Use PARQUET from miniconda path

ci/travis_script_python.sh

@@ -7,6 +7,7 @@ PYTHON_DIR=$TRAVIS_BUILD_DIR/python
 # Re-use conda installation from C++
 export MINICONDA=$TRAVIS_BUILD_DIR/miniconda
 export PATH="$MINICONDA/bin:$PATH"
+export PARQUET_HOME=$MINICONDA
 
 # Share environment with C++
 pushd $CPP_BUILD_DIR

python/pyarrow/tests/test_parquet.py

@@ -27,6 +27,19 @@ from tempfile import mkdtemp
 import os.path
 
 
+def test_single_int64_column(tmpdir):
+    filename = tmpdir.join('single_int64_column.parquet')
+    data = [A.from_pylist(range(5))]
+    table = A.Table.from_arrays(('a', 'b'), data, 'table_name')
+    A.parquet.write_table(table, filename.strpath)
+    table_read = pyarrow.parquet.read_table(filename.strpath)
+    for col_written, col_read in zip(table.itercolumns(), table_read.itercolumns()):
+        assert col_written.name == col_read.name
+        assert col_read.data.num_chunks == 1
+        data_written = col_written.data.chunk(0)
+        data_read = col_read.data.chunk(0)
+        assert data_written == data_read
+
 class TestParquetIO(unittest.TestCase):
 
   def setUp(self):