143

My code is:

SELECT column_name
FROM information.SCHEMA.columns
WHERE table_name = 'aean'

It returns column names of table aean.
Now I have declared an array:

DECLARE colnames text[]

How can I store select's output in colnames array.
Is there any need to initialize colnames?

2

6 Answers 6

234

There are two ways. One is to aggregate:

SELECT array_agg(column_name::TEXT)
FROM information.schema.columns
WHERE table_name = 'aean'

The other is to use an array constructor:

SELECT ARRAY(
    SELECT column_name 
    FROM information_schema.columns 
    WHERE table_name = 'aean'
)

I'm presuming this is for plpgsql. In that case you can assign it like this:

colnames := ARRAY(
    SELECT column_name
    FROM information_schema.columns
    WHERE table_name='aean'
);
4
  • can you give me the code that works on postgres because this is not working on postgres ERROR: could not find array type for data type information_schema.sql_identifier
    – mitesh
    Commented Jun 19, 2011 at 12:15
  • Sorry 'bout that. Mindless copying and pasting got array_agg() included in all three calls. I've also type-casted the initial one to make your PG version happy. Commented Jun 19, 2011 at 13:41
  • 15
    For anyone looking further on the plpgsql part you can DECLARE an array as my_array INTEGER[]; (or whatever the relevant type is). You can also use the array in a query's WHERE clause like WHERE values = ANY(my_array). The ANY takes an array or a set and will check for presence in that array/set, so it functions equivalently to IN () in a sense Commented Jun 12, 2018 at 19:56
  • The examples above should be information_schema, not information.schema. Commented Sep 9, 2021 at 23:14
17

I had exactly the same problem. Just one more working modification of the solution given by Denis (the type must be specified):

SELECT ARRAY(
SELECT column_name::text
FROM information_schema.columns
WHERE table_name='aean'
)
1
  • 2
    This doesn't appear to be the case anymore.
    – Soviut
    Commented Apr 9, 2019 at 1:01
7

Regular:

SELECT post_id FROM posts WHERE(poster_name='John');

output: [
  {'post_id': 1},
  {'post_id': 2},
  {'post_id': 3},
]

Using ARRAY_AGG:

SELECT ARRAY_AGG(post_id) FROM posts WHERE(poster_name='John');

output: [
 {[1, 2, 3]}
]
0
2

Casting to the datatype "TEXT" will ensure that your queries will run without any problem. In plpgsql when we assign to a array variable, we need not use the type casting. My requirement was to get a CSV of all the column names of a particular table. I'd used the following code in plpgsql.

Declare col_list varchar[]:=NULL;
cols varchar:=NULL;
Begin
    col_list := ARRAY(select t.name from frm_columns t where t.tname='emp_mstr');
    cols := array_to_string(col_list,',');
    return cols;
End;
2
CREATE OR REPLACE FUNCTION f_test_array(in _colname text)
returns text as $body$
DECLARE colnames text[];
begin
colnames := ARRAY(
    SELECT column_name FROM information_schema.columns WHERE table_name='customer'
);
    if exists(select _colname = any(colnames))
    then return format('%s it exits.', _colname);
    else return format('%s not exits.', _colname);
end if;
end
$body$
LANGUAGE plpgsql;   

check if the column exists or not. Key point: if exists(select _colname = any(colnames)) We can also using string_agg String_agg usage:

CREATE OR REPLACE FUNCTION f_test_array1(in _colname text)
returns text as $body$
DECLARE colnames text;
begin
colnames := (SELECT string_agg(column_name,',') FROM information_schema.columns WHERE table_name='customer')::text;
if exists(select colnames ilike '%' || quote_literal(_colname) ||'%')
    then return format('column %s  exits.', _colname);
    else return format('column %s does not exits.', _colname);
end if;
end
$body$
LANGUAGE plpgsql;
0

declare colnames text[] array in function and then write following query in begin:

SELECT column_name into colnames FROM information.SCHEMA.columns WHERE table_name = 'aean'

Not the answer you're looking for? Browse other questions tagged or ask your own question.