python-lazr.restful 0.19.3-0ubuntu2 / usr / share / pyshared / lazr / restful / example / base / tests / representation-cache.txt

**********************************
The in-memory representation cache
**********************************

Rather than having lazr.restful calculate a representation of an entry
every time it's requested, you can register an object as the
representation cache. String representations of entries are generated
once and stored in the representation cache.

lazr.restful works fine when there is no representation cache
installed; in fact, this is the only test that uses one.

    >>> from zope.component import getUtility
    >>> from lazr.restful.interfaces import IRepresentationCache
    >>> getUtility(IRepresentationCache)
    Traceback (most recent call last):
    ...
    ComponentLookupError: ...

DictionaryBasedRepresentationCache
==================================

A representation cache can be any object that implements
IRepresentationCache, but for test purposes we'll be using a simple
DictionaryBasedRepresentationCache. This object transforms the
IRepresentationCache operations into operations on a Python dict-like
object.

    >>> from lazr.restful.simple import DictionaryBasedRepresentationCache
    >>> dictionary = {}
    >>> cache = DictionaryBasedRepresentationCache(dictionary)

It's not a good idea to use a normal Python dict in production,
because there's no limit on how large the dict can become. In a real
situation you want something with an LRU implementation. That said,
let's see how the DictionaryBasedRepresentationCache works.

All IRepresentationCache implementations will cache a representation
under a key derived from the object whose representation it is, the
media type of the representation, and a web service version name.

    >>> from lazr.restful.example.base.root import C4 as greens_object
    >>> json = "application/json"
    >>> print cache.get(greens_object, json, "devel")
    None
    >>> print cache.get(greens_object, json, "devel", "missing")
    missing

    >>> cache.set(greens_object, json, "devel", "This is the 'devel' value.")
    >>> print cache.get(greens_object, json, "devel")
    This is the 'devel' value.
    >>> sorted(dictionary.keys())
    ['http://cookbooks.dev/devel/cookbooks/Everyday%20Greens,application/json']

This allows different representations of the same object to be stored
for different versions.

    >>> cache.set(greens_object, json, "1.0", "This is the '1.0' value.")
    >>> print cache.get(greens_object, json, "1.0")
    This is the '1.0' value.
    >>> sorted(dictionary.keys())
    ['http://cookbooks.dev/1.0/cookbooks/Everyday%20Greens,application/json',
     'http://cookbooks.dev/devel/cookbooks/Everyday%20Greens,application/json']

Deleting an object from the cache will remove all its representations.

    >>> cache.delete(greens_object)
    >>> sorted(dictionary.keys())
    []
    >>> print cache.get(greens_object, json, "devel")
    None
    >>> print cache.get(greens_object, json, "1.0")
    None

DO_NOT_CACHE
------------

Representation caches treat the constant object DO_NOT_CACHE
specially. If key_for() returns DO_NOT_CACHE, the resulting
object+type+version combination will not be cached at all.

    >>> class TestRepresentationCache(DictionaryBasedRepresentationCache):
    ...     def key_for(self, object, media_type, version):
    ...         if (object == "Don't cache this."
    ...             and version != "cache everything!"):
    ...             return self.DO_NOT_CACHE
    ...         return object + ',' + media_type + ',' + version

    >>> test_dict = {}
    >>> test_cache = TestRepresentationCache(test_dict)

The key_for() implementation defined above will not cache a
representation of the string "Don't cache this."

    >>> test_cache.set("Cache this.", json, "1.0", "representation")
    >>> test_cache.set("Don't cache this.", json, "1.0", "representation")
    >>> test_dict.keys()
    ['Cache this.,application/json,1.0']

...UNLESS the representation being cached is for the version "cache
everything!"

    >>> test_cache.set("Don't cache this.", json, "cache everything!",
    ...                "representation")
    >>> for key in sorted(test_dict.keys()):
    ...     print key
    Cache this.,application/json,1.0
    Don't cache this.,application/json,cache everything!

Even if an uncacheable value somehow gets into the cache, it's not
retrievable.

    >>> bad_key = "Don't cache this.,application/json,1.0"
    >>> test_dict[bad_key] = "This representation should not be cached."
    >>> print test_cache.get("Don't cache this.", json, "1.0")
    None

A representation cache
======================

Now let's register our DictionaryBasedRepresentationCache as the
representation cache for this web service, and see how it works within
lazr.restful.

    >>> from zope.component import getSiteManager
    >>> sm = getSiteManager()
    >>> sm.registerUtility(cache, IRepresentationCache)

    >>> from lazr.restful.testing.webservice import WebServiceCaller
    >>> webservice = WebServiceCaller(domain='cookbooks.dev')

When we retrieve a JSON representation of an entry, that
representation is added to the cache.

    >>> recipe_url = "/recipes/1"
    >>> ignored = webservice.get(recipe_url)
    >>> [the_only_key] = dictionary.keys()
    >>> print the_only_key
    http://cookbooks.dev/devel/recipes/1,application/json

Note that the cache key incorporates the web service version name
("devel") and the media type of the representation
("application/json").

Associated with the key is a string: the JSON representation of the object.

    >>> import simplejson
    >>> print simplejson.loads(dictionary[the_only_key])['self_link']
    http://cookbooks.dev/devel/recipes/1

If we get a representation of the same resource from a different web
service version, that representation is stored separately.

    >>> ignored = webservice.get(recipe_url, api_version="1.0")
    >>> for key in sorted(dictionary.keys()):
    ...     print key
    http://cookbooks.dev/1.0/recipes/1,application/json
    http://cookbooks.dev/devel/recipes/1,application/json

    >>> key1 = "http://cookbooks.dev/1.0/recipes/1,application/json"
    >>> key2 = "http://cookbooks.dev/devel/recipes/1,application/json"
    >>> dictionary[key1] == dictionary[key2]
    False

Cache invalidation
==================

lazr.restful automatically invalidates the representation
cache when certain changes happen that it knows about.

For instance, when the client sends a PATCH request to modify an
object, the modified representation is returned in preference to the
previously cached representation.

    >>> old_instructions = webservice.get(
    ...     recipe_url, api_version='devel').jsonBody()['instructions']
    >>> print old_instructions
    You can always judge...
    >>> response = webservice.patch(recipe_url, 'application/json',
    ...     simplejson.dumps(dict(instructions="New instructions")),
    ...     api_version='devel')
    >>> print response.status
    209
    >>> print response.jsonBody()['instructions']
    New instructions

The modified representation is immediately available in the cache.

    >>> from lazr.restful.example.base.root import RECIPES
    >>> recipe = [recipe for recipe in RECIPES if recipe.id == 1][0]
    >>> cached_representation = cache.get(recipe, json, 'devel')
    >>> print simplejson.loads(cached_representation)['instructions']
    New instructions

Cleanup.

    >>> recipe.instructions = old_instructions

When the client invokes a named operation using POST, the object is
always removed from the cache, because there's no way to know what the
POST did.

    >>> dictionary.clear()

    >>> from lazr.restful.example.base.root import COOKBOOKS
    >>> cookbook = [cookbook for cookbook in COOKBOOKS
    ...             if cookbook.name == "Everyday Greens"][0]
    >>> cache.set(cookbook, json, 'devel', "Dummy value.")
    >>> print dictionary.keys()[0]
    http://.../devel/cookbooks/Everyday%20Greens,application/json

    >>> from urllib import quote
    >>> greens_url = quote("/cookbooks/Everyday Greens")
    >>> ignore = webservice.named_post(
    ...     greens_url, "replace_cover", cover="foo")
    >>> print len(dictionary.keys())
    0

But unless the web service is the only way to manipulate a data set,
you'll also need to come up with your own cache invalidation
rules. Without those rules, other kinds of changes won't trigger cache
invalidations, and your web service cache will grow stale.

Let's signal a change to a recipe. Suppose someone changed that
recipe, using a web application that has no connection to the web
service except for a shared database. Let's further suppose that our
ORM lets us detect the database change. What do we do when that change
happens?

To remove an object's representation from the cache, we pass it into
the cache's delete() method.

    >>> ignore = webservice.get(recipe_url, api_version='devel')
    >>> print cache.get(recipe, json, 'devel')
    {...}
    >>> cache.delete(recipe)

This deletes all the relevant representations.

    >>> print cache.get(recipe, json, 'devel')
    None
    >>> dictionary.keys()
    []

Data visibility
===============

Only full representations are added to the cache. If the
representation you request includes a redacted field (because you
don't have permission to see that field's true value), the
representation is not added to the cache.

    >>> greens = webservice.get(greens_url).jsonBody()
    >>> print greens['confirmed']
    tag:launchpad.net:2008:redacted

    >>> dictionary.keys()
    []

This means that if your entry resources typically contain data that's
only visible to a select few users, you won't get much benefit out of
a representation cache.

What if a full representation is in the cache, and the user requests a
representation that must be redacted? Let's put some semi-fake data in
the cache and find out.

    >>> import simplejson
    >>> greens['name'] = "This comes from the cache; it is not generated."
    >>> greens['confirmed'] = True
    >>> cache.set(greens_object, json, 'devel', simplejson.dumps(greens))

When we GET the corresponding resource, we get a representation that
definitely comes from the cache, not the original data source.

    >>> cached_greens = webservice.get(greens_url).jsonBody()
    >>> print cached_greens['name']
    This comes from the cache; it is not generated.

But the redacted value is still redacted.

    >>> print cached_greens['confirmed']
    tag:launchpad.net:2008:redacted

Cleanup: clear the cache.

    >>> dictionary.clear()

Unauthorized
============

When a client tries to fetch an object they lack permission to view,
they get a 401 error.

    >>> recipe_url ="/recipes/5"
    >>> response = webservice.get(recipe_url)
    >>> print response.status
    401
    >>> len(dictionary.keys())
    0

This happens even if the forbidden object has a cached representation.
To demonstrate this, we'll temporarily make recipe #5 public and get
its representation.

    >>> from lazr.restful.example.base.root import C3_D2 as recipe
    >>> recipe.private = False

    >>> print webservice.get("/recipes/5").jsonBody()['instructions']
    Without doubt the most famous...

Now there's a representation in the cache.

    >>> len(dictionary.keys())
    1

If we make the recipe private again, the client can no longer retrieve
a representation, even though there's one in the cache. (This happens
when an object's URL construction code raises an Unauthorized
exception.)

    >>> recipe.private = True
    >>> response = webservice.get(recipe_url)
    >>> print response.status
    401

Cleanup: clear the cache.

    >>> dictionary.clear()

Collections
===========

Collections are full of entries, and representations of collections
are built from the cache if possible. We'll demonstrate this with the
collection of recipes.

First, we'll hack the cached representation of a single recipe.

    >>> recipe = webservice.get("/recipes/1").jsonBody()
    >>> recipe['instructions'] = "This representation is from the cache."
    >>> [recipe_key] = dictionary.keys()
    >>> dictionary[recipe_key] = simplejson.dumps(recipe)

Now, we get the collection of recipes.

    >>> recipes = webservice.get("/recipes").jsonBody()['entries']

The fake instructions we put into an entry's cached representation are
also present in the collection.

    >>> for instructions in (
    ...     sorted(recipe['instructions'] for recipe in recipes)):
    ...     print instructions
    A perfectly roasted chicken is...
    Draw, singe, stuff, and truss...
    ...
    This representation is from the cache.

To build the collection, lazr.restful had to generate representations
of all the cookbook entries. As it generated each representation, it
populated the cache.

    >>> for key in sorted(dictionary.keys()):
    ...     print key
    http://cookbooks.dev/devel/recipes/1,application/json
    http://cookbooks.dev/devel/recipes/2,application/json
    http://cookbooks.dev/devel/recipes/3,application/json
    http://cookbooks.dev/devel/recipes/4,application/json

If we request the collection again, all the entry representations will
come from the cache.

    >>> for key in dictionary.keys():
    ...     value = simplejson.loads(dictionary[key])
    ...     value['instructions'] = "This representation is from the cache."
    ...     dictionary[key] = simplejson.dumps(value)

    >>> recipes = webservice.get("/recipes").jsonBody()['entries']
    >>> for instructions in (
    ...     sorted(recipe['instructions'] for recipe in recipes)):
    ...     print instructions
    This representation is from the cache.
    This representation is from the cache.
    This representation is from the cache.
    This representation is from the cache.

enable_server_side_representation_cache
=======================================

A configuration setting allows you to disable the cache without
de-registering it. This is useful when you're debugging a cache
implementation or a cache invalidation algorithm.

    >>> from lazr.restful.interfaces import IWebServiceConfiguration
    >>> config = getUtility(IWebServiceConfiguration)
    >>> print config.enable_server_side_representation_cache
    True

Set this configuration value to False, and representations will not be
served from the cache, even if present.

    >>> config.enable_server_side_representation_cache = False
    >>> recipes = webservice.get("/recipes").jsonBody()['entries']
    >>> for instructions in (
    ...     sorted(recipe['instructions'] for recipe in recipes)):
    ...     print instructions
    A perfectly roasted chicken is...
    Draw, singe, stuff, and truss...
    Preheat oven to...
    You can always judge...

New representations will not be added to the cache.

    >>> before_keys = len(dictionary.keys())
    >>> dishes = webservice.get("/dishes")
    >>> len(dictionary.keys()) == before_keys
    True

The cache is still registered as a utility.

    >>> print getUtility(IRepresentationCache)
    <lazr.restful.simple.DictionaryBasedRepresentationCache object ...>

And it's still populated, as we can see by re-enabling it.

    >>> config.enable_server_side_representation_cache = True
    >>> recipes = webservice.get("/recipes").jsonBody()['entries']
    >>> for instructions in (
    ...     sorted(recipe['instructions'] for recipe in recipes)):
    ...     print instructions
    This representation is from the cache.
    This representation is from the cache.
    This representation is from the cache.
    This representation is from the cache.

Once re-enabled, new representations are once again added to the cache.

    >>> dishes = webservice.get("/dishes")
    >>> len(dictionary.keys()) > before_keys
    True

Cleanup
=======

De-register the cache.

    >>> sm.registerUtility(None, IRepresentationCache)

Of course, the hacks we made to the cached representations have no
effect on the objects themselves. Once the hacked cache is gone, the
representations look just as they did before.

    >>> recipes = webservice.get("/recipes").jsonBody()['entries']
    >>> for instructions in (
    ...     sorted(recipe['instructions'] for recipe in recipes)):
    ...     print instructions
    A perfectly roasted chicken is...
    Draw, singe, stuff, and truss...
    Preheat oven to...
    You can always judge...