SQLAlchemy Create View in PostgresQL

Question

I am trying create a view with SQLAlchemy with Postgresql as the underlying DB. The separate select query to create the view works well and returns results but when I use it in the create view, I get the error sqlalchemy.exc.NoSuchTableError: popular which means the view is not being selected. I am getting the error when I try to select from the view. Creating the view does not throw any error but it does not create the view. Here is my code:

from sqlalchemy import *
import sqlalchemy as db
from sqlalchemy import func
from sqlalchemy import desc
from sqlalchemy import Table
from sqlalchemy.ext.compiler import compiles
from sqlalchemy.sql.expression import Executable, ClauseElement

try:
    engine = db.create_engine('postgresql://user:pass@localhost:5432/db_name')
    connection = engine.connect()

except:
    print('Error establishing DB connection')

# Import metadata
metadata = db.MetaData()

# Import articles, authors and log tables
art = db.Table('articles', metadata, autoload=True, autoload_with=engine)
aut = db.Table('authors', metadata, autoload=True, autoload_with=engine)
log = db.Table('log', metadata, autoload=True, autoload_with=engine)


class CreateView(Executable, ClauseElement):
    def __init__(self, name, select):
        self.name = name
        self.select = select


@compiles(CreateView)
def visit_create_view(element, compiler, **kw):
    return "CREATE VIEW %s AS %s" % (
         element.name,
         compiler.process(element.select, literal_binds=True)
         )



# Method to create view with top three articles
def view_top_three():
    top_three_view = CreateView('popular', db.select([art.columns.title, func.count(log.columns.path)]) \
        .where(func.concat('/article/', art.columns.slug) == log.columns.path) \
        .where(log.columns.path != "/") \
        .group_by(log.columns.path, art.columns.title) \
        .order_by(desc(func.count(log.columns.path))) \
        .limit(3))

    engine.execute(top_three_view)
    v = Table('popular', metadata, autoload=True, autoload_with=engine)
    for r in engine.execute(v.select()):
        print(r)


# Call the method which creates view and selects from view
view_top_three()

Any help will be appreciated.

Answer 1

Since your CreateView inherits from Executable, and ClauseElement, it is not considered a data changing operation. In other words

engine.execute(top_three_view)

executes the CREATE VIEW statement and then implicitly rollbacks, when the connection is returned to the pool.

Instead it should be a subclass of DDLElement, as shown in the usage recipes wiki. Simply changing the baseclass will allow SQLAlchemy autocommit to work properly.