Related
I have a table with ProductID (int) and ProductGroups jsonb.
The ProductGroups just have values in the json rather than tag names. I want to be able to query the following data to get ProductID where ProductGroup contains 69.
ProductID ProductGroups
125481 [134, 83]
128166 [134, 83]
128175 [134, 83]
128172 [134, 83]
131492 [69, 134]
131489 [69, 134]
131860 [128, 131, 133, 100, 71]
128142 [134, 83]
I have queried what I think of as normal jsonb with tag names in different tables with below query where I callout the name and value
SELECT *
FROM trans."TxnHeader" mpt, jsonb_array_elements(mpt."ExtensionProperty") as ext
where 1=1
and jsonb_typeof(mpt."ExtensionProperty") = 'array'
and ext->>'Name' = 'posTranId' and ext->>'Value' = '8539'
You can do it using the containment operator #> as follows :
select *
from products
where ProductGroups #> '69'
Demo here
I've a column with jsonb type and contains list of elements either in string or integer format.
What I want now is to make all of them as same type e.g either all int or all string format
Tried: this way I get single element but I need to update all of the elements inside of the list.
SELECT parent_path -> 1 AS path
FROM abc
LIMIT 10
OR
Update abc SET parent_path = ARRAY[parent_path]::TEXT[] AS parent_path
FROM abc
OR
UPDATE abc SET parent_path = replace(parent_path::text, '"', '') where id=123
Current Output
path
[6123697, 178, 6023099]
[625953521394212864, 117, 6023181]
["153", "6288361", "553248635949090971"]
[553248635358954983, 178320, 174, 6022967]
[6050684, 6050648, 120, 6022967]
[653, 178238, 6239135, 38, 6023117]
["153", "6288496", "553248635977039112"]
[553248635998143523, 6023185]
[553248635976194501, 6022967]
[553248635976195634, 6022967]
Expected Output
path
[6123697, 178, 6023099]
[625953521394212864, 117, 6023181]
[153, 6288361, 553248635949090971] <----
[553248635358954983, 178320, 174, 6022967]
[6050684, 6050648, 120, 6022967]
[653, 178238, 6239135, 38, 6023117]
[153, 6288496, 553248635977039112] <----
[553248635998143523, 6023185]
[553248635976194501, 6022967]
[553248635976195634, 6022967]
Note: Missing double quotes on the list. I've tried several methods from here but no luck
You will have to unnest them, cleanup each element, then aggregate it back to an array:
The following converts all elements to integers:
select (select jsonb_agg(x.i::bigint order by idx)
from jsonb_array_elements_text(a.path) with ordinality as x(i, idx)
) as clean_path
from abc a;
You can use a scalar subquery to select, unnest, and aggregate the elements:
WITH mytable AS (
SELECT row_number() over () as id, col::JSONB
FROM (VALUES ('[6123697, 178, 6023099]'),
('["6123697", "178", "6023099"]')) as bla(col)
)
SELECT id, (SELECT JSONB_AGG(el::int) FROM jsonb_array_elements_text(col) as el)
FROM mytable
I'm having a frustrating issue trying to set a default during an ecto migration
In the migration the code looks like the following:
def encode(binary) do
"\\x" <> Base.encode16(binary, case: :lower)
end
Logger.debug("admin.id = #{inspect admin.id}")
Logger.debug("admin.id = #{inspect UUID.string_to_binary!(admin.id)}")
Logger.debug("admin.id = #{inspect encode(admin.id)}")
alter table(#questions) do
add :owner_id, references(:users, on_delete: :nothing, type: :binary_id), null: false, default: admin.id
end
You can see the attempts I tried above in the logger
I get the error
default values are interpolated as UTF-8 strings and cannot contain null bytes. `<<209, 241,
149, 133, 44, 81, 70, 164, 181, 120, 214, 0, 253, 191, 198, 214>>` is invalid. If you want
to write it as a binary, use "\xd1f195852c5146a4b578d600fdbfc6d6", otherwise refer to
PostgreSQL documentation for instructions on how to escape this SQL type
Any help would be great thanks
When using :binary_id with Postgres, Ecto expects you to pass UUIDs as strings. Your error message implies you tried to pass it as a binary, so you should first convert it to a string:
add :owner_id, references(:users, on_delete: :nothing, type: :binary_id), null: false, default: UUID.binary_to_string!(admin.id)
I'm struggling doing aggregations on a JSONB field in a PostgreSQL database. This is probably easier explained with an example so if create and populate a table called analysis with 2 columns (id and analysis) as follows: -
create table analysis (
id serial primary key,
analysis jsonb
);
insert into analysis
(id, analysis) values
(1, '{"category" : "news", "results" : [1, 2, 3, 4, 5 , 6, 7, 8, 9, 10, 11, 12, 13, 14, null, null]}'),
(2, '{"category" : "news", "results" : [11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, null, 26]}'),
(3, '{"category" : "news", "results" : [31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46]}'),
(4, '{"category" : "sport", "results" : [51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66]}'),
(5, '{"category" : "sport", "results" : [71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86]}'),
(6, '{"category" : "weather", "results" : [91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106]}');
As you can see the analysis JSONB field always contains 2 attributes category and results. The results attribute will always contain an fixed length array of size 16. I've used various functions such as jsonb_array_elements but what I'm trying to do is the following: -
Group by analysis->'category'
Average of each array element
When I want is a statement to return 3 rows grouped by category (i.e. news, sport and weather) and a 16 fixed length array containing averages. To further complicate things, if there are nulls in the array then we should ignore them (i.e. we are not simply summing and averaging by the number of rows). The result should look something like the following: -
category | analysis_average
-----------+--------------------------------------------------------------------------------------------------------------
"news" | [14.33, 15.33, 16.33, 17.33, 18.33, 19.33, 20.33, 21.33, 22.33, 23.33, 24.33, 25.33, 26.33, 27.33, 45, 36]
"sport" | [61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76]
"weather" | [91, 92, 93, 94, 95, 96, 97, 98, 99, 00, 101, 102, 103, 104, 105, 106]
NOTE: Notice the 45 and 36 in the last 2 array itmes on the 1st row which illustrates ignoring the nullss.
I had considered creating a view which exploded the array into 16 columns i.e.
create view analysis_view as
select a.*,
(a.analysis->'results'->>0)::int as result0,
(a.analysis->'results'->>1)::int as result1
/* ... etc for all 16 array entries .. */
from analysis a;
This seems extremely inelegant to me and removes the advantages of using an array in the first place but could probably hack something together using that approach.
Any pointers or tips will be most appreciated!
Also performance is really important here so the higher the performance the better!
This will work for any array length
select category, array_agg(average order by subscript) as average
from (
select
a.analysis->>'category' category,
subscript,
avg(v)::numeric(5,2) as average
from
analysis a,
lateral unnest(
array(select jsonb_array_elements_text(analysis->'results')::int)
) with ordinality s(v,subscript)
group by 1, 2
) s
group by category
;
category | average
----------+----------------------------------------------------------------------------------------------------------
news | {14.33,15.33,16.33,17.33,18.33,19.33,20.33,21.33,22.33,23.33,24.33,25.33,26.33,27.33,45.00,36.00}
sport | {61.00,62.00,63.00,64.00,65.00,66.00,67.00,68.00,69.00,70.00,71.00,72.00,73.00,74.00,75.00,76.00}
weather | {91.00,92.00,93.00,94.00,95.00,96.00,97.00,98.00,99.00,100.00,101.00,102.00,103.00,104.00,105.00,106.00}
table functions - with ordinality
lateral
Because the array is always of the same length, you can use generate_series instead of typing the index of every array element yourself. You CROSS JOIN with that generated series so the index is applied to every category and you can get every element at position s from the array. Then it is just aggregating the data using GROUP BY.
The query then becomes:
SELECT category, array_agg(val ORDER BY s) analysis_average
FROM (
SELECT analysis->'category' category, s, AVG((analysis->'results'->>s)::numeric) val
FROM analysis
CROSS JOIN generate_series(0, 15) s
GROUP BY category,s
) q
GROUP BY category
15 is in this case the last index of the array (16-1).
It can be done in more traditional way like
select
(t.analysis->'category')::varchar,
array_math_avg(array(select jsonb_array_elements_text(t.analysis->'results')::int))::numeric(9,2)[]
from
analysis t
group by 1 order by 1;
but we need to do some preparation:
create type t_array_math_agg as(
c int[],
a numeric[]
);
create or replace function array_math_sum_f(in t_array_math_agg, in numeric[]) returns t_array_math_agg as $$
declare
r t_array_math_agg;
i int;
begin
if $2 is null then
return $1;
end if;
r := $1;
for i in array_lower($2,1)..array_upper($2,1) loop
if coalesce(r.a[i],$2[i]) is null then
r.a[i] := null;
else
r.a[i] := coalesce(r.a[i],0) + coalesce($2[i],0);
r.c[i] := coalesce(r.c[i],0) + 1;
end if;
end loop;
return r;
end; $$ immutable language plpgsql;
create or replace function array_math_avg_final(in t_array_math_agg) returns numeric[] as $$
declare
r numeric[];
i int;
begin
if array_lower($1.a, 1) is null then
return null;
end if;
for i in array_lower($1.a,1)..array_upper($1.a,1) loop
r[i] := $1.a[i] / $1.c[i];
end loop;
return r;
end; $$ immutable language plpgsql;
create aggregate array_math_avg(numeric[]) (
sfunc=array_math_sum_f,
finalfunc=array_math_avg_final,
stype=t_array_math_agg,
initcond='({},{})'
);
Before some time i get here very reliable query for calculating ingredients from normatives but with time I get need for advancing a bit.
Here are example tables:
CREATE TABLE myusedfood
(mybill int, mydate text, food_code int, food_name text, qtyu integer, meas text);
INSERT INTO myusedfood (mybill, mydate, food_code, food_name, qtyu, meas)
VALUES (1, '03.01.2014', 10, 'spaghetti', 3, 'pcs'),
(2, '04.01.2014', 156, 'mayonnaise', 2, 'pcs'),
(3, '06.01.2014', 173, 'ketchup', 1, 'pcs'),
(4, '07.01.2014', 172, 'bolognese sauce', 2, 'pcs'),
(5, '08.01.2014', 173, 'ketchup', 1, 'pcs'),
(6, '15.01.2014', 175, 'worchester sauce', 2, 'pcs'),
(7, '16.01.2014', 177, 'parmesan', 1, 'pcs'),
(8, '17.01.2014', 10, 'spaghetti', 2, 'pcs'),
(9, '18.01.2014', 156, 'mayonnaise', 1, 'pcs'),
(10, '19.01.2014', 10, 'spaghetti', 2, 'pcs'),
(11, '19.01.2014', 1256, 'spaghetti rinf', 100, 'gramm'),
(12, '20.01.2014', 156, 'mayonnaise', 2, 'pcs'),
(13, '21.01.2014', 173, 'ketchup', 1, 'pcs'),
(14, '19.01.2014', 10, 'spaghetti', 2, 'pcs');
DROP TABLE IF EXISTS myingredients;
CREATE TABLE myingredients
(food_code int, ingr_code int, ingr_name text, qtyi decimal(10, 3), meas text);
INSERT INTO myingredients (food_code, ingr_code, ingr_name, qtyi, meas)
VALUES (10, 1256, 'spaghetti rinf', 75, 'gramm'),
(156, 1144, 'salt', 0.3, 'gramm'),
(10, 1144, 'salt', 0.5, 'gramm'),
(156, 1140, 'fresh egg', 50, 'gramm'),
(172, 1138, 'tomato', 80, 'gramm'),
(156, 1139, 'mustard', 5, 'gramm'),
(172, 1136, 'clove', 1, 'gramm'),
(156, 1258, 'oil', 120, 'gramm'),
(172, 1135, 'laurel', 0.4, 'gramm'),
(10, 1258, 'oil', 0.4, 'gramm'),
(172, 1130, 'corned beef', 40, 'gramm');
First table contains bill's number and date, notmative's/ingredient's code, name, selled quantity.
Second represent normative's code, ingredient's code, ingredient's name, ingredient's quantity in related normative.
With this query I get listed all ingredients from all normatives plus normatives which haven't any ingredients. That work OK.
SELECT SUM(f.qtyu) AS used,
COALESCE(i.ingr_code, f.food_code) AS code,
COALESCE(i.ingr_name, f.food_name) AS f_name,
SUM(COALESCE(i.qtyi, 1) * f.qtyu) AS qty,
COALESCE(i.meas, f.meas) AS meas
FROM myusedfood f LEFT JOIN myingredients i
ON f.food_code = i.food_code
GROUP BY COALESCE(i.ingr_code, f.food_code),
COALESCE(i.ingr_name, f.food_name),
COALESCE(i.meas, f.meas)
ORDER BY code;
With time my real tables becames huge and queries slower. I often need to get results for just one food code but I don't know how to do that properly regarding of speed.
For example with showed tables if we query for ingredient code 1256 result should be
used ing.code ing.name qty meas
------------------------------------------------------
109, 1256, "spaghetti rinf", 775.000, "gramm"
Where article 1256 is "used" 9 times in normative (10) and 100g :) as alone, all in qty of 775 gramms.
Here I have 2 questions:
How to optimize upper query to list only one code instead of all, say code 1256?
How to get list of all records related to code 1256.
Result should look like this:
bill date code name qty
------------------------------------------------------
1, '03.01.2014', 10, 'spaghetti', 225
8, '17.01.2014', 10, 'spaghetti', 150
10, '19.01.2014', 10, 'spaghetti', 150
11, '19.01.2014', 1256, 'spaghetti rinf', 100
14, '19.01.2014', 10, 'spaghetti', 150
------------------------------------------------------
775
Note that ingredient 1256 may be selled in normative (10) or alone as is.
Please help to get those two needed queries.