fixes #46 support for bytes input for higher galois fields, thanks to…

… another-pjohnson

Signed-off-by: Stephen L. <lrq3000@gmail.com>

Makefile

@@ -65,6 +65,9 @@ testsetuppost:
 testcoverage:
      # This is the preferred way to run the tests since Python 3.10
 	@+make coverclean
+     # Build the Cython extension
+	python setup.py build_ext --inplace
+     # Run the tests
 	# nosetests reedsolo --with-coverage --cover-package=reedsolo --cover-erase --cover-min-percentage=80 -d -v
      # With PyTest, it is now necessary to first install the python module so that it is found (--cov=<module>)
      #python setup.py develop

README.rst

@@ -310,7 +310,7 @@ To use the Cython implementation, you need to ``pip install cython`` and a C++ c
 
 Then, use `import RSCodec from creedsolo` instead of importing from the `reedsolo` module, and finally only feed `bytearray()` objects to the `RSCodec` object. Exclusively using bytearrays is one of the reasons creedsolo is faster than reedsolo. You can convert any string by specifying the encoding: `bytearray("Hello World", "UTF-8")`.
 
-Note that there is an inherent limitation of the C implementation which cannot work with higher galois fields than 8 (= characters of max 255 value) because the C implementation only works with bytearrays, and bytearrays only support characters up to 255. If you want to use higher galois fields, you need to use the pure python version, or rewrite the C implementation to use lists instead of bytearrays (which will be MUCH slower so this defeats the purpose and you are better off simply using the pure python version under PyPy).
+Note that there is an inherent limitation of the C implementation which cannot work with higher galois fields than 8 (= characters of max 255 value) because the C implementation only works with bytearrays, and bytearrays only support characters up to 255. If you want to use higher galois fields, you need to use the pure python version, which includes a fake `_bytearray` function that overloads the standard bytearray in case galois fields higher than 8 are used to `init_tables()`, or rewrite the C implementation to use lists instead of bytearrays (which will be MUCH slower so this defeats the purpose and you are better off simply using the pure python version under PyPy - an older version of the C implementation was doing just that, and without bytearrays, all performance gains were lost, hence why the bytearrays were kept despite the limitations).
 
 Edge cases
 -------------

reedsolo.py

@@ -276,6 +276,8 @@ def _bytearray(obj = 0, encoding = "latin-1"):
                     raise(ValueError, "Type of object not recognized!")
             elif isinstance(obj, int):  # compatibility with list preallocation bytearray(int)
                 obj = [0] * obj
+            elif isinstance(obj, bytes):
+                obj = [int(b) for b in obj]
             # Else obj is a list of int, it's ok
             return array("i", obj)
 

tests/test_reedsolo.py

@@ -281,6 +281,13 @@ def test_multiple_RSCodec(self):
         bytearray(rs_256.decode(rs_256.encode(mes))[0])
         # At this point, there should not have been any exception raised!
 
+    def test_higher_galois_fields_bytes(self):
+        rs = RSCodec(12, c_exp=12) # same as nsize=4095
+        str_msg = "This is a message"
+        bytes_msg = b"This is a binary message"
+        result = rs.encode(str_msg) # this always worked
+        result_b = rs.encode(bytes_msg) # this is the dege case that used to fail
+
 class TestGFArithmetics(unittest.TestCase):
     '''Test Galois Field arithmetics'''