fix: only percent-encode characters in the userinfo encode set (#1852)

Author: davidhewitt

src/url.rs

@@ -7,7 +7,7 @@ use std::sync::OnceLock;
 
 use idna::punycode::decode_to_string;
 use jiter::{PartialMode, StringCacheMode};
-use percent_encoding::{percent_encode, NON_ALPHANUMERIC};
+use percent_encoding::{percent_encode, AsciiSet, CONTROLS};
 use pyo3::exceptions::PyValueError;
 use pyo3::pyclass::CompareOp;
 use pyo3::sync::OnceLockExt;
@@ -602,8 +602,36 @@ fn is_punnycode_domain(lib_url: &Url, domain: &str) -> bool {
     scheme_is_special(lib_url.scheme()) && domain.split('.').any(|part| part.starts_with(PUNYCODE_PREFIX))
 }
 
+/// See <https://url.spec.whatwg.org/#userinfo-percent-encode-set>
+const USERINFO_ENCODE_SET: &AsciiSet = &CONTROLS
+    // query percent-encodes is controls plus the below
+    .add(b' ')
+    .add(b'"')
+    .add(b'#')
+    .add(b'<')
+    .add(b'>')
+    // path percent-encodes is query percent-encodes plus the below
+    .add(b'?')
+    .add(b'^')
+    .add(b'`')
+    .add(b'{')
+    .add(b'}')
+    // userinfo percent-encodes is path percent-encodes plus the below
+    .add(b'/')
+    .add(b':')
+    .add(b';')
+    .add(b'=')
+    .add(b'@')
+    .add(b'[')
+    .add(b'\\')
+    .add(b']')
+    .add(b'|')
+    // https://datatracker.ietf.org/doc/html/rfc3986.html#section-2.4
+    // we must also percent-encode '%'
+    .add(b'%');
+
 fn encode_userinfo_component(value: &str) -> Cow<'_, str> {
-    let encoded = percent_encode(value.as_bytes(), NON_ALPHANUMERIC).to_string();
+    let encoded = percent_encode(value.as_bytes(), USERINFO_ENCODE_SET).to_string();
     if encoded == value {
         Cow::Borrowed(value)
     } else {

tests/validators/test_url.py

@@ -1318,31 +1318,38 @@ def test_multi_url_build() -> None:
     assert str(url) == 'postgresql://testuser:testpassword@127.0.0.1:5432/database?sslmode=require#test'
 
 
-def test_multi_url_build_encodes_credentials() -> None:
-    url = MultiHostUrl.build(
+@pytest.mark.parametrize('url_type', [Url, MultiHostUrl])
+def test_url_build_encodes_credentials(url_type: type[Union[Url, MultiHostUrl]]) -> None:
+    url = url_type.build(
         scheme='postgresql',
         username='user name',
-        password='p@ss/word?#',
+        password='p@ss/word?#__',
         host='example.com',
         port=5432,
     )
-    assert url == MultiHostUrl('postgresql://user%20name:p%40ss%2Fword%3F%23@example.com:5432')
-    assert str(url) == 'postgresql://user%20name:p%40ss%2Fword%3F%23@example.com:5432'
-    assert url.hosts() == [
-        {'username': 'user%20name', 'password': 'p%40ss%2Fword%3F%23', 'host': 'example.com', 'port': 5432}
-    ]
+    assert url == url_type('postgresql://user%20name:p%40ss%2Fword%3F%23__@example.com:5432')
+    assert str(url) == 'postgresql://user%20name:p%40ss%2Fword%3F%23__@example.com:5432'
+    if url_type is Url:
+        assert url.username == 'user%20name'
+        assert url.password == 'p%40ss%2Fword%3F%23__'
+    else:
+        assert url.hosts() == [
+            {'username': 'user%20name', 'password': 'p%40ss%2Fword%3F%23__', 'host': 'example.com', 'port': 5432}
+        ]
 
 
 def test_multi_url_build_hosts_encodes_credentials() -> None:
     hosts = [
-        {'host': 'example.com', 'password': 'p@ss/word?#', 'username': 'user name', 'port': 5431},
-        {'host': 'example.org', 'password': 'pa%ss', 'username': 'other', 'port': 5432},
+        {'host': 'example.com', 'password': 'p@ss/word?#__', 'username': 'user name', 'port': 5431},
+        {'host': 'example.org', 'password': 'p@%ss__', 'username': 'other', 'port': 5432},
     ]
     url = MultiHostUrl.build(scheme='postgresql', hosts=hosts)
-    assert str(url) == 'postgresql://user%20name:p%40ss%2Fword%3F%23@example.com:5431,other:pa%25ss@example.org:5432'
+    assert (
+        str(url) == 'postgresql://user%20name:p%40ss%2Fword%3F%23__@example.com:5431,other:p%40%25ss__@example.org:5432'
+    )
     assert url.hosts() == [
-        {'username': 'user%20name', 'password': 'p%40ss%2Fword%3F%23', 'host': 'example.com', 'port': 5431},
-        {'username': 'other', 'password': 'pa%25ss', 'host': 'example.org', 'port': 5432},
+        {'username': 'user%20name', 'password': 'p%40ss%2Fword%3F%23__', 'host': 'example.com', 'port': 5431},
+        {'username': 'other', 'password': 'p%40%25ss__', 'host': 'example.org', 'port': 5432},
     ]