Use a SQL view with a SQLAlchemy ORM class?
in my Postgres DB I have one table called actions. I found myself doing the same complex query over and over, so I decided to create a view (portfolio).
CREATE TABLE actions(
id UUID NOT NULL PRIMARY KEY,
utc_date TIMESTAMP WITHOUT TIME ZONE NOT NULL,
action_type TEXT NOT NULL,
coin TEXT NOT NULL,
action_id BYTEA NOT NULL,
amount NUMERIC(13, 8) NOT NULL,
investment NUMERIC(7, 2) NOT NULL DEFAULT 0.00,
wallet TEXT NOT NULL,
FOREIGN KEY(coin) REFERENCES coins(coin_token),
FOREIGN KEY(wallet) REFERENCES wallets(name),
FOREIGN KEY(action_type) REFERENCES action_type(name)
);
CREATE VIEW portfolio AS
SELECT coin,
SUM(amount) AS amount,
SUM(investment) AS investment,
SUM(investment) / SUM(amount) AS min_price
FROM actions
GROUP BY coin
HAVING SUM(amount) > 0.00
ORDER BY coin;
CREATE VIEW actual_investment AS
SELECT SUM(investment)
FROM actions
WHERE action_type IN ('DEPOSIT', 'WITHDRAW');
in my python code, I declared a class for actions using sqlalchemy
. So far it works as expected, I can read&write to that table. But I haven't found a way to map portfolio something like a read-only table. The workaround that I have implemented is to emulate the SQL query in Python (which I don't like a bit).
is this even possible?
NOTE: I found this thread but I don't have a primary key in the view Is possible to mapping view with class using mapper in SqlAlchemy?
class Portfolio(Base):
__table__ = Table(
"portfolio",
Base.metadata,
Column("coin", String, primary_key=True),
Column("amount", Float(precision=13), nullable=False),
Column("investment", Float(precision=7), nullable=False),
)
the above code produces the following error:
sqlalchemy.exc.ArgumentError: Mapper mapped class Portfolio->portfolio could not assemble any primary key columns for mapped table 'portfolio'
As always any help or hint will be appreciate it :-)
Solution 1:
I found a "solution" that works in my use case. Since I know the column "coin" in my view will hold unique values, I re-defined that column as a primary key. I guess this will be true for any column that is being grouped by.
I hope this solution works for everyone who runs into this post :-)