I ll describe here a 'non-perfect but suitable' solution to make a tag cloud, like the one i use for this blog, with django.
Principle
We are going to use a many to many field with link a blog anetry to many tags. It will create three tables blog_entry, blog_entry_tags and blog_tag. The tags need some field to remember the number of blog entries they appears in, as well as there current font size for the tag cloud.
The algorithm to distributes n elements between b buckets, keeping a uniform distribution without having tags with same number of references in different buckets, was the tricky part. You can find discussion about font distribution alogorithm around. one here.
Related items (the things being saved into the many-to-many relation) are not saved as part of a model's save method, as you have discovered. Instead, the Add- and ChangeManipulators save the many-to-many items later. In fact, for adding a new item, this is basically required, because you need to know the new instance's primary key value before you can save a reference to it in the m2m join table -- and that value does not necessarily exist before it is saved to the database.Malcolm Tredinnick link.
An important point: When do we do this processing ? at every blog view ? in a post_save method for our blog Entry model? The problem comes from many to many relation object saving, the post_save method for entry still happen before the related tags in blog_tag and blog_entry_tags are saved, see the quote for more details.
Since i dont see how to overcome the many to many saving process and i dont want to recalculate the cloud tag values at every view, i will use a post_save dispatcher which basicaly acts like a post_save model method.
Here comes the dodgy solution. On the first save of your entry (which you do by clicking save in the edit/create form for entry objects), the dispatcher will connect to your tag_cloud generation method. But, since the tags arent saved yet, you need to save your entry twice. On the second time the tags are there and the tag cloud generation method will update the entry tags total references and font-size field. Hope you got it, my explainations are not famous for their clearness.
In order to update the tag cloud we need to save the entry object twice.
Models
model.py file for my blog application.
class Tag(models.Model):
name = models.CharField(maxlength=200, primary_key='True', core=True)
total_ref = models.IntegerField(blank=True, default=0)
font_size = models.IntegerField(blank=True, default=0)
def __str__(self):
return self.name
def get_absolute_url(self):
return "/blog/tag/%s/" % (self.name)
def __cmp__(self, other):
return cmp(self.total_ref, other.total_ref)
class Entry(models.Model):
[...]
tags = models.ManyToManyField(Tag)
Dispatcher
In our blog/view.py we can now connect the dispatcher to the tag cloud generation method. Tweak the final result using *nbr_of_buckets and base_font_size.
[...]
from coulix_org.blog.models import Entry, Tag
from django.db.models import signals
from django.dispatch import dispatcher
def process_cloud_tag(instance):
''' distribution algo n tags to b bucket, where b represents
font size. '''
entry = instance
# be sure you save twice the same entry, otherwise it wont update the new tags.
entry_tag_list = entry.tags.all()
for tag in entry_tag_list:
tag.total_ref = tag.entry_set.all().count();
tag.save()
tag_list = Tag.objects.all()
nbr_of_buckets = 8
base_font_size = 11
tresholds = []
max_tag = max(tag_list)
min_tag = min(tag_list)
delta = (float(max_tag.total_ref) - float(min_tag.total_ref)) / (float(nbr_of_buckets))
# set a treshold for all buckets
for i in range(nbr_of_buckets):
tresh_value = float(min_tag.total_ref) + (i+1) * delta
tresholds.append(tresh_value)
# set font size for tags (per bucket)
for tag in tag_list:
font_set_flag = False
for bucket in range(nbr_of_buckets):
if font_set_flag == False:
if (tag.total_ref <= tresholds[bucket]):
tag.font_size = base_font_size + bucket * 2
tag.save()
font_set_flag = True
# connect signal
dispatcher.connect(process_cloud_tag,
sender = Entry,
signal = signals.post_save)
Templatetags
We use a template tag to return a list of all tags to our template. Add the code to blog/templatetags/whatever.py file.
from coulix_org.blog.models import Tag
register = template.Library()
# use for tag cloud
def show_tag_list(parser, token):
""" {% get_tag_list %}"""
return TagListObject()
class TagListObject(template.Node):
def render(self, context):
context['blog_tags'] = Tag.objects.all()
return ''
register.tag('get_tag_list', show_tag_list)
Template
Here comes the template which generates the proper html code to display a nice cloud tag.
{% load whatever %}
{% get_tag_list %}
{% for tag in blog_tags %}
<span style="font-size: {{ tag.font_size }}px;">
<a class="link-typeA" title="Number of entries: {{ tag.total_ref }}" href="/blog/tags/all/{{ tag.name }}">{{ tag.name }}</a>
</span>
{% endfor %}
Currently it only uses font-size, i may add some color gradient later. Please comment and propose better solutions than the two step saving dodgy method :).
8:27 a.m.
Maybe Django signals can be used to preserve the Tag Cloud in the Cache
8:49 a.m.
That's what happening with dispatcher.connect(process_cloud_tag, Isn't it what you mean ?
12:16 a.m.
I added post_delete dispatcher to delete tags not in use:
tag_list = Tag.objects.all() for tag in tag_list: total = tag.entry_set.all().count() if total == 0: tag.delete() else: tag.total_ref = total tag.save()
tag_list = Tag.objects.all() nbr_of_buckets = 8 ...
Excellent work!
5:45 a.m.
class Tag(Model): def save(self): Model.save(self) # I think that you could also shove all that stuff here rather than using signals Model.save(self)8:06 a.m.
Just Curious - Will signals works if the sender and receiver are in two different machines accessing the same database?
11:13 a.m.
No idea, a good question for #django :)
10:59 a.m.
signals are process-local
this code will result in invalid data given a fitting race-condition
7:03 a.m.
very nice tutorial. i've been use it for my pages
6:31 p.m.
Sorry, I think another way!
7:55 p.m.
what other way ? bad ? :/