Look for a unique value using two variables - postgresql

Question

I need to implement the ISSUES column by using REFERENCE and LOCALISATION variables for each rows with unique values stocked on Table_issues_Localisation.
The problem is, those two variables make a two dimension table, so I have to dynamically select the right column of LOCALISATION.
Here is an explanation of what I need to do with an image. I'm sorry for posting an image but I think this is way more understandable.
I tried to make a query to update row by row the Table_Observation.ISSUES column with informations stocked on variable columns (=SELECT(SELECT))) of Table_issues_Localisation.
On Table_Observation.ROW_NUMBER column indicates the number of each rows. It is used for the loop.
DO $$
DECLARE
my_variable TEXT;
BEGIN
FOR i IN 1..35 LOOP
my_variable = SELECT((SELECT LOCALISATION FROM Table_Observation WHERE Table_Observation.ROW_NUMBER = i) FROM Table__issues_Localisation ON Table_Observation.REFERENCE = Table__issues_Localisation.REFERENCE)
UPDATE Table_Observation
SET ISSUES = my_variable
WHERE Table_Observation.ROW_NUMBER = i
END LOOP;
END;
$$
Postgres v 9.4
I hope I'm clear enough,

Answer 1

You don't need PL/pgSQL or a loop for this. You can do that with a single update statement:
update observation o
set issues = row_to_json(il) ->> o.localisation
from issues_localisation il
where il.reference = o.reference;
This requires that the values in observation.localisation exactly map to the column names in issues_localisation.
With the following test data:
create table observation
(
rn integer primary key,
reference integer,
localisation text,
issues text
);
create table issues_localisation
(
reference integer,
issues_12 text,
issues_17 text,
issues_27 text,
issues_34 text
);
insert into observation (rn, reference, localisation)
values
(1, 27568, 'issues_27'),
(2, 6492, 'issues_34'),
(3, 1529, 'issues_34'),
(4, 1529, 'issues_34'),
(5, 709, 'issues_12');
insert into issues_localisation (reference, issues_12, issues_17, issues_27, issues_34)
values
(29, 'FB', 'FB', 'TFB', null),
(506, 'M', null, 'M', null),
(709, 'TF', null, null, null),
(1234, null, 'TF', 'TF', null),
(1529, 'FB', 'FB', 'FB', 'M'),
(3548, null, 'M', null, null),
(6492, 'FB', 'FB', 'FB', null),
(18210, 'TFB', null, 'TFB', 'TFB'),
(27568, 'TF', null, 'TF', 'TF');
The update will result in this data in the table observation:
rn | reference | localisation | issues
---+-----------+--------------+-------
1 | 27568 | issues_27 | TF
2 | 6492 | issues_34 |
3 | 1529 | issues_34 | M
4 | 1529 | issues_34 | M
5 | 709 | issues_12 | TF
Online example: http://rextester.com/OCGFM81609
For your next question you should supply the sample data (and the expected output) the way I did in my answer.
I removed the completely useless prefix table_ from the table names as well. That is horrible naming convention.

Answer 2

And here is an (unfinished, still need to execute) example of dynamic sql:
CREATE FUNCTION bagger (_name text) RETURNS text
AS
$func$
DECLARE upd text;
BEGIN
upd := format('
UPDATE observation dst
SET issues = src.%I
FROM issues_localisation src
WHERE src.reference = dst.reference
AND dst.localisation = %L
;', _name, _name);
-- RAISE NOTICE upd;
RETURN upd;
END
$func$
LANGUAGE plpgsql
;
-- SELECT bagger('issues_12' );
WITH loc AS (
SELECT DISTINCT localisation AS loc
FROM observation
)
SELECT bagger( loc.loc)
FROM loc
;