On PostgreSQL 9.3.4, I have a JSON type column called "person" and the data stored in it is in the format {dogs: [{breed: <>, name: <>}, {breed: <>, name: <>}]}. I want to retrieve the breed of dog at index 0. Here are the two queries I ran:
Doesn't work
db=> select person->'dogs'->>0->'breed' from people where id = 77;
ERROR: operator does not exist: text -> unknown
LINE 1: select person->'dogs'->>0->'bree...
^
HINT: No operator matches the given name and argument type(s). You might need to add explicit type casts.
Works
select (person->'dogs'->>0)::json->'breed' from es_config_app_solutiondraft where id = 77;
?column?
-----------
"westie"
(1 row)
Why is the type casting necessary? Isn't it inefficient? Am I doing something wrong or is this necessary for postgres JSON support?
This is because operator ->> gets JSON array element as text. You need a cast to convert its result back to JSON.
You can eliminate this redundant cast by using operator ->:
select person->'dogs'->0->'breed' from people where id = 77;
Related
I need to clean JSON data that could look like:
{
"reference":"0000010-CAJ",
"product_code":"00000-10",
"var_name":"CAJ-1",
"doc_date":"2020-02-09T21:01:01-05:00",
"due_date":"2020-03-10T21:01:01-05:00",
}
However, this is just one of many other possibilities (is for a log aggregation that gets data from many sources).
I need to replace "-" with "_", but without break the dates like "2020-03-10T21:01:01-05:00", so can't simply cast to string and do a replace. I wonder if exist an equivalent of:
for (k,v) in json:
if is_text(v):
v = replace(...)
You can check with a regex if the value looks like a timestamp:
update the_table
set the_column = (select
jsonb_object_agg(
key,
case
when value ~ '^[0-9]{4}-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}.*' then value
else replace(value, '-', '_')
end)
from jsonb_each_text(the_column) as t(key, value))
This iterates over all key/value pairs in the JSON column (using jsonb_each_text()) and assembles all of them back into a JSON again (using jsonb_object_agg()). Values that look like a timestamp are left unchanged, for all others, the - is replaced with a _.
First week having to deal with a MYSQL database and JSON field types and I cannot seem to figure out why values are encoded automatically and then returned in encoded format.
Given the following SQL
-- create a multiline string with a tab example
SET #str ="Line One
Line 2 Tabbed out
Line 3";
-- encode it
SET #j = JSON_OBJECT("str", #str);
-- extract the value by name
SET #strOut = JSON_EXTRACT(#J, "$.str");
-- show the object and attribute value.
SELECT #j, #strOut;
You end up with what appears to be a full formed JSON object with a single attribute encoded.
#j = {"str": "Line One\n\tLine 2\tTabbed out\n\tLine 3"}
but using JSON_EXTRACT to get the attribute value I get the encoded version including outer quotes.
#strOut = "Line One\n\tLine 2\tTabbed out\n\tLine 3"
I would expect to get my original string with the \n \t all unescaped to the original values and no outer quotes. as such
Line One
Line 2 Tabbed out
Line 3
I can't seem to find any JSON_DECODE or JSON_UNESCAPE or similar functions.
I did find a JSON_ESCAPE() function but that appears to be used to manually build a JSON object structure in a string.
What am I missing to extract the values to the original format?
I like to use handy operator ->> for this.
It was introduced in MySQL 5.7.13, and basically combines JSON_EXTRACT() and JSON_UNQUOTE():
SET #strOut = #J ->> '$.str';
You are looking for the JSON_UNQUOTE function
SET #strOut = JSON_UNQUOTE( JSON_EXTRACT(#J, "$.str") );
The result of JSON_EXTRACT() is intentionally a JSON document, not a string.
A JSON document may be:
An object enclosed in { }
An array enclosed in [ ]
A scalar string value enclosed in " "
A scalar number or boolean value
A null — but this is not an SQL NULL, it's a JSON null. This leads to confusing cases because you can extract a JSON field whose JSON value is null, and yet in an SQL expression, this fails IS NULL tests, and it also fails to be equal to an SQL string 'null'. Because it's a JSON type, not a scalar type.
My json field data is like this:
{"active":true,"id":"xxxxxxxxxxxxxxxxx","settings":{"secret":"xxxxxxxxxxxxxxxxxxxxxxxxxx","token":"xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx","expires":"2019-12-16 01:11:23"},"plan":"Sample"}
Then I tried to query the field like this:
select * from integrations.accounts where field -> 'id' = 'xxxxxx';
But it gives me an error of:
SQL Error [42883]: ERROR: operator does not exist: json -> unknown
I found that the arrow operator (->) is not supported in version 9.2:
Unsupported versions: 9.3 / 9.2
Is there any alternative way to do this?
Since there is no true support for JSON in your version (It is really recommended to upgrade!) you have to parse your JSON text manually in any way. For example with help of regexp:
demo:db<>fiddle
SELECT
(regexp_matches(my_json_value::text, '"id":"(.*?)"'))[1]
-> is used for traversing the nested JSON
->> is used for the selection of querying level
In the above example if you wanna query on token which is inside settings you can use the following command
select * from integrations.accounts where field -> 'settings' ->> 'token' = 'xxxxxx';
To query on id which is at the top level use the following command
select * from integrations.accounts where field ->> 'id' = 'xxxxxx';
I have to extract data from a json file who contains spatial information. The content of this file is
{"vertices":[{"lat":46.744628268759314,"lon":6.569952920654968},
{"lat":46.74441692818192,"lon":6.570487107359068},
{"lat":46.74426116111054,"lon":6.570355867853787},
{"lat":46.74447250168793,"lon":6.569821681149689}],
"name":"demo-field",
"cropType":"sugarbeet",
"cropPlantDistance":0.18000000715255737,
"rowDistance":0.5,"numberOfRows":[28,12,12],"seedingDate":"2016-08-17T07:39+00:00"}
I've created a table then copied the content of this file into it
create table field(data json);
COPY field(data) FROM '/home/guest-pc5/field.json';
I now I can query my data
SELECT json_array_elements(data->'vertices') from field;
{"lat":46.744628268759314,"lon":6.569952920654968}
{"lat":46.74441692818192,"lon":6.570487107359068}
{"lat":46.74426116111054,"lon":6.570355867853787}
{"lat":46.74447250168793,"lon":6.569821681149689}
(4 rows)
The problem is that I can't use it like that. I would like to catch only values of "lat" and "lon" to put them in the field table
I've tried to use the function json_to_recordset without success
select * from json_to_recordset('[{"lat":46.744628268759314,"lon":6.569952920654968},{"lat":46.74441692818192,"lon":6.570487107359068},{"lat":46.74426116111054,"lon":6.570355867853787},{"lat":46.74447250168793,"lon":6.569821681149689}]') as (lat numeric, lon numeric);
ERROR: function json_to_recordset(unknown) does not exist
LINE 1: select * from json_to_recordset('[{"lat":46.744628268759314,...
^
HINT: No function matches the given name and argument types. You might need to add explicit type casts.
You can use json manipulator operator ->> to get the value you want as text out of json_array_elements output. To make it easier, you can call json_array_elements in FROM clause (which is a lateral call to a set-returning function):
SELECT
f.data AS original_json,
CAST((e.element->>'lat') AS numeric) AS lat,
CAST((e.element->>'lon') AS numeric) AS lon
FROM
field AS f,
json_array_elements(f.data->'vertices') AS e(element);
With that, you can simple create a table (or use INSERT into an existent one):
CREATE TABLE coordinates AS
SELECT
f.data AS original_json,
CAST((e.element->>'lat') AS numeric) AS lat,
CAST((e.element->>'lon') AS numeric) AS lon
FROM
field AS f,
json_array_elements(f.data->'vertices') AS e(element);
OBS: The LATERAL there is implicit, as the LATERAL keyword is optional for set-returning function calls, but you could make it really explicit, as:
FROM
field f
CROSS JOIN LATERAL json_array_elements(f.data->'vertices') AS e(element);
Also, LATERAL is 9.3+ only, although you are certainly above that as you are using json_array_elements (also 9.3+ only).
In a PostgreSQL 9.5 table I have an integer column social.
When I try to update it in a stored procedure given the following JSON data (an array with 2 objects, each having a "social" key) in the in_users variable of type jsonb:
'[{"sid":"12345284239407942","auth":"ddddc1808197a1161bc22dc307accccc",**"social":3**,"given":"Alexander1","family":"Farber","photo":"https:\/\/graph.facebook.com\/1015428423940942\/picture?type=large","place":"Bochum,
Germany","female":0,"stamp":1450102770},
{"sid":"54321284239407942","auth":"ddddc1808197a1161bc22dc307abbbbb",**"social":4**,"given":"Alxander2","family":"Farber","photo":null,"place":"Bochum,
Germany","female":0,"stamp":1450102800}]'::jsonb
Then the following code is failing:
FOR t IN SELECT * FROM JSONB_ARRAY_ELEMENTS(in_users)
LOOP
UPDATE words_social SET
social = t->'social',
WHERE sid = t->>'sid';
END LOOP;
with the error message:
ERROR: column "social" is of type integer but expression is of type jsonb
LINE 3: social = t->'social',
^
HINT: You will need to rewrite or cast the expression.
I have tried changing that line to:
social = t->'social'::int,
but then I get the error:
ERROR: invalid input syntax for integer: "social"
LINE 3: social = t->'social'::int,
^
Why doesn't PostgreSQL recognize that the data is integer?
From the JSON-TYPE-MAPPING-TABLE I was having the impression that JSON number would be auto-converted to PostgreSQL numeric type.
A single set-based SQL command is far more efficient than looping:
UPDATE words_social w
SET social = (iu->>'social')::int
FROM JSONB_ARRAY_ELEMENTS(in_users) iu -- in_user = function variable
WHERE w.sid = iu->>'sid'; -- type of sid?
To answer your original question:
Why doesn't PostgreSQL recognize that the data is integer?
Because you were trying to convert the jsonb value to integer. In your solution you already found that you need the ->> operator instead of -> to extract text, which can be cast to integer.
Your second attempt added a second error:
t->'social'::int
In addition to the above: operator precedence. The cast operator :: binds stronger than the json operator ->. Like you found yourself already, you really want:
(t->>'social')::int
Very similar case on dba.SE:
Querying JSONB in PostgreSQL
I've ended up using:
FOR t IN SELECT * FROM JSONB_ARRAY_ELEMENTS(in_users)
LOOP
UPDATE words_social SET
social = (t->>'social')::int
WHERE sid = t->>'sid';
IF NOT FOUND THEN
INSERT INTO words_social (social)
VALUES ((t->>'social')::int);
END IF;
END LOOP;