How can one combine pyramid_beaker decorator-style caching and pyramid_tm?

Question

With Pyramid / pyramid_tm I end up passing around the dbsession object that joined the current transaction to methods that encapsulate my code. For example

def get_item_name(dbsession, item_id):
    item = dbsession.query(...)
    return item.name

def view_func(request):
    item_name = get_item_name(request.dbsession, request.GET['item_id']
    ...

Now if I wanted to cache the results of get_item_name using the beaker cache decorator @cache_region for example, I would never get a cache hit since dbsession is part of the cache key.

What is a common pattern to solve this elegantly?

Since pyramid, I am using dogpile cache for caching. I am pretty sure it has some mechanism to ignore the first argument when calculating the key, but I cannot find any documentation :-( — Petr Blahos, May 05 '21 at 06:26

score 1 · Accepted Answer · answered May 12 '21 at 04:07

1

At least one pattern is to decorate a closure that only accepts the args you want to cache and accesses other variables like request/dbsession from nonlocal scope.

def view_func(request):
    @cache.region('short_term')
    def get_item_name(item_id):
        item = request.dbsession.query(...)
        return item.name
    item_name = get_item_name(request.GET['item_id'])

This pattern is common in the beaker docs.

answered May 12 '21 at 04:07

Michael Merickel

23,153
3
54
70

Thanks for confirming, that's also the only solution that I came up with spontaneously. – barbaz May 12 '21 at 09:34

How can one combine pyramid_beaker decorator-style caching and pyramid_tm?

1 Answers1