diff --git a/.github/workflows/python-package.yml b/.github/workflows/python-package.yml
new file mode 100644
index 0000000..59f9ae9
--- /dev/null
+++ b/.github/workflows/python-package.yml
@@ -0,0 +1,39 @@
+# This workflow will install Python dependencies, run tests and lint with a variety of Python versions
+# For more information see: https://help.github.com/actions/language-and-framework-guides/using-python-with-github-actions
+
+name: Python package
+
+on:
+  push:
+    branches: [ master ]
+  pull_request:
+    branches: [ master ]
+
+jobs:
+  build:
+
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: [2.7, 3.5, 3.6, 3.7, 3.8]
+
+    steps:
+    - uses: actions/checkout@v2
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v2
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install flake8
+        if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+    - name: Lint with flake8
+      run: |
+        # stop the build if there are Python syntax errors or undefined names
+        flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
+        # exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
+        flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
+    - name: Test with pytest
+      run: |
+        python -m unittest -v
diff --git a/devmem/__init__.py b/devmem/__init__.py
index e8c9358..952e60f 100644
--- a/devmem/__init__.py
+++ b/devmem/__init__.py
@@ -12,67 +12,60 @@
     http://www.python.org/dev/peps/pep-0008/
 
 """
+# TODO delete when dropping Python 2 support
+from __future__ import print_function
 
 import os
 import sys
 import mmap
 import struct
 
+
 class DevMemBuffer:
     """This class holds data for objects returned from DevMem class. It allows an easy way to print hex data"""
 
-    def __init__(self, base_addr, data):
+    lut = {1: "B", 2: "H", 4: "I"}
+
+    def __init__(self, base_addr, data, word=4):
         self.data = data
         self.base_addr = base_addr
+        self.word = word
 
     def __len__(self):
         return len(self.data)
 
-    def __getitem__(self, key):
-        return self.data[key]
+    def __getitem__(self, i):
+        start = i * self.word
+        stop = start + self.word
+
+        if stop > len(self.data):
+            raise IndexError
 
-    def __setitem__(self, key, value):
-        self.data[key] = value
+        return self.data[start:stop]
 
-    def hexdump(self, word_size = 4, words_per_row = 4):
+    def __setitem__(self, i, value):
+        raise NotImplementedError("DevMemBuffer is immutable")
+        # self.data[i] = value
+
+    def hexdump(self, words_per_row=4):
         # Build a list of strings and then join them in the last step.
         # This is more efficient then concat'ing immutable strings.
 
-        d = self.data
         dump = []
+        row_step = words_per_row * self.word
 
-        word = 0
-        while (word < len(d)):
-            dump.append('0x{0:02x}:  '.format(self.base_addr
-                                              + word_size * word))
-
-            max_col = word + words_per_row
-            if max_col > len(d): max_col = len(d)
-
-            while word < max_col:
-                # If the word is 4 bytes, then handle it and continue the
-                # loop, this should be the normal case
-                if word_size == 4:
-                    dump.append(" {0:08x} ".format(d[word]))
+        for row_pos in range(0, len(self.data), row_step):
+            row = ["0x{0:02x}: ".format(self.base_addr + row_pos)]
 
-                # Otherwise the word_size is not an int, pack it so it can be
-                # un-packed to the desired word size.  This should blindly
-                # handle endian problems (Verify?)
-                elif word_size == 2:
-                    for halfword in struct.unpack('HH', struct.pack('I',(d[word]))):
-                        dump.append(" {0:04x}".format(halfword))
+            row_stop = min(row_pos + row_step, len(self.data))
+            for col_pos in range(row_pos, row_stop, self.word):
+                data = struct.unpack_from(self.lut[self.word], self.data, col_pos)
+                for b in data:
+                    row.append(" {0:0{width}x}".format(b, width=self.word * 2))
 
-                elif word_size == 1:
-                    for byte in struct.unpack('BBBB', struct.pack('I',(d[word]))):
-                        dump.append(" {0:02x}".format(byte))
+            dump.append(" ".join(row))
 
-                word += 1
-
-            dump.append('\n')
-
-        # Chop off the last new line character and join the list of strings
-        # in to a single string
-        return ''.join(dump[:-1])
+        return "\n".join(dump)
 
     def __str__(self):
         return self.hexdump()
@@ -81,40 +74,48 @@ def __str__(self):
 class DevMem:
     """Class to read and write data aligned to word boundaries of /dev/mem"""
 
-    # Size of a word that will be used for reading/writing
-    word = 4
-    mask = ~(word - 1)
     f = None
 
-    def __init__(self, base_addr, length = 1, filename = '/dev/mem',
-                 debug = 0):
+    def __init__(self, base_addr, length=1, filename="/dev/mem", debug=0, word=4):
 
-        if base_addr < 0 or length < 0: raise AssertionError
+        if base_addr < 0 or length < 0:
+            raise AssertionError
         self._debug = debug
 
+        # Size of a word that will be used for reading/writing
+        self.word = word
+        self.mask = ~(word - 1)
+
         self.base_addr = base_addr & ~(mmap.PAGESIZE - 1)
         self.base_addr_offset = base_addr - self.base_addr
 
         stop = base_addr + length * self.word
-        if (stop % self.mask):
+        if stop % self.mask:
             stop = (stop + self.word) & ~(self.word - 1)
 
         self.length = stop - self.base_addr
         self.fname = filename
 
         # Check filesize (doesn't work with /dev/mem)
-        #filesize = os.stat(self.fname).st_size
-        #if (self.base_addr + self.length) > filesize:
+        # filesize = os.stat(self.fname).st_size
+        # if (self.base_addr + self.length) > filesize:
         #    self.length = filesize - self.base_addr
 
-        self.debug('init with base_addr = {0} and length = {1} on {2}'.
-                format(hex(self.base_addr), hex(self.length), self.fname))
+        self.debug(
+            "init with base_addr = {0} and length = {1} on {2}".format(
+                hex(self.base_addr), hex(self.length), self.fname
+            )
+        )
 
         # Open file and mmap
         self.f = os.open(self.fname, os.O_RDWR | os.O_SYNC)
-        self.mem = mmap.mmap(self.f, self.length, mmap.MAP_SHARED,
-                mmap.PROT_READ | mmap.PROT_WRITE,
-                offset=self.base_addr)
+        self.mem = mmap.mmap(
+            self.f,
+            self.length,
+            mmap.MAP_SHARED,
+            mmap.PROT_READ | mmap.PROT_WRITE,
+            offset=self.base_addr,
+        )
 
     def __del__(self):
         if self.f:
@@ -123,59 +124,59 @@ def __del__(self):
     def read(self, offset, length):
         """Read length number of words from offset"""
 
-        if offset < 0 or length < 0: raise AssertionError
+        if offset < 0 or length < 0:
+            raise AssertionError
 
-        # Make reading easier (and faster... won't resolve dot in loops)
-        mem = self.mem
-
-        self.debug('reading {0} bytes from offset {1}'.
-                   format(length * self.word, hex(offset)))
+        self.debug(
+            "reading {0} bytes from offset {1}".format(length * self.word, hex(offset))
+        )
 
         # Compensate for the base_address not being what the user requested
         # and then seek to the aligned offset.
         virt_base_addr = self.base_addr_offset & self.mask
-        mem.seek(virt_base_addr + offset)
+        self.mem.seek(virt_base_addr + offset)
 
-        # Read length words of size self.word and return it
-        data = []
-        for i in range(length):
-            data.append(struct.unpack('I', mem.read(self.word))[0])
+        # Read length words of size self.word and return bytes()
+        data = self.mem.read(length * self.word)
 
         abs_addr = self.base_addr + virt_base_addr
-        return DevMemBuffer(abs_addr + offset, data)
-
+        return DevMemBuffer(abs_addr + offset, data, self.word)
 
     def write(self, offset, din):
         """Write length number of words to offset"""
 
-        if offset < 0 or len(din) <= 0: raise AssertionError
-
-        self.debug('writing {0} bytes to offset {1}'.
-                format(len(din) * self.word, hex(offset)))
+        if offset < 0 or len(din) <= 0:
+            raise AssertionError
 
-        # Make reading easier (and faster... won't resolve dot in loops)
-        mem = self.mem
+        self.debug(
+            "writing {0} bytes to offset {1}".format(len(din) * self.word, hex(offset))
+        )
 
         # Compensate for the base_address not being what the user requested
         # fix double plus offset
-        #offset += self.base_addr_offset
+        # offset += self.base_addr_offset
 
         # Check that the operation is going write to an aligned location
-        if (offset & ~self.mask): raise AssertionError
+        if offset & ~self.mask:
+            raise AssertionError
 
         # Seek to the aligned offset
         virt_base_addr = self.base_addr_offset & self.mask
-        mem.seek(virt_base_addr + offset)
+        self.mem.seek(virt_base_addr + offset)
 
         # Read until the end of our aligned address
         for i in range(len(din)):
-            self.debug('writing at position = {0}: 0x{1:x}'.
-                        format(self.mem.tell(), din[i]))
+            self.debug(
+                "writing at position = {0}: 0x{1:x}".format(
+                    self.self.mem.tell(), din[i]
+                )
+            )
             # Write one word at a time
-            mem.write(struct.pack('I', din[i]))
+            self.mem.write(struct.pack("I", din[i]))
 
     def debug_set(self, value):
         self._debug = value
 
     def debug(self, debug_str):
-        if self._debug: print('DevMem Debug: {0}'.format(debug_str))
+        if self._debug:
+            print("DevMem Debug: {0}".format(debug_str), file=sys.stderr)
diff --git a/devmem/__main__.py b/devmem/__main__.py
index 30daf32..d236234 100644
--- a/devmem/__main__.py
+++ b/devmem/__main__.py
@@ -1,5 +1,8 @@
 #!/usr/bin/env python3
 
+# TODO delete when dropping Python 2 support
+from __future__ import print_function
+
 import argparse
 import os
 import sys
@@ -13,14 +16,12 @@
 # * https://bugs.python.org/issue22240
 # pip workaround:
 # * https://github.com/pypa/pip/blob/08c99b6e00135ca8df2e98db58aa0b701b971c64/src/pip/_internal/utils/misc.py#L124-L134
-def get_prog() -> str:
+def get_prog():
     """Determine the program name if invoked directly or as a module"""
 
-    name = (
-        sys.argv[0]
-        if globals().get("__spec__") is None
-        else __spec__.name.partition(".")[0]
-    )
+    spec = globals().get("__spec__")
+    name = sys.argv[0] if spec is None else spec.name.partition(".")[0]
+
     try:
         prog = os.path.basename(sys.argv[0])
         if prog in ("__main__.py", "-c"):
@@ -33,7 +34,7 @@ def get_prog() -> str:
     return name
 
 
-def main() -> int:
+def main():
     """Main function with useful demo application"""
 
     parser = argparse.ArgumentParser(prog=get_prog())
@@ -88,6 +89,13 @@ def main() -> int:
         "-d", "--debug", action="store_true", help="provide debugging information"
     )
 
+    parser.add_argument(
+        "-b",
+        "--binary",
+        action="store_true",
+        help="output data as a binary stream",
+    )
+
     args = parser.parse_args()
 
     # Check for sane arguments
@@ -105,15 +113,21 @@ def main() -> int:
     addr = args.write[0] if args.write else args.read
 
     # Create the Dev Mem object that does the magic
-    mem = DevMem(addr, length=args.num, filename=args.mmap, debug=args.debug)
+    mem = DevMem(
+        addr, length=args.num, filename=args.mmap, debug=args.debug, word=args.word_size
+    )
+
+    bytes_per_row = 16
+    words_per_row = int(bytes_per_row / args.word_size)
 
     # Perform the actual read or write
     if args.write is not None:
         if args.verbose:
             print(
                 "Value before write:\t{0}".format(
-                    mem.read(0x0, args.num).hexdump(args.word_size)
-                )
+                    mem.read(0x0, args.num).hexdump(words_per_row)
+                ),
+                file=sys.stderr,
             )
 
         mem.write(0x0, [args.write[1]])
@@ -121,11 +135,22 @@ def main() -> int:
         if args.verbose:
             print(
                 "Value after write:\t{0}".format(
-                    mem.read(0x0, args.num).hexdump(args.word_size)
-                )
+                    mem.read(0x0, args.num).hexdump(words_per_row)
+                ),
+                file=sys.stderr,
             )
+
+    elif args.binary:
+        buf = mem.read(0x0, args.num)
+
+        # TODO Delete when dropping Python2 support
+        if sys.version_info.major == 2:
+            sys.stdout.write(buf.data)
+        else:
+            sys.stdout.buffer.write(buf.data)
+
     else:
-        print(mem.read(0x0, args.num).hexdump(args.word_size))
+        print(mem.read(0x0, args.num).hexdump(words_per_row))
 
 
 if __name__ == "__main__":
diff --git a/test.py b/test.py
new file mode 100644
index 0000000..37c9cb4
--- /dev/null
+++ b/test.py
@@ -0,0 +1,22 @@
+import unittest
+import devmem
+
+class TestDevMem(unittest.TestCase):
+
+    def test_upper(self):
+        #devmem.main(...)
+        self.assertEqual('foo'.upper(), 'FOO')
+
+    def test_isupper(self):
+        self.assertTrue('FOO'.isupper())
+        self.assertFalse('Foo'.isupper())
+
+    def test_split(self):
+        s = 'hello world'
+        self.assertEqual(s.split(), ['hello', 'world'])
+        # check that s.split fails when the separator is not a string
+        with self.assertRaises(TypeError):
+            s.split(2)
+
+if __name__ == '__main__':
+    unittest.main()