ARROW-7311: [Python] Return filesystem and path from URI

This should supersede #5977

Closes #6197 from kszucs/fs-from-uri and squashes the following commits:

523972f <Krisztián Szűcs> relative paths
4b1f80b <Krisztián Szűcs> filesystem from uri
4ec33b1 <Krisztián Szűcs> from URI

Authored-by: Krisztián Szűcs <szucs.krisztian@gmail.com>
Signed-off-by: Joris Van den Bossche <jorisvandenbossche@gmail.com>

python/pyarrow/_fs.pyx

@@ -190,6 +190,29 @@ cdef class FileSystem:
                         "the subclasses instead: LocalFileSystem or "
                         "SubTreeFileSystem")
 
+    @staticmethod
+    def from_uri(uri):
+        """Create a new FileSystem from by URI
+
+        A scheme-less URI is considered a local filesystem path.
+        Recognized schemes are "file", "mock", "hdfs" and "viewfs".
+
+        Parameters
+        ----------
+        uri : string
+            URI-based path, for example: file:///some/local/path
+
+        Returns
+        -------
+        With (filesystem, path) tuple where path is the abtract path inside the
+        FileSystem instance.
+        """
+        cdef:
+            c_string path
+            CResult[shared_ptr[CFileSystem]] result
+        result = CFileSystemFromUri(tobytes(uri), &path)
+        return FileSystem.wrap(GetResultValue(result)), frombytes(path)
+
     cdef init(self, const shared_ptr[CFileSystem]& wrapped):
         self.wrapped = wrapped
         self.fs = wrapped.get()

python/pyarrow/includes/libarrow_fs.pxd

@@ -79,6 +79,9 @@ cdef extern from "arrow/filesystem/api.h" namespace "arrow::fs" nogil:
         CResult[shared_ptr[COutputStream]] OpenAppendStream(
             const c_string& path)
 
+    CResult[shared_ptr[CFileSystem]] CFileSystemFromUri \
+        "arrow::fs::FileSystemFromUri"(const c_string& uri, c_string* out_path)
+
     cdef cppclass CLocalFileSystemOptions "arrow::fs::LocalFileSystemOptions":
         c_bool use_mmap
 

python/pyarrow/tests/test_fs.py

@@ -590,3 +590,24 @@ def test_hdfs_options(hdfs_server):
     uri = "hdfs://{}:{}/?user={}".format(host, port, user)
     fs = HadoopFileSystem(uri)
     assert fs.get_target_stats(FileSelector('/'))
+
+
+@pytest.mark.parametrize(('uri', 'expected_klass', 'expected_path'), [
+    # leading slashes are removed intentionally, becuase MockFileSystem doesn't
+    # have a distinction between relative and absolute paths
+    ('mock:', _MockFileSystem, ''),
+    ('mock:foo/bar', _MockFileSystem, 'foo/bar'),
+    ('mock:/foo/bar', _MockFileSystem, 'foo/bar'),
+    ('mock:///foo/bar', _MockFileSystem, 'foo/bar'),
+    ('file:', LocalFileSystem, ''),
+    ('file:foo/bar', LocalFileSystem, 'foo/bar'),
+    ('file:/foo/bar', LocalFileSystem, '/foo/bar'),
+    ('file:///foo/bar', LocalFileSystem, '/foo/bar'),
+    ('', LocalFileSystem, ''),
+    ('foo/bar', LocalFileSystem, 'foo/bar'),
+    ('/foo/bar', LocalFileSystem, '/foo/bar'),
+])
+def test_filesystem_from_uri(uri, expected_klass, expected_path):
+    fs, path = FileSystem.from_uri(uri)
+    assert isinstance(fs, expected_klass)
+    assert path == expected_path