django-adv-cache-tag 0.2.0

Versioning

{% load adv_cache %}
{% cache 0 myobj_main_template obj.pk obj.date_last_updated %}
  {{ obj }}
{% endcache %}

Primary key

:1:template.cache.your_fragment_name.64223ccf70bbb65a3a4aceac37e21016

:1:template.cache.your_fragment_name.your_pk.64223ccf70bbb65a3a4aceac37e21016

{% cache 0 myobj_main_template obj.pk obj.date_last_updated %}

Compression

Choose your cache backend

Partial caching

With the default django cache templatetag, your templates are cached and you can’t update them before display, so you can’t cache big parts of html with a little dynamic fragment in it, for the user name, the current date or whatever. You can cheat and save two templates surrounding your dynamic part, but you will have more accesses to your cache backend.

django-adv-cache-tag allow the use of one or many {% nocache %} blocks (closed by {% endnocache %}) to put in your {% cache %} blocks. These {% nocache %} block will be saved “as is” in the cache, while the rest of the block will be rendered to html. It’s only when the template is finally displayed that the no-cached parts will be rendered.

You can have as many of these blocks you want.

{% cache 0 myobj_main_template obj.pk obj.date_last_updated %}
    <p>This is the cached part of the template for {{ obj }}, evaluated at {% now "r" %}.</p>
    {% nocache %}
        <p>This part will be evaluated each time : {% now "r" %}</p>
    {% endnocache %}
    <p>This is another cached part</p>
{% endcache %}

The fragment name

{% cache 0 fragment_name obj.pk obj.date_last_updated %}

{% cache 0 "myobj_main_template" obj.pk obj.date_last_updated %}

{% cache 0 "myobj_main_template" obj.pk obj.date_last_updated %}

Basic override

Imagine you don’t want to change the default settings (all to False, and using the default backend) but want a templatetag with versioning activated :

Create a new templatetag file (myapp/templatetags/my_cache_tags.py) with this:

from adv_cache_tag.tag import CacheTag

class VersionedCacheTag(CacheTag):
    class Meta(CacheTag.Meta):
        versioning = True

from django import template
register = template.Library()

VersionedCacheTag.register(register, 'ver_cache')

With these simple lines, you now have a new templatetag to use when you want versioning:

{% load my_cache_tags %}
{% ver_cache 0 myobj_main_template obj.pk obj.date_last_updated %}
    obj
{% endver_cache %}

As you see, just replace {% load adv_cache %} (or the django default {% load cache %}) by {% load my_cache_tags %} (your templatetag module), and the {% cache %} templatetag by your new defined one, {% ver_cache ... %}. Don’t forget to replace the closing tag too: {% endver_cache %}. But the {% nocache %} will stay the same, except if you want a new one. For this, just add a parameter to the register method:

MyCacheTag.register(register, 'ver_cache', 'ver_nocache')

{% ver_cache ... %}
    cached
    {% ver_nocache %}not cached{% endver_nocache %}
{% endver_cache %}

Note that you can keep the name cache for your tag if you know that you will not load in your template another templatetag module providing a cache tag. To do so, the simplest way is:

MyCacheTag.register(register)  # 'cache' and 'nocache' are the default values

All the django-adv-cache-tag settings have a matching variable in the Meta class, so you can override one or many of them in your own classes. See the “Settings” part to see them.

Internal version

When your template file is updated, the only way to invalidate all cached versions of this template is to update the fragment name or the arguments passed to the templatetag.

With django-adv-cache-tag you can do this with versioning, by managing your own version as the last argument to the templatetag. But if you want to use the power of the versioning system of django-adv-cache-tag, it can be too verbose:

{% load adv_cache %}
{% with template_version=obj.date_last_updated|stringformat:"s"|add:"v1" %}
    {% cache 0 myobj_main_template obj.pk template_version %}
    ...
    {% endcache %}
{% endwith %}

django-adv-cache-tag provides a way to do this easily, with the ADV_CACHE_VERSION setting. But by updating it, all cached versions will be invalidated, not only those you updated.

To do this, simply create your own tag with a specific internal version:

class MyCacheTag(CacheTag):
    class Meta(CacheTag.Meta):
       internal_version = "v1"

MyCacheTag.register('my_cache')

And then in your template, you can simply do

{% load my_cache_tags %}
{% my_cache 0 myobj_main_template obj.pk obj.date_last_updated %}
...
{% endmy_cache %}

Each time you update the content of your template and want invalidation, simply change the internal_version in your MyCacheTag class (or you can use a settings for this).

Change the cache backend

If you want to change the cache backend for one templatetag, it’s easy:

class MyCacheTag(CacheTag):
    class Meta:
        cache_backend = 'templates'

But you can also to this by overriding a method:

from django.core.cache import get_cache

class MyCacheTag(CacheTag):
    def get_cache_object(self):
        return get_cache('templates')

And if you want a cache backend for old objects, and another, faster, for recent ones:

from django.core.cache import get_cache

class MyCacheTag(CacheTag):
    class Meta:
        cache_backend = 'fast_templates'

    def get_cache_object(self):
        cache_backend = self.options.cache_backend
        if self.get_pk() < 1000:
            cache_backend = 'slow_templates'
        return get_cache(cache_backend)

The value returned by the get_cache_object should be a cache backend object, but as we only use the set and get methods on this object, it can be what you want if it provides these two methods. And even more, you can override the cache_set and cache_get methods of the CacheTag class if you don’t want to use the default set and get methods of the cache backend object.

Note that we also support the django way of changing the cache backend in the template-tag, using the using argument, to be set at the last parameter (without any space between using and the name of the cache backend).

{% cache 0 myobj_main_template obj.pk obj.date_last_updated using=foo %}

Change the cache key

The CacheTag class provides three classes to create the cache key:

• get_base_cache_key, which returns a formattable string (“template.%(nodename)s.%(name)s.%(pk)s.%(hash)s” by default if include_pk is True or “template.%(nodename)s.%(name)s.%(hash)s” if False

• get_cache_key_args, which returns the arguments to use in the previous string

• get_cache_key, which combine the two

The arguments are:

• nodename parameter is the name of the templatetag: it’s “my_cache” in {% my_cache ... %}

• name is the “fragment name” of your templatetag, the value after the expire-time

• pk is used only if self.options.include_pk is True, and is returned by this.get_pk()

• hash is the hash of all arguments after the fragment name, excluding the last one which is the version number (this exclusion occurs only if self.options.versioning is True)

If you want to remove the “template.” part at the start of the cache key (useless if you have a cache backend dedicated to template caching), you can do this:

class MyCacheTag(CacheTag):
    def get_base_cache_key(self):
        cache_key = super(MyCacheTag, self).get_base_cache_key()
        return cache_key[len('template:'):]  # or [9:]

Add an argument to the templatetag

By default, the templatetag provided by CacheTag takes the same arguments as the default django cache templatetag.

If you want to add an argument, it’s easy as the class provides a get_template_node_arguments method, which will work as for normal django templatetags, taking a list of tokens, and returning ones that will be passed to the real templatetag, a Node class tied to the CacheTag.

Say you want to add a foo argument between the expire time and the fragment name:

from django import template

from adv_cache_tag.tag import CacheTag, Node

class MyNode(Node):
    def __init__(self, nodename, nodelist, expire_time, foo, fragment_name, vary_on):
        """ Save the foo variable in the node (not resolved yet) """
        super(MyNode, self).__init__(self, nodename, nodelist, expire_time, fragment_name, vary_on)
        self.foo = foo


class MyCacheTag(CacheTag):

    Node = MyNode

    def prepare_params(self):
        """ Resolve the foo variable to it's real content """
        super(MyCacheTag, self).prepare_params()
        self.foo = template.Variable(self.node.foo).resolve(self.context)

    @classmethod
    def get_template_node_arguments(cls, tokens):
        """ Check validity of tokens and return them as ready to be passed to the Node class """
        if len(tokens) < 4:
            raise template.TemplateSyntaxError(u"'%r' tag requires at least 3 arguments." % tokens[0])
        return (tokens[1], tokens[2], tokens[3], tokens[4:])

Prepare caching of templates

This one is not about overriding the class, but it can be useful. When an object is updated, it can be better to regenerate the cached template at this moment rather than when we need to display it.

It’s easy. You can do this by catching the post_save signal of your model, or just by overriding its save method. For this example we will use this last solution.

The only special thing is to know the path of the template where your templatetag is. In my case, i have a template just for this (included in other ones for general use), so it’s easier to find it and regenerate it as in this example.

As we are not in a request, we don’t have the Request object here, so context processors are not working, we must create a context object that will be used to render the template, with all needed variables.

from django.template import loader, Context

class MyModel(models.Model):
    # your fields

    def save(self, *args, **kwargs):
        super(MyModel, self.save(*args, **kwargs)

        template = 'path/to/my_template_file_with_my_cache_block.html'

        context = Context({
            'obj': self,

            # as you have no request, we have to add stuff from context processors manually if we need them
            'STATIC_URL': settings.STATIC_URL,

            # the line below indicates that we force regenerating the cache, even if it exists
            '__regenerate__': True,

            # the line below indicates if we only want html, without parsing the nocache parts
            '__partial__': True,

        })

        loader.get_template(template).render(context)

Load data from database before rendering

This is a special case. Say you want to display a list of objects but you have only ids and versions retrieved from redis (with ZSET, with id as value and updated date (which is used as a version) as score , for example)

If you know you always have a valid version of your template in cache, because they are regenerated every time they are saved, as seen above, it’s fine, just add the object’s primary key as the pk in your templatetag arguments, and the cached template will be loaded.

But if it’s not the case, you will have a problem: when django will render the template, the only part of the object present in the context is the primary key, so if you need the name or whatever field to render the cached template, it won’t work.

With django-adv-cache-tag it’s easy to resolve this, as we can load the object from the database and adding it to the context.

def my_view(request):
    objects = [
        dict(
            pk=val[0],
            date_last_updated=val[1]
        )
        for val in
            redis.zrevrange('my_objects', 0, 19, withscores=True)
    ]
    return render(request, "my_results.html", dict(objects=objects))

{% for obj in objects %}
    {% include "my_result.html" %}
{% endfor %}

{% load my_cache_tags %}
{% my_cache 0 myobj_main_template obj.pk obj.date_last_updated %}
    {{ obj }}
{% endmy_cache %}

from my_app.models import MyModel

class MyCacheTag(CacheTag):

    class Meta(CacheTag.Meta):
        """ Force options """
        include_pk = True
        versioning = True

    def create_content(self):
        """ If the object in context is not a real model, load it from db """
        if not isinstance(context['obj'], MyObject):
            context['obj'] = MyModel.objects.get(id=self.get_pk())
        super(MyCacheTag, self).create_content()

MyCacheTag.register('my_cache')

And more…

If you want to do more, feel free to look at the source code of the CacheTag class (in tag.py), all methods are documented.

Partial caching

Your template :

{% load adv_cache %}
{% cache ... %}
    foo
    {% nocache %}
        bar
    {% endnocache %}
    baz
{% endcache %}

Cached version (we ignore versioning and compress here, just to see how it works):

foo
{% endRAW_xyz %}
    bar
{% RAW_xyz %}
baz

When cached version is loaded, we parse :

{% RAW_xyz %}
foo
{% endRAW_xyz %}
    bar
{% RAW_xyz %}
baz
{% endRAW_xyz %}

The first {% RAW_xyz %} and the last {% endRAW_xyz %} are not included in the cached version and added before parsing, only to save some bytes.

Parts between {% RAW_xyz %} and {% endRAW_xyz %} are not parsed at all (seen as a TextNode by django)

The xyz part of the RAW and endRAW templatetags depends on the SECRET_KEY and so is unique for a given site.

It allows to avoid at max the possible collisions with parsed content in the cached version.

We could have used {% nocache %} and {% endnocache %} instead of {% RAW_xyz %} and {% endRAW_xyz %} but in the parsed template, stored in the cache, if the html includes one of these strings, our final template would be broken, so we use long ones with a hash (but we can not be sure at 100% these strings could not be in the cached html, but for common usages it should suffice)