Merge pull request #2196 from mosra/absolute-url-merging

Make URL part joining aware of absolute URLs

pelican/contents.py

@@ -11,7 +11,7 @@ import sys
 import pytz
 
 import six
-from six.moves.urllib.parse import urlparse, urlunparse
+from six.moves.urllib.parse import urljoin, urlparse, urlunparse
 
 from pelican import signals
 from pelican.settings import DEFAULT_CONFIG
@@ -228,35 +228,31 @@ class Content(object):
         key = key if self.in_default_lang else 'lang_%s' % key
         return self._expand_settings(key)
 
-    def _update_content(self, content, siteurl):
+    def _link_replacer(self, siteurl, m):
-        """Update the content attribute.
-
-        Change all the relative paths of the content to relative paths
-        suitable for the output content.
-
-        :param content: content resource that will be passed to the templates.
-        :param siteurl: siteurl which is locally generated by the writer in
-                        case of RELATIVE_URLS.
-        """
-        if not content:
-            return content
-
-        instrasite_link_regex = self.settings['INTRASITE_LINK_REGEX']
-        regex = r"""
-            (?P<markup><[^\>]+  # match tag with all url-value attributes
-                (?:href|src|poster|data|cite|formaction|action)\s*=\s*)
-
-            (?P<quote>["\'])      # require value to be quoted
-            (?P<path>{0}(?P<value>.*?))  # the url value
-            \2""".format(instrasite_link_regex)
-        hrefs = re.compile(regex, re.X)
-
-        def replacer(m):
         what = m.group('what')
         value = urlparse(m.group('value'))
         path = value.path
         origin = m.group('path')
 
+        # urllib.parse.urljoin() produces `a.html` for urljoin("..", "a.html")
+        # so if RELATIVE_URLS are enabled, we fall back to os.path.join() to
+        # properly get `../a.html`. However, os.path.join() produces
+        # `baz/http://foo/bar.html` for join("baz", "http://foo/bar.html")
+        # instead of correct "http://foo/bar.html", so one has to pick a side
+        # as there is no silver bullet.
+        if self.settings['RELATIVE_URLS']:
+            joiner = os.path.join
+        else:
+            joiner = urljoin
+
+            # However, it's not *that* simple: urljoin("blog", "index.html")
+            # produces just `index.html` instead of `blog/index.html` (unlike
+            # os.path.join()), so in order to get a correct answer one needs to
+            # append a trailing slash to siteurl in that case. This also makes
+            # the new behavior fully compatible with Pelican 3.7.1.
+            if not siteurl.endswith('/'):
+                siteurl += '/'
+
         # XXX Put this in a different location.
         if what in {'filename', 'attach'}:
             if path.startswith('/'):
@@ -283,7 +279,7 @@ class Content(object):
                             "%s used {attach} link syntax on a "
                             "non-static file. Use {filename} instead.",
                             self.get_relative_source_path())
-                    origin = '/'.join((siteurl, linked_content.url))
+                origin = joiner(siteurl, linked_content.url)
                 origin = origin.replace('\\', '/')  # for Windows paths.
             else:
                 logger.warning(
@@ -292,13 +288,13 @@ class Content(object):
                         'limit_msg': ("Other resources were not found "
                                       "and their urls not replaced")})
         elif what == 'category':
-                origin = '/'.join((siteurl, Category(path, self.settings).url))
+            origin = joiner(siteurl, Category(path, self.settings).url)
         elif what == 'tag':
-                origin = '/'.join((siteurl, Tag(path, self.settings).url))
+            origin = joiner(siteurl, Tag(path, self.settings).url)
         elif what == 'index':
-                origin = '/'.join((siteurl, self.settings['INDEX_SAVE_AS']))
+            origin = joiner(siteurl, self.settings['INDEX_SAVE_AS'])
         elif what == 'author':
-                origin = '/'.join((siteurl, Author(path, self.settings).url))
+            origin = joiner(siteurl, Author(path, self.settings).url)
         else:
             logger.warning(
                 "Replacement Indicator '%s' not recognized, "
@@ -313,7 +309,30 @@ class Content(object):
         return ''.join((m.group('markup'), m.group('quote'), origin,
                         m.group('quote')))
 
-        return hrefs.sub(replacer, content)
+    def _update_content(self, content, siteurl):
+        """Update the content attribute.
+
+        Change all the relative paths of the content to relative paths
+        suitable for the output content.
+
+        :param content: content resource that will be passed to the templates.
+        :param siteurl: siteurl which is locally generated by the writer in
+                        case of RELATIVE_URLS.
+        """
+        if not content:
+            return content
+
+        instrasite_link_regex = self.settings['INTRASITE_LINK_REGEX']
+        regex = r"""
+            (?P<markup><[^\>]+  # match tag with all url-value attributes
+                (?:href|src|poster|data|cite|formaction|action)\s*=\s*)
+
+            (?P<quote>["\'])      # require value to be quoted
+            (?P<path>{0}(?P<value>.*?))  # the url value
+            \2""".format(instrasite_link_regex)
+        hrefs = re.compile(regex, re.X)
+
+        return hrefs.sub(lambda m: self._link_replacer(siteurl, m), content)
 
     def get_siteurl(self):
         return self._context.get('localsiteurl', '')

pelican/tests/test_contents.py

@@ -397,6 +397,54 @@ class TestPage(LoggedTestCase):
             '</blockquote>'
         )
 
+    def test_intrasite_link_absolute(self):
+        """Test that absolute URLs are merged properly."""
+
+        args = self.page_kwargs.copy()
+        args['settings'] = get_settings(
+            STATIC_URL='http://static.cool.site/{path}',
+            ARTICLE_URL='http://blog.cool.site/{slug}.html')
+        args['source_path'] = 'content'
+        args['context']['filenames'] = {
+            'images/poster.jpg': Static('',
+                                        settings=args['settings'],
+                                        source_path='images/poster.jpg'),
+            'article.rst': Article('',
+                                   settings=args['settings'],
+                                   metadata={'slug': 'article',
+                                             'title': 'Article'})
+        }
+
+        # Article link will go to blog
+        args['content'] = (
+            '<a href="{filename}article.rst">Article</a>'
+        )
+        content = Page(**args).get_content('http://cool.site')
+        self.assertEqual(
+            content,
+            '<a href="http://blog.cool.site/article.html">Article</a>'
+        )
+
+        # Page link will go to the main site
+        args['content'] = (
+            '<a href="{index}">Index</a>'
+        )
+        content = Page(**args).get_content('http://cool.site')
+        self.assertEqual(
+            content,
+            '<a href="http://cool.site/index.html">Index</a>'
+        )
+
+        # Image link will go to static
+        args['content'] = (
+            '<img src="{filename}/images/poster.jpg"/>'
+        )
+        content = Page(**args).get_content('http://cool.site')
+        self.assertEqual(
+            content,
+            '<img src="http://static.cool.site/images/poster.jpg"/>'
+        )
+
     def test_intrasite_link_markdown_spaces(self):
         # Markdown introduces %20 instead of spaces, this tests that
         # we support markdown doing this.