Draft
Commerce Coordinator is a broker for network calls to other services. Network calls to other services are expensive (e.g. time-consuming). We want to introduce caching to remove unwanted network calls to other services and improve response time.
Caching introduces a fair amount of complexity. Use caching in Coordinator only for endpoints where we expect a large volume of responses.
Coordinator will use LocMemCache for development and PyMemcacheCache for production.
Use PyMemcacheCache instead of MemcachedCache because it is now deprecated as of Django 3.2. See deprecation notice on Django's Memcache documentation.
Use PyMemcacheCache instead of PyLibMCCache because seems to be more used and is pure Python, and is faster and better maintained. See this discussion.
Use a default cache expiration time of 30 minutes. This is roughly the length of our average user payment checkout session.
Use TieredCache and get_cache_key from edx_django_utils.cache. See documentation.
Create wrapper functions around edx_django_utils.cache functions in Coordinator's core app.
Only use these cache wrapper functions to use the cache from Coordinator's other apps. This will help us centrally organize Coordinator's caching.
Prefer versioned cache keys instead of overwriting existing cache entries. This will prevent accidental overwrites.
For example, instead of constantly overwriting payment.<payment_number> every time the state of a payment changes, prefer creating a cache entry for each state, like payment.<payment_number>.<state_name>.
Here is an example wrapper function
from edx_django_utils.cache import get_cache_key
def get_payment_state_cache_key(payment_number, payment_state):
cache_key = get_cache_key(
cache_name='payment',
identifier=payment_number,
version=payment_state
)
return cache_keyAvoid overwriting cache entries. When overwriting, add a comment as to why overwriting is necessary.