do not override orc global mark in test_orc.py

docs/source/developers/python.rst

@@ -360,8 +360,6 @@ Now, we build and install Arrow C++ libraries
    set ARROW_HOME=C:\thirdparty
    cmake -G "Visual Studio 14 2015 Win64" ^
          -DCMAKE_INSTALL_PREFIX=%ARROW_HOME% ^
-         -DCMAKE_BUILD_TYPE=Release ^
-         -DARROW_BUILD_TESTS=on ^
          -DARROW_CXXFLAGS="/WX /MP" ^
          -DARROW_GANDIVA=on ^
          -DARROW_PARQUET=on ^
@@ -380,7 +378,9 @@ Now, we can build pyarrow:
 .. code-block:: shell
 
    cd python
-   python setup.py build_ext --inplace --with-parquet
+   set PYARROW_WITH_GANDIVA=1
+   set PYARROW_WITH_PARQUET=1
+   python setup.py build_ext --inplace
 
 Then run the unit tests with:
 

python/pyarrow/orc.py

@@ -20,6 +20,7 @@ from numbers import Integral
 
 from pyarrow import types
 from pyarrow.lib import Schema
+import pyarrow._orc as _orc
 
 
 def _is_map(typ):
@@ -69,8 +70,7 @@ class ORCFile(object):
         see pyarrow.io.PythonFileInterface or pyarrow.io.BufferReader.
     """
     def __init__(self, source):
-        from pyarrow._orc import ORCReader
-        self.reader = ORCReader()
+        self.reader = _orc.ORCReader()
         self.reader.open(source)
         self._column_index_lookup = _schema_to_indices(self.schema)
 

python/pyarrow/tests/test_orc.py

@@ -33,11 +33,6 @@ except ImportError:
     pass
 
 
-# TODO(wesm): The ORC tests depend a lot on pandas currently, so all excluded
-# when it is not installed
-pytestmark = pytest.mark.pandas
-
-
 @pytest.fixture(scope='module')
 def datadir(datadir):
     return datadir / 'orc'
@@ -119,6 +114,7 @@ def check_example_file(orc_path, expected_df, need_fix=False):
     assert json_pos == orc_file.nrows
 
 
+@pytest.mark.pandas
 @pytest.mark.parametrize('filename', [
     'TestOrcFile.test1.orc',
     'TestOrcFile.testDate1900.orc',