Cache management

Here we describe the strategy we use to manage the cache.

We have three kinds of services:

  • Without any cache (no_cache).

  • Cache related to the application version (public_cache).

  • Cache related to the application version and the current role (private_cache).

For some services, we also set the header Vary to Accept-Language to have one cache per language.

Server side

On the server side, we set the Cache-Control header with a different value depending on the cache type.

  • no_cache: no-cache.

  • public_cache: max-age=864000, public.

  • private_cache: max-age=864000, public or max-age=864000, private if we are logged in.

The max-age is configurable in a vars file in the default_max_age attribute.

Client side

On the client side, some attributes are added in the URL parameters to force to have the right version of the cache. It will be done automatically by the route_url method.

  • no_cache: nothing to add (be sure that the lib does not add custom cache).

  • public_cache: add the query param named cache_version.

  • private_cache: add the query param named cache_version and role.

Service list

No cache

  • interface pages

  • dynamic

  • api js

  • api help

  • auto login form (loginform403)

  • login

  • logout

  • login change password

  • echo

  • full text search

  • layers/read_many

  • layers/read_one

  • layers/count

  • layers/create

  • layers/update

  • layers/delete

  • WMS/GetMap

  • WMS/GetFeatureInfo

  • WFS/GetFeature

  • WFS all POST

  • PDF report

  • print/report.pdf

  • print/status

  • print/get file

  • profile

  • raster

  • shortener

Public cache

  • login form

  • layers/enumerate_attribute_values

  • WMS/GetLegendGraphic

Private cache

  • themes

  • layers/metadata

  • WMS/GetCapabilities

  • WMS/DescribeLayer

  • WFS/GetCapabilities

  • WFS/DescribeFeaturesType

  • print/capabilities.json

Internal cache

We have two different internal cache types, one (std) for the serializable objects (Like JSON) and one (obj) for the non serializable objects.

The std cache is stored on redis, and the obj cache is stored in memory.

Example usage:

from c2cgeoportal_geoportal.caching import get_region
CACHE_REGION = get_region("std")

@CACHE_REGION.cache_on_arguments()
def method_to_be_cached():
  return {
    "key": "value"
  }

from c2cgeoportal_geoportal.caching import get_region
CACHE_REGION_OBJ = get_region("obj")

@CACHE_REGION_OBJ.cache_on_arguments()
def method_to_be_cached():
  return AnyObject()