Query a specific JSON column (postgres) with sqlalchemy

Question

I have a model with a JSON field:

class Item(db.Model)
   ...
   data = db.Column(JSON, nullable=False)
   ...

The data contains some JSON such as:

{
    "cost": 10.00,
    "passengers": 2,
    "surcharge": 1.6
}

I want to be able to get a sum of the cost across all rows in the table with a filter. I tried the following but that didn't seem to work.

db.session.query(func.count(Item.data['cost'])).filter(
          Item.data["surcharge"].cast(Float) > 1
      ).scalar()

Ilja Everilä · Accepted Answer · 2017-08-15T11:35:10.853

9

You're using the wrong aggregate. count(expression) counts the number of rows for which the expression is not null. If you want a sum, use sum(expression):

db.session.query(func.sum(Item.data['cost'].astext.cast(Numeric))).\
    filter(Item.data['surcharge'].astext.cast(Numeric) > 1).\
    scalar()

Note that monetary values and binary floating point math is a bad mixture due to binary floats not being able to represent all decimal values. Instead use a proper monetary type, or Numeric in which case SQLAlchemy uses Decimal to represent the results in Python.

edited Aug 15 '17 at 11:35

answered Aug 15 '17 at 11:01

Ilja Everilä

50,538
7
126
127

Though PostgreSQL `JSONB` stores numbers in numeric, and `JSON` as decimal strings... However, needs some special handling with Python ... – Antti Haapala -- Слава Україні Aug 15 '17 at 11:14
Great answer but I had to use `as_string()` instead of `astext` (SQLAlchemy 1.3.23) – Urik Dec 02 '21 at 10:47

score 0 · Answer 2 · answered Aug 15 '17 at 09:05

0

You need to use Float instead of Integer as argument of cast

db.session.query(func.count(Item)).filter(
          Item.data['surcharge'].cast(Float) > 1
      ).all()

answered Aug 15 '17 at 09:05

vishes_shell

22,409
6
71
81

Fair point, I've updated the query but that still doesn't work. It just returns 0. – skittles789 Aug 15 '17 at 09:18

Query a specific JSON column (postgres) with sqlalchemy

2 Answers2

Linked