Optimise queries run for creating pagination links

We have been running in to the problem in
https://github.com/pallets/flask-sqlalchemy/issues/518 where
our page loads very slow when viewing a single page of notifications
for a service in the admin app. Tracing this back and using SQL
explain analyze I can see that getting the notifications takes about
a second but the second query to count how many notifications there
are (to work out if there is a next page of pagination) can take up
to 100 seconds.

As suggested in that issue, we do the pagination ourselves.
Our pagination doesn't need us to know exactly how many notifications
there are, just whether there are any on the next page and that can
be done without running the slow query to count how many
notifications in total by using `count_pages=False`.
This commit is contained in:
David McDonald
2021-12-03 17:07:03 +00:00
parent 989ef9c21a
commit c68d1a2f23
2 changed files with 33 additions and 10 deletions

View File

@@ -241,7 +241,8 @@ def get_notifications_for_service(
include_from_test_key=False,
older_than=None,
client_reference=None,
include_one_off=True
include_one_off=True,
error_out=True
):
if page_size is None:
page_size = current_app.config['PAGE_SIZE']
@@ -280,7 +281,8 @@ def get_notifications_for_service(
return query.order_by(desc(Notification.created_at)).paginate(
page=page,
per_page=page_size,
count=count_pages
count=count_pages,
error_out=error_out,
)

View File

@@ -420,6 +420,8 @@ def get_all_notifications_for_service(service_id):
include_from_test_key = data.get('include_from_test_key', False)
include_one_off = data.get('include_one_off', True)
# count_pages is not being used for whether to count the number of pages, but instead as a flag
# for whether to show pagination links
count_pages = data.get('count_pages', True)
pagination = notifications_dao.get_notifications_for_service(
@@ -427,7 +429,7 @@ def get_all_notifications_for_service(service_id):
filter_dict=data,
page=page,
page_size=page_size,
count_pages=count_pages,
count_pages=False,
limit_days=limit_days,
include_jobs=include_jobs,
include_from_test_key=include_from_test_key,
@@ -442,25 +444,44 @@ def get_all_notifications_for_service(service_id):
else:
notifications = notification_with_template_schema.dump(pagination.items, many=True).data
# We try and get the next page of results to work out if we need provide a pagination link to the next page
# in our response if it exists. Note, this could be done instead by changing `count_pages` in the previous
# call to be True which will enable us to use Flask-Sqlalchemy to tell if there is a next page of results but
# this way is much more performant for services with many results (unlike Flask SqlAlchemy, this approach
# doesn't do an additional query to count all the results of which there could be millions but instead only
# asks for a single extra page of results).
next_page_of_pagination = notifications_dao.get_notifications_for_service(
service_id,
filter_dict=data,
page=page + 1,
page_size=page_size,
count_pages=False,
limit_days=limit_days,
include_jobs=include_jobs,
include_from_test_key=include_from_test_key,
include_one_off=include_one_off,
error_out=False # False so that if there are no results, it doesn't end in aborting with a 404
)
def get_prev_next_pagination_links(pagination, endpoint, **kwargs):
def get_prev_next_pagination_links(current_page, next_page_exists, endpoint, **kwargs):
if 'page' in kwargs:
kwargs.pop('page', None)
links = {}
if pagination.has_prev:
links['prev'] = url_for(endpoint, page=pagination.prev_num, **kwargs)
if pagination.has_next:
links['next'] = url_for(endpoint, page=pagination.next_num, **kwargs)
if page > 1:
links['prev'] = url_for(endpoint, page=page - 1, **kwargs)
if next_page_exists:
links['next'] = url_for(endpoint, page=page + 1, **kwargs)
return links
return jsonify(
notifications=notifications,
page_size=page_size,
links=get_prev_next_pagination_links(
pagination,
page,
len(next_page_of_pagination.items),
'.get_all_notifications_for_service',
**kwargs
)
) if count_pages else {}
), 200