Fix trailing slash behavior (#488)

* upath: ensure stripped trailing slashes in cloudpaths

* tests: data path doesn't do joins

* tests: adjust test cases to match intent of having identical __init__, joinpath, truediv behavior

* tests: adjust trailing slash behavior of gcs and azure path

* tests: add multiple args in init to trailing slash conditions

Author: ap--

upath/core.py

@@ -573,12 +573,10 @@ def __init__(
         # FIXME: normalization needs to happen in unchain already...
         chain = Chain.from_list(Chain.from_list(segments).to_list())
         if len(args) > 1:
-            chain = chain.replace(
-                path=WrappedFileSystemFlavour.from_protocol(protocol).join(
-                    chain.active_path,
-                    *args[1:],
-                )
-            )
+            flavour = WrappedFileSystemFlavour.from_protocol(protocol)
+            joined = flavour.join(chain.active_path, *args[1:])
+            stripped = flavour.strip_protocol(joined)
+            chain = chain.replace(path=stripped)
         self._chain = chain
         self._chain_parser = chain_parser
         self._raw_urlpaths = args

upath/implementations/cloud.py

@@ -72,7 +72,7 @@ def __str__(self) -> str:
 
     @property
     def path(self) -> str:
-        self_path = super().path
+        self_path = super().path.rstrip(self.parser.sep)
         if (
             self._relative_base is None
             and self_path

upath/tests/cases.py

@@ -677,3 +677,37 @@ def test_relative_to(self):
         child = self.path / "folder1" / "file1.txt"
         relative = child.relative_to(base)
         assert str(relative) == "folder1/file1.txt"
+
+    def test_trailing_slash_joinpath_is_identical(self):
+        # setup
+        cls = type(self.path)
+        protocol = self.path.protocol
+        path = self.path.path
+        sopts = self.path.storage_options
+        if not path:
+            path = "something"
+            path_with_slash = "something/"
+        elif path.endswith("/"):
+            path_with_slash = path
+            path = path.removeprefix("/")
+        else:
+            path_with_slash = path + "/"
+        key = "key/"
+
+        # test
+        a = cls(path_with_slash + key, protocol=protocol, **sopts)
+        b = cls(path_with_slash, key, protocol=protocol, **sopts)
+        c = cls(path_with_slash, protocol=protocol, **sopts).joinpath(key)
+        d = cls(path_with_slash, protocol=protocol, **sopts) / key
+        assert a.path == b.path == c.path == d.path
+
+    def test_trailing_slash_is_stripped(self):
+        has_meaningful_trailing_slash = getattr(
+            self.path.parser, "has_meaningful_trailing_slash", False
+        )
+        if has_meaningful_trailing_slash:
+            assert not self.path.joinpath("key").path.endswith("/")
+            assert self.path.joinpath("key/").path.endswith("/")
+        else:
+            assert not self.path.joinpath("key").path.endswith("/")
+            assert not self.path.joinpath("key/").path.endswith("/")

upath/tests/implementations/test_data.py

@@ -287,3 +287,11 @@ def test_move_into_memory(self, clear_fsspec_memory_cache):
     @pytest.mark.skip(reason="DataPath does not support relative_to")
     def test_relative_to(self):
         pass
+
+    @pytest.mark.skip(reason="DataPath does not support joins")
+    def test_trailing_slash_joinpath_is_identical(self):
+        pass
+
+    @pytest.mark.skip(reason="DataPath does not support joins")
+    def test_trailing_slash_is_stripped(self):
+        pass