Document changing host for rotating-url and assume same version based only on checksum and file size

README.md

@@ -205,6 +205,22 @@ The version number for the latest version can be detected in two ways:
   URL for the latest version, and the first match group is taken to be the
   version. (This follows the convention used by
   [`debian/watch`](https://wiki.debian.org/debian/watch) files.)
+  * If the application download is behind load-balanced URL that changes regularly
+    (e.g. `https://stable.dl2.example.com` and `dl.example.com`),
+    the regex needs to be adjusted to extract the version number in all cases.
+    Otherwise this project assumes a new version was published.
+  * You can use a non-capturing group for this use-case. For the use case above, use
+    the pattern below to allow downloads from:
+    * `https://dl.example.com/foo-v1.9.tar.gz`
+    * `https://stable.dl1.example.com/foo-v1.9.tar.gz`
+    * ...
+```json
+"x-checker-data": {
+    "type": "rotating-url",
+    "url": "http://example.com/last-version",
+    "pattern": "https://(?:dl|stable.dl\\d).example.com/foo-v([0-9.]+).tar.gz"
+}
+```
 
 Some upstream vendors may add unwanted GET query parameters to
 the download URL, such as identifiers for counting unique downloads.

src/checkers/urlchecker.py

@@ -61,6 +61,26 @@ def extract_version(checker_data, url):
     return m.group(1)
 
 
+def is_same_version(checker_data, current_url, new_version):
+    """
+    Check if the new application version is the same with the current one. If the
+    version number could be extracted, those strings are compared with each other
+    to be resilient against load-balanced urls pointing to the same file.
+    """
+    if new_version.version is None:
+        # No pattern given or failed parsing the new version, so check only
+        # if the url is different
+        return current_url == new_version.url
+
+    current_version_string = extract_version(checker_data, current_url)
+    if current_version_string is None:
+        # If the pattern failed to apply to the old/current version,
+        # check again only the url. Otherwise we would compare None values
+        return current_url == new_version.url
+
+    return current_version_string == new_version.version
+
+
 class URLChecker(Checker):
     PRIORITY = 99
     CHECKER_DATA_TYPE = "rotating-url"
@@ -138,9 +158,10 @@ async def check(self, external_data: ExternalBase):
         if not is_rotating:
             new_version = new_version._replace(url=url)  # pylint: disable=no-member
 
+        same_version = is_same_version(
+            external_data.checker_data, external_data.current_version.url, new_version
+        )
         external_data.set_new_version(
             new_version,
-            is_update=(
-                is_rotating and external_data.current_version.url != new_version.url
-            ),
+            is_update=is_rotating and not same_version,
         )

src/lib/externaldata.py

@@ -319,10 +319,8 @@ def json(self) -> t.Dict[str, t.Any]:
     def matches(self, other: ExternalFile):
         for i in (self, other):
             assert i.checksum is None or isinstance(i.checksum, MultiDigest), i.checksum
-        return (
-            self.url == other.url
-            and self.checksum == other.checksum
-            and (self.size is None or other.size is None or self.size == other.size)
+        return self.checksum == other.checksum and (
+            self.size is None or other.size is None or self.size == other.size
         )
 
     def is_same_version(self, other: ExternalFile):