Using EXISTS with bound parameters in sub-query with Sequelize - node.js

Question

Suppose I have a table of something like cars, where inside is a JSONB object specifying possible customizations:
+----+-----------------------------------------+
| id | customizations JSONB |
+----+-----------------------------------------+
| 1 | {"color": "blue", "lights": "led", ...} |
+----+-----------------------------------------+
| 2 | {"color": "red"} |
+----+-----------------------------------------+
| 3 | {} |
+----+-----------------------------------------+
If I want to query for a certain customization based on case-insensitive value or partial value (i.e., ILIKE), I can do something like:
SELECT * FROM "Cars" WHERE EXISTS (
SELECT 1 FROM JSONB_EACH_TEXT("customizations") WHERE "value" ~* 'BLU'
);
This pattern works fine in Postgres, but now I am trying to translate it over to Sequelize as best as I can. The important thing here is that the search term ('BLU' from the example) is passed as a parameter.
const cars = await cars.findAll({
where: // ???? Maybe something like db.fn('EXISTS', ...), but how do I bind the parameter?
});
How can I use an EXISTS query here, but bind a parameter? I know I could use db.literal(), but if I did that, I'd have to escape the search term string myself before interpolating it into the query. (Is there at least a proper method for doing this data escaping in Sequelize?)
Note that the JSONB object in customizations can have many keys, a single key, or even no keys.
Bounty Note: Answers using modern versions of Postgres are fine, but I would also like an answer for PostgreSQL v10.12, as that's all that is available with AWS Aurora Serverless. I'll happily assign a separate bounty to both answers!

Answer 1

If on postgresql 12+, it would be possible to use the json path expression to extract all values & cast to text in order to regex match
The raw query would be
SELECT *
FROM cars
WHERE jsonb_path_query_array(customizations, '$.*')::TEXT ~* 'BLU'
in sequelize:
where: Sequelize.where(
Sequelize.literal("jsonb_path_query_array(customizations, '$.*')::TEXT"),
Op.iRegexp,
'BLU'
)
On older versions, i would probably use a raw query with bound parameters that maps back to the Car object.
How can I use an EXISTS query here, but bind a parameter? I know I could use db.literal(), but if I did that, I'd have to escape the search term string myself before interpolating it into the query. (Is there at least a proper method for doing this data escaping in Sequelize?)
you can bind parameters to raw sql queries using either $1 or $name for positional (array) and named (object) arguments.
// cars.js
const Sequelize = require('sequelize');
const path = 'postgresql://hal:hal#localhost:5432/hal'
const sequelize = new Sequelize(path);
let Car = sequelize.define('cars', {
id: { type: Sequelize.INTEGER, primaryKey: true },
customizations: Sequelize.JSONB
});
let cars = [{ id: 1, customizations: {color: "blue", lights: "led"} },
{ id: 2, customizations: {color: "red"} },
{ id: 3, customizations: {} }]
const search_pat = 'BLU';
const stmt = `
select distinct cars.*
from cars, jsonb_each_text(customizations) kv(kk, vv)
where vv ~* $search_pattern
`;
sequelize
.sync()
.then(() => {
Car.bulkCreate(cars)
})
.then(() => {
sequelize.query(stmt, {
bind: {search_pattern: search_pat},
type: sequelize.QueryTypes.SELECT,
model: Car,
mapToModel: true
}).then(cars => {console.log(cars);})
});
executing this script (node cars.js) produces the following output:
Executing (default): CREATE TABLE IF NOT EXISTS "cars" ("id" INTEGER , "customizations" JSONB, "createdAt" TIMESTAMP WITH TIME ZONE NOT NULL, "updatedAt" TIMESTAMP WITH TIME ZONE NOT NULL, PRIMARY KEY ("id"));
Executing (default): SELECT i.relname AS name, ix.indisprimary AS primary, ix.indisunique AS unique, ix.indkey AS indkey, array_agg(a.attnum) as column_indexes, array_agg(a.attname) AS column_names, pg_get_indexdef(ix.indexrelid) AS definition FROM pg_class t, pg_class i, pg_index ix, pg_attribute a WHERE t.oid = ix.indrelid AND i.oid = ix.indexrelid AND a.attrelid = t.oid AND t.relkind = 'r' and t.relname = 'cars' GROUP BY i.relname, ix.indexrelid, ix.indisprimary, ix.indisunique, ix.indkey ORDER BY i.relname;
Executing (default): INSERT INTO "cars" ("id","customizations","createdAt","updatedAt") VALUES (1,'{"color":"blue","lights":"led"}','2021-04-14 04:42:50.446 +00:00','2021-04-14 04:42:50.446 +00:00'),(2,'{"color":"red"}','2021-04-14 04:42:50.446 +00:00','2021-04-14 04:42:50.446 +00:00'),(3,'{}','2021-04-14 04:42:50.446 +00:00','2021-04-14 04:42:50.446 +00:00') RETURNING "id","customizations","createdAt","updatedAt";
Executing (default): select distinct cars.*
from cars, jsonb_each_text(customizations) kv(kk, vv)
where vv ~* $1
[
cars {
dataValues: {
id: 1,
customizations: [Object],
createdAt: 2021-04-14T04:42:50.446Z,
updatedAt: 2021-04-14T04:42:50.446Z
},
_previousDataValues: {
id: 1,
customizations: [Object],
createdAt: 2021-04-14T04:42:50.446Z,
updatedAt: 2021-04-14T04:42:50.446Z
},
_changed: Set(0) {},
_options: {
isNewRecord: false,
_schema: null,
_schemaDelimiter: '',
raw: true,
attributes: undefined
},
isNewRecord: false
}
]
note that i've used an alternative (and likely less efficient select statement in the example above)
You can use your original query, which IMO is the most straight-forward query for the target postgresql version (10.x), i.e.
const stmt = `
SELECT *
FROM cars
WHERE EXISTS (
SELECT 1
FROM JSONB_EACH_TEXT(customizations) WHERE "value" ~* $search_pattern
)
`;

Answer 2

If you modify your condition like this:
SELECT * FROM "Cars" WHERE (
SELECT "value" FROM JSONB_EACH_TEXT("customizations")) ~* 'BLU'
then you can use Sequelize.where in a conjunction with Sequelize.literal:
where: Sequelize.where(
Sequelize.literal('(SELECT "value" FROM JSONB_EACH_TEXT("customizations"))'),
Op.iRegexp,
'BLU'
)
Upd.
This solution will work only if a subquery returns 1 record.