Ordering and pagination in SQL-alchemy using non-sql ranking

Question

I have a python algorithm which returns a rank ordering of database rows for a given user. This algorithm outputs a list of primary key ids (these keys could be joined with post.id). It looks like the below, except that there are potentially thousands of matches:

result_rank = [1286, 1956, 6759, 3485, 2309, 3409, 9023, 912, 13098, 23489, 19023, 1239]

I want to instruct sqlalchemy to select these rows, AND order them as they are ordered in the list. The catch is that I would like to perform pagination on this

results = posts.query().filter(posts.id.in_(
    resultIds)).order_by(??? how can I order by post.id = result_rank ???).paginate(page_num, posts_per_page)

I am using Postgresql as the database.

Answer 1

Unless there is a good solution, I'm going to hack together my own paginate object:

class paginate_obj:

    """ Pagination dummy object. Takes a list and paginates it similar to sqlalchemy paginate() """
    def __init__(self, paginatable, page, per_page):
        self.has_next = (len(paginatable)/per_page) > page
        self.has_prev = bool(page - 1)
        self.next = page + self.has_next
        self.prev = page - self.has_prev
        self.items = paginatable[(page-1)*(per_page):(page)*(per_page)]

I think the only way to do ordering is to create a list of all results and sort it in python according to some lambda function:

results = my_table.query().all()
results.sort(key=lamba x: distance(x.lat, x.long, user_lat, user_long)
paginated_results = paginate_obj(results, 1, 10) #returns the first page of 10 elements

Answer 2

I think that the ordering is more important, because without it the database level pagination is completely useless. Having noted it, my answer does not cover pagination aspect at all, but I assume that even the answer provided by @mgoldwasser can be used for this.

This is what I came up with in order to be able to select some objects and preserve the order of them as per initial filter list. The code is self explanatory:

# input
post_ids = [3, 4, 1]

# create helper (temporary in-query table with two columns: post_id, sort_order)
# this table looks like this:
# key | sort_order
#   3 |          0
#   4 |          1
#   1 |          2
q_subq = "\nUNION ALL\n".join(
    "SELECT {} AS key, {} AS sort_order".format(_id, i)
    for i, _id in enumerate(post_ids)
)

# wrap it in a `Selectable` so that we can use JOINs
s = (select([literal_column("key", Integer),
             literal_column("sort_order", Integer)])
     .select_from(text("({}) AS helper".format(text(q_subq))))
     ).alias("helper")

# actual query which is both the filter and sorter
q = (session.query(Post)
     .join(s, Post.id == s.c.key)  # INNER JOIN will filter implicitly
     .order_by(s.c.sort_order)  # apply sort order
     )

It works on both postgresql and sqlite.