Postgres distinct on sqlalchemy join：子查询必须只返回一列

Question

我在这些帖子上有一张帖子表和一张 cmets 表。没有外键约束，但是评论上的 post_id 指的是帖子上的 id。

我正在使用这样的查询来获取最近的帖子列表，其中包含对该帖子的最新评论的文本：

SELECT posts.*, comments.text
FROM posts
LEFT JOIN (
    SELECT DISTINCT ON ("post_id") *
    FROM comments
    ORDER BY "post_id", "timestamp" DESC
) comments ON posts.id=comments.post_id
WHERE posts.timestamp >= start_date;

我想将此查询转换为与 sqlalchemy 表达式一起使用。我试过这个：

query = select(posts, comments.c.text).\
    join(comments, and_(select(comments).
        distinct(comments.c.post_id).
        order_by(comments.c.post_id, comments.c.timestamp.desc()),
        posts.c.id == comments.c.post_id), isouter=True).\
    where(posts.c.timestamp >= start_date)

但它创建的查询略有不同：

SELECT posts.id, posts.[other posts columns...], comments.text
FROM posts 
LEFT OUTER JOIN comments ON (
    SELECT DISTINCT ON (comments.post_id) comments.text, comments.[other comments columns...]
    FROM comments 
    ORDER BY comments.post_id, comments.timestamp DESC
) AND posts.id = comments.post_id 
WHERE posts.timestamp >= start_date;

导致错误：sqlalchemy.exc.ProgrammingError: (psycopg2.errors.SyntaxError) subquery must return only one column

Answer 1

试试这个

comments_subquery = db.session.query(comments).distinct(comments.post_id, comments.text).subquery()

query = db.session.query(posts, comments.text)\
        .outerjoin(comments_subquery, comments_subquery.c.post_id==posts.id)
        .filter(posts.timestamp >= start_date)
        .all()