Rework of the URL subtraction feature

CHANGES/1392.feature.rst

@@ -0,0 +1,22 @@
+Added support for using the :meth:`subtraction operator <yarl.URL.__sub__>`
+to get the relative path between URLs.
+
+Note that both URLs must have the same scheme, user, password, host and port:
+
+.. code-block:: pycon
+
+   >>> target = URL("http://example.com/path/index.html")
+   >>> base = URL("http://example.com/")
+   >>> target - base
+   URL('path/index.html')
+
+URLs can also be relative:
+
+.. code-block:: pycon
+
+   >>> target = URL("/")
+   >>> base = URL("/path/index.html")
+   >>> target - base
+   URL('..')
+
+-- by :user:`oleksbabieiev`

docs/api.rst

@@ -997,6 +997,20 @@ The path is encoded if needed.
          >>> base.join(URL('//python.org/page.html'))
          URL('http://python.org/page.html')
 
+.. method:: URL.__sub__(url)
+
+   Return a new URL with a relative *path* between two other URL objects.
+   *scheme*, *user*, *password*, *host*, *port*, *query* and *fragment* are removed.
+
+   .. doctest::
+
+      >>> target = URL('http://example.com/path/index.html')
+      >>> base = URL('http://example.com/')
+      >>> target - base
+      URL('path/index.html')
+
+   .. versionadded:: 1.18
+
 Human readable representation
 -----------------------------
 

tests/test_url.py

@@ -61,6 +61,89 @@ def test_str():
     assert str(url) == "http://example.com:8888/path/to?a=1&b=2"
 
 
+@pytest.mark.parametrize(
+    ("target", "base", "expected"),
+    [
+        ("http://example.com/path/to", "http://example.com/", "path/to"),
+        ("http://example.com/path/to", "http://example.com/spam", "path/to"),
+        ("http://example.com/this/is/a/test", "http://example.com/this/", "is/a/test"),
+        (
+            "http://example.com/this/./is/a/test",
+            "http://example.com/this/",
+            "is/a/test",
+        ),
+        ("http://example.com/this/is/../a//test", "http://example.com/this/", "a/test"),
+        ("http://example.com/path/to", "http://example.com/spam/", "../path/to"),
+        ("http://example.com/path", "http://example.com/path/to/", ".."),
+        ("http://example.com/path", "http://example.com/other/../path/to/", ".."),
+        ("http://example.com/", "http://example.com/", "."),
+        ("http://example.com", "http://example.com", "."),
+        ("http://example.com/", "http://example.com", "."),
+        ("http://example.com", "http://example.com/", "."),
+        ("//example.com", "//example.com", "."),
+        ("/path/to", "/spam/", "../path/to"),
+        ("path/to", "spam/", "../path/to"),
+        ("path/../to", "path/", "../to"),
+        ("path/..", ".", "path/.."),
+        ("path/../replace/me", "path/../replace", "replace/me"),
+        ("path/../replace/me", "path/../replace/", "me"),
+        ("path/to", "spam", "path/to"),
+        ("..", ".", ".."),
+        (".", "..", "."),
+    ],
+)
+def test_sub(target: str, base: str, expected: str):
+    expected_url = URL(expected)
+    result_url = URL(target) - URL(base)
+    assert result_url == expected_url
+
+
+@pytest.mark.xfail(reason="Empty segments are not preserved")
+@pytest.mark.parametrize(
+    ("target", "base", "expected"),
+    [
+        (
+            "http://example.com/////path/////to",
+            "http://example.com/////spam",
+            "path/////to",
+        ),
+        (
+            "http://example.com////path/////to",
+            "http://example.com/////spam",
+            "..//path/////to",
+        ),
+    ],
+)
+def test_sub_empty_segments(target: str, base: str, expected: str):
+    expected_url = URL(expected)
+    result_url = URL(target) - URL(base)
+    assert result_url == expected_url
+
+
+def test_sub_with_different_schemes():
+    expected_error_msg = r"^Both URLs should have the same scheme$"
+    with pytest.raises(ValueError, match=expected_error_msg):
+        URL("http://example.com/") - URL("https://example.com/")
+
+
+def test_sub_with_different_netlocs():
+    expected_error_msg = r"^Both URLs should have the same netloc$"
+    with pytest.raises(ValueError, match=expected_error_msg):
+        URL("https://spam.com/") - URL("https://ham.com/")
+
+
+def test_sub_with_different_anchors():
+    expected_error_msg = r"^'path/to' and '/path' have different anchors$"
+    with pytest.raises(ValueError, match=expected_error_msg):
+        URL("path/to") - URL("/path/from")
+
+
+def test_sub_with_two_dots_in_base():
+    expected_error_msg = r"^'..' segment in '/path/..' cannot be walked$"
+    with pytest.raises(ValueError, match=expected_error_msg):
+        URL("path/to") - URL("/path/../from")
+
+
 def test_repr():
     url = URL("http://example.com")
     assert "URL('http://example.com')" == repr(url)

tests/test_url_benchmarks.py

@@ -18,11 +18,14 @@
 QUERY_URL = URL(QUERY_URL_STR)
 URL_WITH_PATH_STR = "http://www.domain.tld/req"
 URL_WITH_PATH = URL(URL_WITH_PATH_STR)
+URL_WITH_LONGER_PATH = URL("http://www.domain.tld/req/req/req")
 REL_URL = URL("/req")
 QUERY_SEQ = {str(i): tuple(str(j) for j in range(10)) for i in range(10)}
 SIMPLE_QUERY = {str(i): str(i) for i in range(10)}
 SIMPLE_INT_QUERY = {str(i): i for i in range(10)}
 QUERY_STRING = "x=y&z=1"
+URL_VERY_LONG_PATH = URL("http://www.domain.tld/" + "req/" * 100)
+URL_LONG_PATH = URL("http://www.domain.tld/" + "req/" * 30)
 
 
 class _SubClassedStr(str):
@@ -586,6 +589,20 @@ def _run() -> None:
             url.query
 
 
+def test_url_subtract(benchmark: BenchmarkFixture) -> None:
+    @benchmark
+    def _run() -> None:
+        for _ in range(100):
+            URL_WITH_LONGER_PATH - URL_WITH_PATH
+
+
+def test_url_subtract_long_urls(benchmark: BenchmarkFixture) -> None:
+    @benchmark
+    def _run() -> None:
+        for _ in range(100):
+            URL_VERY_LONG_PATH - URL_LONG_PATH
+
+
 def test_url_host_port_subcomponent(benchmark: BenchmarkFixture) -> None:
     cache_non_default = URL_WITH_NOT_DEFAULT_PORT._cache
     cache = BASE_URL._cache

yarl/_path.py

@@ -2,6 +2,8 @@
 
 from collections.abc import Sequence
 from contextlib import suppress
+from itertools import chain
+from pathlib import PurePosixPath
 
 
 def normalize_path_segments(segments: Sequence[str]) -> list[str]:
@@ -39,3 +41,31 @@ def normalize_path(path: str) -> str:
 
     segments = path.split("/")
     return prefix + "/".join(normalize_path_segments(segments))
+
+
+def calculate_relative_path(target: str, base: str) -> str:
+    """Return the relative path between two other paths.
+
+    If the operation is not possible, raise ValueError.
+    """
+
+    target = target or "/"
+    base = base or "/"
+
+    target_path = PurePosixPath(target)
+    base_path = PurePosixPath(base)
+
+    if base[-1] != "/":
+        base_path = base_path.parent
+
+    for step, path in enumerate(chain((base_path,), base_path.parents)):
+        if path == target_path or path in target_path.parents:
+            break
+        elif path.name == "..":
+            raise ValueError(f"'..' segment in {str(base_path)!r} cannot be walked")
+    else:
+        raise ValueError(
+            f"{str(target_path)!r} and {str(base_path)!r} have different anchors"
+        )
+    offset = len(path.parts)
+    return str(PurePosixPath(*("..",) * step, *target_path.parts[offset:]))

yarl/_url.py

@@ -20,7 +20,7 @@
     split_url,
     unsplit_result,
 )
-from ._path import normalize_path, normalize_path_segments
+from ._path import calculate_relative_path, normalize_path, normalize_path_segments
 from ._query import (
     Query,
     QueryVariable,
@@ -516,6 +516,32 @@ def __truediv__(self, name: str) -> "URL":
     def __mod__(self, query: Query) -> "URL":
         return self.update_query(query)
 
+    def __sub__(self, other: object) -> "URL":
+        """Return a new URL with a relative path between two other URL objects.
+
+        Note that both URLs must have the same scheme and netloc.
+
+        Example:
+        >>> target = URL("http://example.com/path/index.html")
+        >>> base = URL("http://example.com/")
+        >>> target - base
+        URL('path/index.html')
+        """
+
+        if type(other) is not URL:
+            return NotImplemented
+
+        target_scheme, target_netloc, target_path, _, _ = self._val
+        base_scheme, base_netloc, base_path, _, _ = other._val
+
+        if target_scheme != base_scheme:
+            raise ValueError("Both URLs should have the same scheme")
+        if target_netloc != base_netloc:
+            raise ValueError("Both URLs should have the same netloc")
+
+        path = calculate_relative_path(target_path, base_path)
+        return self._from_parts("", "", path, "", "")
+
     def __bool__(self) -> bool:
         return bool(self._netloc or self._path or self._query or self._fragment)