Add PEP691 support

Author: chriskuehl

README.md

@@ -57,15 +57,21 @@ You can set these environment variables to configure the server:
 * `PYPI_BROWSER_PYPI_URL`: URL for the PyPI server to use (defaults to
   `https://pypi.org`)
 
-  If your registry supports the pypi.org-compatible JSON API (e.g.
-  `{registry}/pypi/{package}/json`), specify your base registry URL without
-  appending `/simple` (e.g. `https://my-registry`).
+  For best results (i.e. most available metadata), choose the first option from
+  this list which is supported by your PyPI registry:
 
-  If your registry only supports the traditional HTML "simple" index, specify
-  the registry URL with `/simple` at the end (e.g.
-  `https://my-registry/simple`).
+    * If your registry supports [PEP691](pep691) JSON "simple" indexes, specify
+      your base registry URL with `/simple` appended (e.g.
+      `https://my-registry/simple`).
+
+    * If your registry supports the legacy pypi.org-compatible JSON API (e.g.
+      `{registry}/pypi/{package}/json`), specify your base registry URL without
+      a suffix (e.g. `https://my-registry`).
+
+    * Otherwise, if your registry only supports the traditional HTML "simple"
+      index, specify the registry URL with `/simple` at the end (e.g.
+      `https://my-registry/simple`).
 
-  Note that the [PEP691][pep691] JSON-based "simple" API is not yet supported.
 
 * `PYPI_BROWSER_PACKAGE_CACHE_PATH`: Filesystem path to use for caching
   downloaded files. This will grow forever (the app does not clean it up) so

pypi_browser/pypi.py

@@ -37,29 +37,40 @@ def handle_starttag(self, tag: str, attrs: list[tuple[str, str | None]]) -> None
 
 @dataclasses.dataclass(frozen=True)
 class SimpleRepository(PythonRepository):
-    """Old-style "simple" PyPI registry serving HTML files."""
-    # TODO: Also handle PEP691 JSON simple repositories.
+    """"Simple" PyPI registry serving either JSON or HTML files."""
     pypi_url: str
 
     async def files_for_package(self, package_name: str) -> dict[str, str]:
         async with httpx.AsyncClient() as client:
             resp = await client.get(
                 f'{self.pypi_url}/{package_name}',
                 follow_redirects=True,
+                headers={
+                    'Accept': ', '.join((
+                        'application/vnd.pypi.simple.v1+json',
+                        'application/vnd.pypi.simple.v1+html;q=0.2',
+                        'text/html;q=0.01',
+                    )),
+                },
             )
             if resp.status_code == 404:
                 raise PackageDoesNotExist(package_name)
-            parser = HTMLAnchorParser()
-            parser.feed(resp.text)
 
             def clean_url(url: str) -> str:
                 parsed = urllib.parse.urlparse(urllib.parse.urljoin(str(resp.url), url))
                 return parsed._replace(fragment='').geturl()
 
-            return {
-                (urllib.parse.urlparse(url).path).split('/')[-1]: clean_url(url)
-                for url in parser.anchors
-            }
+            if resp.headers.get('Content-Type') == 'application/vnd.pypi.simple.v1+json':
+                result = resp.json()
+                return {file_['filename']: clean_url(file_['url']) for file_ in result['files']}
+            else:
+                parser = HTMLAnchorParser()
+                parser.feed(resp.text)
+
+                return {
+                    (urllib.parse.urlparse(url).path).split('/')[-1]: clean_url(url)
+                    for url in parser.anchors
+                }
 
 
 @dataclasses.dataclass(frozen=True)