BUG: read_csv throws UnicodeDecodeError with unicode aliases

doc/source/whatsnew/v0.19.0.txt

@@ -81,3 +81,4 @@ Performance Improvements
 
 Bug Fixes
 ~~~~~~~~~
+- Bug in ``read_csv()``, where aliases for utf-xx (e.g. UTF-xx, UTF_xx, utf_xx) raised UnicodeDecodeError (:issue:`13549`)

pandas/io/parsers.py

@@ -339,6 +339,10 @@ def _validate_nrows(nrows):
 def _read(filepath_or_buffer, kwds):
     "Generic reader of line files."
     encoding = kwds.get('encoding', None)
+    if encoding is not None:
+        encoding = re.sub('_', '-', encoding).lower()
+        kwds['encoding'] = encoding
+
     skipfooter = kwds.pop('skipfooter', None)
     if skipfooter is not None:
         kwds['skip_footer'] = skipfooter

pandas/io/tests/parser/common.py

@@ -1469,3 +1469,34 @@ def test_memory_map(self):
 
         out = self.read_csv(mmap_file, memory_map=True)
         tm.assert_frame_equal(out, expected)
+
+    def test_read_csv_utf_aliases():
+        # see gh issue 13549
+        engines = ['c', 'python', None]
+        path = 'test.csv'
+        expected = DataFrame({"A": [0, 1], "B": [2, 3]})
+        expected.to_csv(path, encoding='utf-8', index=False)
+        test_encodings = ['utf-8', 'utf_8', 'UTF_8', 'UTF-8']
+
+        for encoding in test_encodings:
+            for engine in engines:
+                out = pd.io.parsers.read_csv(
+                    path,
+                    engine=engine,
+                    encoding=encoding)
+                tm.assert_frame_equal(out, expected)
+
+        os.remove("test.csv")
+
+        expected.to_csv(path, encoding='utf-16', index=False)
+        test_encodings = ['utf-16', 'utf_16', 'UTF_16', 'UTF-16']
+
+        for encoding in test_encodings:
+            for engine in engines:
+                out = pd.io.parsers.read_csv(
+                    path,
+                    engine=engine,
+                    encoding=encoding)
+                tm.assert_frame_equal(out, expected)
+
+        os.remove("test.csv")