fix: correctly handle absolute URLs in parse_simpleapi_html.bzl (#2112)

This PR addresses a typo and improves the handling of absolute URLs in
the
`parse_simpleapi_html.bzl` file and corrects a minor issue in the
`simpleapi_download.bzl` file.

Summary:
1. parse_simpleapi_html.bzl:
- Introduced a new private function `_get_root_directory(url)` to
extract
     the root directory from a given URL.
- Enhanced `_absolute_url` function to correctly handle absolute URLs by
utilizing the `_get_root_directory` function. This ensures that URLs
starting with a "/" are correctly resolved to their full path, avoiding
     potential incorrect concatenation.

2. simpleapi_download.bzl: Corrected the handling of the `real_url`
variable in
the `_read_simpleapi` function, ensuring that the correct URL is passed
to
   `_read_index_result` when using non-blocking downloads.

Author: WSUFan

CHANGELOG.md

@@ -35,6 +35,7 @@ A brief description of the categories of changes:
     * `3.12 -> 3.12.4`
 
 ### Fixed
+* (rules) correctly handle absolute URLs in parse_simpleapi_html.bzl.
 * (rules) Fixes build targets linking against `@rules_python//python/cc:current_py_cc_libs`
   in host platform builds on macOS, by editing the `LC_ID_DYLIB` field of the hermetic interpreter's
   `libpython3.x.dylib` using `install_name_tool`, setting it to its absolute path under Bazel's

python/private/pypi/parse_simpleapi_html.bzl

@@ -96,7 +96,25 @@ def parse_simpleapi_html(*, url, content):
         whls = whls,
     )
 
+def _get_root_directory(url):
+    scheme_end = url.find("://")
+    if scheme_end == -1:
+        fail("Invalid URL format")
+
+    scheme = url[:scheme_end]
+    host_end = url.find("/", scheme_end + 3)
+    if host_end == -1:
+        host_end = len(url)
+    host = url[scheme_end + 3:host_end]
+
+    return "{}://{}".format(scheme, host)
+
 def _absolute_url(index_url, candidate):
+    if candidate.startswith("/"):
+        # absolute url
+        root_directory = _get_root_directory(index_url)
+        return "{}{}".format(root_directory, candidate)
+
     if not candidate.startswith(".."):
         return candidate
 

python/private/pypi/simpleapi_download.bzl

@@ -185,10 +185,10 @@ def _read_simpleapi(ctx, url, attr, cache, **download_kwargs):
     if download_kwargs.get("block") == False:
         # Simulate the same API as ctx.download has
         return struct(
-            wait = lambda: _read_index_result(ctx, download.wait(), output, url, cache, cache_key),
+            wait = lambda: _read_index_result(ctx, download.wait(), output, real_url, cache, cache_key),
         )
 
-    return _read_index_result(ctx, download, output, url, cache, cache_key)
+    return _read_index_result(ctx, download, output, real_url, cache, cache_key)
 
 def _read_index_result(ctx, result, output, url, cache, cache_key):
     if not result.success:

tests/pypi/parse_simpleapi_html/parse_simpleapi_html_tests.bzl

@@ -221,6 +221,40 @@ def _test_whls(env):
                 yanked = False,
             ),
         ),
+        (
+            struct(
+                attrs = [
+                    'href="/whl/torch-2.0.0-cp38-cp38-manylinux2014_aarch64.whl#sha256=deadbeef"',
+                ],
+                filename = "torch-2.0.0-cp38-cp38-manylinux2014_aarch64.whl",
+                url = "https://download.pytorch.org/whl/cpu/torch",
+            ),
+            struct(
+                filename = "torch-2.0.0-cp38-cp38-manylinux2014_aarch64.whl",
+                metadata_sha256 = "",
+                metadata_url = "",
+                sha256 = "deadbeef",
+                url = "https://download.pytorch.org/whl/torch-2.0.0-cp38-cp38-manylinux2014_aarch64.whl",
+                yanked = False,
+            ),
+        ),
+        (
+            struct(
+                attrs = [
+                    'href="/whl/torch-2.0.0-cp38-cp38-manylinux2014_aarch64.whl#sha256=notdeadbeef"',
+                ],
+                filename = "torch-2.0.0-cp38-cp38-manylinux2014_aarch64.whl",
+                url = "http://download.pytorch.org/whl/cpu/torch",
+            ),
+            struct(
+                filename = "torch-2.0.0-cp38-cp38-manylinux2014_aarch64.whl",
+                metadata_sha256 = "",
+                metadata_url = "",
+                sha256 = "notdeadbeef",
+                url = "http://download.pytorch.org/whl/torch-2.0.0-cp38-cp38-manylinux2014_aarch64.whl",
+                yanked = False,
+            ),
+        ),
     ]
 
     for (input, want) in tests: