULT ' || (
SELECT pg_catalog.pg_get_expr(d.adbin, d.adrelid)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = pa.attrelid
AND d.adnum = pa.attnum
AND pa.atthasdef
),
'') || ' ' ||
CASE pa.attnotnull
WHEN TRUE THEN 'NOT NULL'
ELSE 'NULL'
END, E',\n') ||
coalesce((SELECT E',\n' || string_agg('CONSTRAINT ' || pc1.conname || ' ' || pg_get_constraintdef(pc1.oid), E',\n' ORDER BY pc1.conindid)
FROM pg_constraint pc1
WHERE pc1.conrelid = pa.attrelid), '') ||
E');'
FROM pg_catalog.pg_attribute pa
JOIN pg_catalog.pg_class pc
ON pc.oid = pa.attrelid
AND pc.relname = 'table_name'
JOIN pg_catalog.pg_namespace pn
ON pn.oid = pc.relnamespace
AND pn.nspname = 'schema_name'
WHERE pa.attnum > 0
AND NOT pa.attisdropped
GROUP BY pn.nspname, pc.relname, pa.attrelid;
If you have PgAdmin4, then open it. Go to your database--> schema---> table--> right click on table name whose create script you want---> Scripts---> CREATE SCRIPT
Here is a bit improved version of shekwi's query.
It generates the primary key constraint and is able to handle temporary tables:
with pkey as
(
select cc.conrelid, format(E',
constraint %I primary key(%s)', cc.conname,
string_agg(a.attname, ', '
order by array_position(cc.conkey, a.attnum))) pkey
from pg_catalog.pg_constraint cc
join pg_catalog.pg_class c on c.oid = cc.conrelid
join pg_catalog.pg_attribute a on a.attrelid = cc.conrelid
and a.attnum = any(cc.conkey)
where cc.contype = 'p'
group by cc.conrelid, cc.conname
)
select format(E'create %stable %s%I\n(\n%s%s\n);\n',
case c.relpersistence when 't' then 'temporary ' else '' end,
case c.relpersistence when 't' then '' else n.nspname || '.' end,
c.relname,
string_agg(
format(E'\t%I %s%s',
a.attname,
pg_catalog.format_type(a.atttypid, a.atttypmod),
case when a.attnotnull then ' not null' else '' end
), E',\n'
order by a.attnum
),
(select pkey from pkey where pkey.conrelid = c.oid)) as sql
from pg_catalog.pg_class c
join pg_catalog.pg_namespace n on n.oid = c.relnamespace
join pg_catalog.pg_attribute a on a.attrelid = c.oid and a.attnum > 0
join pg_catalog.pg_type t on a.atttypid = t.oid
where c.relname = :table_name
group by c.oid, c.relname, c.relpersistence, n.nspname;
Use table_name parameter to specify the name of the table.
This is the variation that works for me:
pg_dump -U user_viktor -h localhost unit_test_database -t floorplanpreferences_table --schema-only
In addition, if you're using schemas, you'll of course need to specify that as well:
pg_dump -U user_viktor -h localhost unit_test_database -t "949766e0-e81e-11e3-b325-1cc1de32fcb6".floorplanpreferences_table --schema-only
You will get an output that you can use to create the table again, just run that output in psql.
pg_dump -h XXXXXXXXXXX.us-west-1.rds.amazonaws.com -U anyuser -t tablename -s
Like the other answers mentioned, there is no built in function that does this.
Here is a function that attempts to get all of the information that would be needed to replicate the table - or to compare deployed and checked in ddl.
This function outputs:
columns (w/ precision, null/not-null, default value)
constraints
indexes
CREATE OR REPLACE FUNCTION public.show_create_table(
in_schema_name varchar,
in_table_name varchar
)
RETURNS text
LANGUAGE plpgsql VOLATILE
AS
$$
DECLARE
-- the ddl we're building
v_table_ddl text;
-- data about the target table
v_table_oid int;
-- records for looping
v_column_record record;
v_constraint_record record;
v_index_record record;
BEGIN
-- grab the oid of the table; https://www.postgresql.org/docs/8.3/catalog-pg-class.html
SELECT c.oid INTO v_table_oid
FROM pg_catalog.pg_class c
LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace
WHERE 1=1
AND c.relkind = 'r' -- r = ordinary table; https://www.postgresql.org/docs/9.3/catalog-pg-class.html
AND c.relname = in_table_name -- the table name
AND n.nspname = in_schema_name; -- the schema
-- throw an error if table was not found
IF (v_table_oid IS NULL) THEN
RAISE EXCEPTION 'table does not exist';
END IF;
-- start the create definition
v_table_ddl := 'CREATE TABLE ' || in_schema_name || '.' || in_table_name || ' (' || E'\n';
-- define all of the columns in the table; https://stackoverflow.com/a/8153081/3068233
FOR v_column_record IN
SELECT
c.column_name,
c.data_type,
c.character_maximum_length,
c.is_nullable,
c.column_default
FROM information_schema.columns c
WHERE (table_schema, table_name) = (in_schema_name, in_table_name)
ORDER BY ordinal_position
LOOP
v_table_ddl := v_table_ddl || ' ' -- note: two char spacer to start, to indent the column
|| v_column_record.column_name || ' '
|| v_column_record.data_type || CASE WHEN v_column_record.character_maximum_length IS NOT NULL THEN ('(' || v_column_record.character_maximum_length || ')') ELSE '' END || ' '
|| CASE WHEN v_column_record.is_nullable = 'NO' THEN 'NOT NULL' ELSE 'NULL' END
|| CASE WHEN v_column_record.column_default IS NOT null THEN (' DEFAULT ' || v_column_record.column_default) ELSE '' END
|| ',' || E'\n';
END LOOP;
-- define all the constraints in the; https://www.postgresql.org/docs/9.1/catalog-pg-constraint.html && https://dba.stackexchange.com/a/214877/75296
FOR v_constraint_record IN
SELECT
con.conname as constraint_name,
con.contype as constraint_type,
CASE
WHEN con.contype = 'p' THEN 1 -- primary key constraint
WHEN con.contype = 'u' THEN 2 -- unique constraint
WHEN con.contype = 'f' THEN 3 -- foreign key constraint
WHEN con.contype = 'c' THEN 4
ELSE 5
END as type_rank,
pg_get_constraintdef(con.oid) as constraint_definition
FROM pg_catalog.pg_constraint con
JOIN pg_catalog.pg_class rel ON rel.oid = con.conrelid
JOIN pg_catalog.pg_namespace nsp ON nsp.oid = connamespace
WHERE nsp.nspname = in_schema_name
AND rel.relname = in_table_name
ORDER BY type_rank
LOOP
v_table_ddl := v_table_ddl || ' ' -- note: two char spacer to start, to indent the column
|| 'CONSTRAINT' || ' '
|| v_constraint_record.constraint_name || ' '
|| v_constraint_record.constraint_definition
|| ',' || E'\n';
END LOOP;
-- drop the last comma before ending the create statement
v_table_ddl = substr(v_table_ddl, 0, length(v_table_ddl) - 1) || E'\n';
-- end the create definition
v_table_ddl := v_table_ddl || ');' || E'\n';
-- suffix create statement with all of the indexes on the table
FOR v_index_record IN
SELECT indexdef
FROM pg_indexes
WHERE (schemaname, tablename) = (in_schema_name, in_table_name)
LOOP
v_table_ddl := v_table_ddl
|| v_index_record.indexdef
|| ';' || E'\n';
END LOOP;
-- return the ddl
RETURN v_table_ddl;
END;
$$;
example
SELECT * FROM public.show_create_table('public', 'example_table');
produces
CREATE TABLE public.example_table (
id bigint NOT NULL DEFAULT nextval('test_tb_for_show_create_on_id_seq'::regclass),
name character varying(150) NULL,
level character varying(50) NULL,
description text NOT NULL DEFAULT 'hello there!'::text,
CONSTRAINT test_tb_for_show_create_on_pkey PRIMARY KEY (id),
CONSTRAINT test_tb_for_show_create_on_level_check CHECK (((level)::text = ANY ((ARRAY['info'::character varying, 'warn'::character varying, 'error'::character varying])::text[])))
);
CREATE UNIQUE INDEX test_tb_for_show_create_on_pkey ON public.test_tb_for_show_create_on USING btree (id);
YOu can also use a free DB management tool, such as DBeaver, which allows you to view DDL for the tables, here's an example:
DataGrip has the same functionality as pgAdmin. You can right click on a table and you will see option to auto-generate create table statement.
Use this and get your output in ddl.out file
~/bin/pg_dump -p 30000 -d <db_name> -U <db_user> --schema=<schema_name> -t <table_name> --schema-only >> /tmp/ddl.out
So this will generate DDL in the path: /tmp/ddl.out
Here is a solution if you don't want to create a function and just want the query to create a basic table structure.
select 'CREATE TABLE ' || table_name ||'(' ||STRING_AGG (
column_name || ' ' || data_type ,
','
ORDER BY
table_name,
ordinal_position
) ||');'
from
information_schema.columns
where table_schema = 'public'
group by
table_name
A simple solution, in pure single SQL.
You get the idea, you may extend it to more attributes you like to show.
with c as (
SELECT table_name, ordinal_position,
column_name|| ' ' || data_type col
, row_number() over (partition by table_name order by ordinal_position asc) rn
, count(*) over (partition by table_name) cnt
FROM information_schema.columns
WHERE table_name in ('pg_index', 'pg_tables')
order by table_name, ordinal_position
)
select case when rn = 1 then 'create table ' || table_name || '(' else '' end
|| col
|| case when rn < cnt then ',' else '); ' end
from c
order by table_name, rn asc;
Output:
create table pg_index(indexrelid oid,
indrelid oid,
indnatts smallint,
indisunique boolean,
indisprimary boolean,
indisexclusion boolean,
indimmediate boolean,
indisclustered boolean,
indisvalid boolean,
indcheckxmin boolean,
indisready boolean,
indislive boolean,
indisreplident boolean,
indkey ARRAY,
indcollation ARRAY,
indclass ARRAY,
indoption ARRAY,
indexprs pg_node_tree,
indpred pg_node_tree);
create table pg_tables(schemaname name,
tablename name,
tableowner name,
tablespace name,
hasindexes boolean,
hasrules boolean,
hastriggers boolean,
rowsecurity boolean);
Another easy option was to use [HeidiSQL client][1] for PostgreSQL database.
How to go into the database tab where all the databases and tables are listed.
Click on any of the table/View which you wanted to see the DDL/create a statement of the particular table.
Now there this client do the following jobs for you for that table, on the right-hand side windows:
The first window would be for data of table
Second for your SQL Host information
Third for database-level information like which tables and what is the size
Forth which we are more concern about table/view information tab will have the create table statement readily available for you.
I can not show you in the snapshot as working with confidential data, Try it with yourself and let me know if any issues you guys found.
In pgadminIII database>>schemas>>tables>> right click on 'Your table'>>scripts>> 'Select any one (Create,Insert,Update,Delete..)'
Here is a query with some edits,
select 'CREATE TABLE ' || a.attrelid::regclass::text || '(' ||
string_agg(a.attname || ' ' || pg_catalog.format_type(a.atttypid,
a.atttypmod)||
CASE WHEN
(SELECT substring(pg_catalog.pg_get_expr(d.adbin, d.adrelid) for 128)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = a.attrelid AND d.adnum = a.attnum AND a.atthasdef) IS NOT NULL THEN
' DEFAULT '|| (SELECT substring(pg_catalog.pg_get_expr(d.adbin, d.adrelid) for 128)
FROM pg_catalog.pg_attrdef d
WHERE d.adrelid = a.attrelid AND d.adnum = a.attnum AND a.atthasdef)
ELSE
'' END
||
CASE WHEN a.attnotnull = true THEN
' NOT NULL'
ELSE
'' END,E'\n,') || ');'
FROM pg_catalog.pg_attribute a join pg_class on a.attrelid=pg_class.oid
WHERE a.attrelid::regclass::varchar =
'TABLENAME_with_or_without_schema'
AND a.attnum > 0 AND NOT a.attisdropped and pg_class.relkind='r'
group by a.attrelid;
To generate the SQL (DDL) behind the creation of a particular table.
We can simply use this SQL query -
SHOW TABLE your_schema_name.your_table_name