I am trying to convert this into a sqlalchemy expression compatible with the language, I do not know if this is possible out of the box and hope that someone more experienced can help me. The backend is PostgreSQL, and if I cannot do it as an expression, I will create a string instead of ::
SELECT
DISTINCT date_trunc('month', x.x) as date,
COALESCE(b.res1, 0) AS res1,
COALESCE(b.res2, 0) AS res2
FROM
generate_series(
date_trunc('year', now() - interval '1 years'),
date_trunc('year', now() + interval '1 years'),
interval '1 months'
) AS x
LEFT OUTER JOIN(
SELECT
date_trunc('month', access_datetime) AS when,
count(NULLIF(resource_id != 1, TRUE)) AS res1,
count(NULLIF(resource_id != 2, TRUE)) AS res2
FROM tracking_entries
GROUP BY
date_trunc('month', access_datetime)
) AS b
ON (date_trunc('month', x.x) = b.when)
First of all, I got the TrackingEntry class mapped to tracking_entries, the select statement inside the external connection can be converted to something like (pseudo-code) ::
from sqlalchemy.sql import func, select
from datetime import datetime, timedelta
stmt = select([
func.date_trunc('month', TrackingEntry.resource_id).label('when'),
func.count(func.nullif(TrackingEntry.resource_id != 1, True)).label('res1'),
func.count(func.nullif(TrackingEntry.resource_id != 2, True)).label('res2')
],
group_by=[func.date_trunc('month', TrackingEntry.access_datetime), ])
Given the external selection operator, I have no idea how to build it, I think something like:
outer = select([
func.distinct(func.date_trunc('month', ?)).label('date'),
func.coalesce(?.res1, 0).label('res1'),
func.coalesce(?.res2, 0).label('res2')
],
from_obj=[
func.generate_series(
datetime.now(),
datetime.now() + timedelta(days=365),
timedelta(days=1)
).label(x)
])
Then I suggest that I need to bind these statements together without using foreign keys:
outer.outerjoin(stmt???).??(func.date_trunc('month', ?.?), ?.when)
Anyone have suggestions or even a better solution?
http://pastie.org/994367