Full outer join missing values with null

Question

I'm trying to use full outer join on postgresql database to get a union of tables that has null values for the missing rows. However, it is not working for me.

Here is the example:

create temp table nutrient_names (
    name text
);

insert into nutrient_names values
('fat'),
('sugar'),
('sodium'),
('total fat');

create temp table nutrients (
    food_id int,
    name text,
    quantity float8
);

insert into nutrients values
(1, 'fat', 0.3),
(1, 'sugar', 15),
(1, 'sodium', 10),
(1, 'total fat', 25),
(2, 'sugar', 10),
(2, 'sodium', 4);

Here is the output:

select n.name, n.food_id, n.quantity from nutrient_names nn
full outer join nutrients n
on nn.name = n.name
order by n.food_id, n.name;

+---------------------------------+
|name           |food_id |quantity|
+---------------------------------+
|    'fat'      |1       |'0.3'   |
|    'sodium'   |1       |'10'    |
|    'sugar'    |1       |'15'    |
|    'total fat'|1       |'25'    |
|    'sodium'   |2       |'4'     |
|    'sugar'    |2       |'10'    |
+---------------------------------+

What I want:

+---------------------------------+
|name           |food_id |quantity|
+---------------------------------+
|    'fat'      |1       |'0.3'   |
|    'sodium'   |1       |'10'    |
|    'sugar'    |1       |'15'    |
|    'total fat'|1       |'25'    |
|    'fat'      |2       |null    | <----
|    'sodium'   |2       |'4'     |
|    'sugar'    |2       |'10'    |
|    'total fat'|2       |null    | <----
+---------------------------------+

Answer 1

I suggest having two tables representing nutrition types and types of food. Then have an extra table specifying the nutrition values.

You can do a cross join on food and nutritions and then do a left join on the nutrition values to get your results.

I made a fiddle for you to look at here http://sqlfiddle.com/#!17/d974b/3

The reason that you cannot do a full outer join is that if you have the quantity on a row you have the id aswell. They are in the same dataset.

Answer 2

It doesn't look like FULL JOIN suits here. Based on your example you want to list all rows from nutrient_names as many times as you have food IDs. This is usually done with CROSS JOIN .

If you don't have a separate table with the list of food IDs, you can build it on the fly and then join to it:

WITH
CTE_IDs
AS
(
    SELECT DISTINCT
        food_id
    FROM nutrients
)
SELECT
    nutrient_names.name
    ,CTE_IDs.food_id
    ,nutrients.quantity
FROM
    CTE_IDs
    CROSS JOIN nutrient_names
    LEFT JOIN nutrients 
        ON  nutrients.name = nutrient_names.name
        AND nutrients.food_id = CTE_IDs.food_id
;

Answer 3

if you have 2 tables (t1 and t2) with cols: dimension_1, dimension_2, measure_1, measure_2

you can write

select dimension_1, dimension_2, sum(measure_1) as measure_1, sum(measure_2) as measure_2
from t1
full join t2 using(dimension_1, dimension_2)
group by dimension_1, dimension_2