Add ORDINALITY to expanded JSON array in Postgres 11.7 - postgresql

Question

I'm taking two JSONB arrays, unpacking them, and combing the results. I'm trying to add WITH ORDINALITY to the JSON array unpacking. I've been unable to figure out how to add WITH ORDINALITY. For some reason, I can't find WITH ORDINALITY in the documentation for Postgres 11's JSON tools:
https://www.postgresql.org/docs/11/functions-json.html
I've seen examples using jsonb_array_elements....WITH ORDINALITY, but haven't been able to get it to work. First, a functional example based on Postgres arrays:
WITH
first AS (
SELECT * FROM
UNNEST (ARRAY['Charles','Jane','George','Percy']) WITH ORDINALITY AS x(name_, index)
),
last AS (
SELECT * FROM
UNNEST (ARRAY['Dickens','Austen','Eliot']) WITH ORDINALITY AS y(name_, index)
)
SELECT first.name_ AS first_name,
last.name_ AS last_name
FROM first
JOIN last ON (last.index = first.index)
This gives the desired output:
first_name last_name
Charles Dickens
Jane Austen
George Eliot
I'm using the ORDINALITY index to make the JOIN, as I'm combining two lists for pair-wise comparison. I can assume my lists are equally sized.
However, my input is going to be a JSON array, not a Postgres array. I've got the unpacking working with jsonb_to_recordset, but have not got the ordinality generation working. Here's a sample that does the unpacking part correctly:
DROP FUNCTION IF EXISTS tools.try_ordinality (jsonb, jsonb);
CREATE OR REPLACE FUNCTION tools.try_ordinality (
base_jsonb_in jsonb,
comparison_jsonb_in jsonb)
RETURNS TABLE (
base_text citext,
base_id citext,
comparison_text citext,
comparison_id citext)
AS $BODY$
BEGIN
RETURN QUERY
WITH
base_expanded AS (
select *
from jsonb_to_recordset (
base_jsonb_in)
AS base_unpacked (text citext, id citext)
),
comparison_expanded AS (
select *
from jsonb_to_recordset (
comparison_jsonb_in)
AS comparison_unpacked (text citext, id citext)
),
combined_lists AS (
select base_expanded.text AS base_text,
base_expanded.id AS base_id,
comparison_expanded.text AS comparison_text,
comparison_expanded.id AS comparison_id
from base_expanded,
comparison_expanded
)
select *
from combined_lists;
END
$BODY$
LANGUAGE plpgsql;
select * from try_ordinality (
'[
{"text":"Fuzzy Green Bunny","id":"1"},
{"text":"Small Gray Turtle","id":"2"}
]',
'[
{"text":"Red Large Special","id":"3"},
{"text":"Blue Small","id":"4"},
{"text":"Green Medium Special","id":"5"}
]'
);
But that's a CROSS JOIN
base_text base_id comparison_text comparison_id
Fuzzy Green Bunny 1 Red Large Special 3
Fuzzy Green Bunny 1 Blue Small 4
Fuzzy Green Bunny 1 Green Medium Special 5
Small Gray Turtle 2 Red Large Special 3
Small Gray Turtle 2 Blue Small 4
Small Gray Turtle 2 Green Medium Special 5
I'm after a pair-wise result with only two rows:
Fuzzy Green Bunny 1 Red Large Special 3
Small Gray Turtle 2 Blue Small 4
I've tried switching to jsonb_array_elements, as in this snippet:
WITH
base_expanded AS (
select *
from jsonb_array_elements (
base_jsonb_in)
AS base_unpacked (text citext, id citext)
),
I get back
ERROR: a column definition list is only allowed for functions returning "record"
Is there a straightforward way to get ordinality on an unpacked JSON array? It's very easy with UNNEST on a Postgres array.
I'm happy to learn I've screwed up the syntax.
I can CREATE TYPE, if it's of any help.
I can convert to a Postgres array, if that's straightforward to do.
Thanks for any suggestions.

Answer 1

You do it exactly the same way.
with first as (
select *
from jsonb_array_elements('[
{"text":"Fuzzy Green Bunny","id":"1"},
{"text":"Small Gray Turtle","id":"2"}
]'::jsonb) with ordinality as f(element, idx)
), last as (
select *
from jsonb_array_elements('[
{"text":"Red Large Special","id":"3"},
{"text":"Blue Small","id":"4"},
{"text":"Green Medium Special","id":"5"}
]'::jsonb) with ordinality as f(element, idx)
)
SELECT first.element ->> 'text' AS first_name,
last.element ->> 'text' AS last_name
FROM first
JOIN last ON last.idx = first.idx