20.11 Sitemaps: Sitemap and the sitemap views

What a sitemap says of each item

Sitemap.items returns the objects to list, a list or a queryset, by default none. Sitemap.location, lastmod, changefreq and priority each say one thing of an item, as a value for all or a method that takes the item. Sitemap._get gives None for one the sitemap lacks, calls a callable with the item, and otherwise returns the value. Of the four, the base class defines only location, the item’s get_absolute_url, a path; the template writes the other three only where the item has a value for them.

Sitemap._urls makes a dictionary for each item of the page asked for, with the keys "item", "location", "lastmod", "changefreq", "priority" and "alternates"; the location joins the protocol, the domain and the path, and the priority is a string, empty for none. The list is the template’s "urlset"; a project’s template finds each object at "item", paired with a language code where each language is listed. _urls also stores the page’s latest lastmod on the sitemap, where the sitemap view reads it as "latest_lastmod", but only if every item of the page had one.

Pages and the domain

Sitemap.limit is 50,000, a limit its comment calls Google’s, citing sitemaps.org. Sitemap.paginator makes a new Paginator of the items, limit to a page, each time it is read, calling items again. The paginator warns of a queryset with no ordering: a GenericSitemap of every lighthouse raised UnorderedObjectListWarning (Pagination: Paginator and Page).

Sitemap.get_urls takes a page, a site and a protocol. The protocol is Sitemap.get_protocol’s: the sitemap’s own Sitemap.protocol, else the one passed, else "https". The domain is Sitemap.get_domain’s: the site passed, else the current site by SITE_ID; without the sites application or that row, ImproperlyConfigured, “To use sitemaps, either enable the sites framework or pass a Site/RequestSite object in your view.”. The sitemap view always passes the request’s site and scheme, so the "https" and the lookup serve code that builds a sitemap outside a request.

A URL for each language

With Sitemap.i18n true, each item is listed once for each of its languages, which are Sitemap.get_languages_for_item’s, by default Sitemap.languages or the codes of LANGUAGES. Sitemap._items makes the pairs of item and code before pagination, so that, its comment says, pages are cut with every language counted: limit counts URLs, not objects. _get hands a method the object alone, and Sitemap._location calls location inside django.utils.translation.override of the pair’s language, so under i18n_patterns reverse writes that language’s prefix (The active language, Language prefixes and translated routes).

With Sitemap.alternates also true, each URL links to the item in each of its languages, and with Sitemap.x_default once more, as "x-default", when LANGUAGE_CODE is among them: the URL in that language with the first path segment that is its code cut out. With the lighthouses’ URL pattern moved under i18n_patterns, LANGUAGES English and Maltese and LANGUAGE_CODE "en", the office lists them with:

# notes/sitemaps.py
from django.contrib.sitemaps import Sitemap

from notes.models import Lighthouse


class LighthouseSitemap(Sitemap):
    i18n = True
    alternates = True
    x_default = True

    def items(self):
        return Lighthouse.objects.order_by("name")

St Elmo’s two URLs, at /en/ and /mt/, carry the same three links; the first, one element to a line:

<url>
<loc>http://harbour.example/en/lighthouses/1/</loc>
<xhtml:link rel="alternate" hreflang="en" href="http://harbour.example/en/lighthouses/1/"/>
<xhtml:link rel="alternate" hreflang="mt" href="http://harbour.example/mt/lighthouses/1/"/>
<xhtml:link rel="alternate" hreflang="x-default" href="http://harbour.example/lighthouses/1/"/>
</url>

GenericSitemap

GenericSitemap is a sitemap made from a dictionary: "queryset" gives the items and an optional "date_field" names the field read as an item’s lastmod. priority, changefreq and protocol may be passed as arguments, and count only where the class’s own value is false, such as None. It goes into the dictionary as an instance, which lives as long as the URLconf. GenericSitemap.items returns a clone of the queryset, by QuerySet.filter with no arguments, against premature evaluation, its comment says: the queryset the instance keeps is never evaluated and holds no rows between requests.

The index asks each sitemap for Sitemap.get_latest_lastmod, which returns a lastmod that is a value, and calls one that is a method on every item and takes the greatest; None if it has none or some item’s is None. GenericSitemap.get_latest_lastmod, given a "date_field", asks the database for one value, and otherwise returns None. For the bulletins, the first selected every column of every bulletin; a GenericSitemap with "date_field" "published" sent SELECT "notes_bulletin"."published" AS "published" FROM "notes_bulletin" ORDER BY 1 DESC LIMIT 1.

The views: index and sitemap

Both views call a value of the dictionary that is callable, a class, to make a sitemap for the request, and use an instance as it is. Both take the domain from get_current_site and the protocol from the request’s scheme unless a sitemap sets its own (Sites: Site, get_current_site and the shortcut view), return a TemplateResponse of "sitemap_index.xml" or "sitemap.xml" by default, as "application/xml" (SimpleTemplateResponse and TemplateResponse), and are wrapped by x_robots_tag, which sets X-Robots-Tag to "noindex, noodp, noarchive".

index reverses the URL its sitemap_url_name argument names, by default "django.contrib.sitemaps.views.sitemap", with the section as "section", and makes a SitemapIndexItem of the absolute URL and the sitemap’s get_latest_lastmod; each page after the first adds the URL with "?p=" and the page’s number, and the same date. sitemap renders the section named, raising Http404 for a key the dictionary lacks, or, given none, every sitemap in one file. The page is the query string’s "p", 1 by default, asked of every sitemap rendered; the paginator’s EmptyPage and PageNotAnInteger become Http404.

Each view sends Last-Modified only when every sitemap it covers gives a date: index from each get_latest_lastmod, sitemap from the "latest_lastmod" that _urls left on the sitemap; an instance in the dictionary keeps it between requests, so a later page with an undated item carried an earlier page’s date. _get_latest_lastmod keeps the greatest, a date taken as midnight and a naive datetime as UTC.

With the bulletins’ sitemap given a limit of 2, the views gave these; the last three called the view directly:

/sitemap.xml 200
    <loc>http://harbour.example/sitemap-bulletins.xml</loc>
    <lastmod>2026-10-14T07:15:00+02:00</lastmod>
    <loc>http://harbour.example/sitemap-bulletins.xml?p=2</loc>
    <lastmod>2026-10-14T07:15:00+02:00</lastmod>
    <loc>http://harbour.example/sitemap-lighthouses.xml</loc>
    Last-Modified: None
/sitemap-bulletins.xml 200
    <loc>http://harbour.example/bulletins/3/</loc>
    <loc>http://harbour.example/bulletins/2/</loc>
    Last-Modified: Wed, 14 Oct 2026 05:15:00 GMT
/sitemap-bulletins.xml?p=2 200
    <loc>http://harbour.example/bulletins/1/</loc>
    Last-Modified: Mon, 12 Oct 2026 06:00:00 GMT
/sitemap-bulletins.xml?p=3     Http404: Page 3 empty
/sitemap-bulletins.xml?p=two   Http404: No page 'two'
/sitemap-ships.xml             Http404: No sitemap available for section: 'ships'

The lighthouses give no dates, so the index has no Last-Modified; in it both pages of the bulletins carry the sitemap’s latest date, while each file’s header carries its own page’s.

Flat pages and geometries

FlatPageSitemap lists the current site’s flat pages that require no login. It asks for the current site without the request, so needs SITE_ID even where the views find the site by host, and raises ImproperlyConfigured without the sites application (Flat pages and redirects: the fallback middleware). GeoDjango’s KMLSitemap and KMZSitemap list geometry fields, at views that write KML (GeoDjango’s forms, admin, feeds and data tools).