Read-only python3 codepath in nbt.pyx

setup_mceditlib.py

@@ -23,7 +23,9 @@ include_dirs = [numpy.get_include()]
 mceditlib_ext_modules = cythonize([
     "src/mceditlib/nbt.pyx",
     "src/mceditlib/relight/with_cython.pyx"
-])
+],
+    compile_time_env={'IS_PY2': True},
+    )
 
 for m in mceditlib_ext_modules:
     m.include_dirs = include_dirs

src/mceditlib/nbt.pyx

@@ -38,25 +38,42 @@ DEF UNICODE_NAMES = True
 # For each NBT file loaded, cache all of the unicode strings used for tag names. Saves some hundred kilobytes per
 # file since tag names often appear multiple times
 
-DEF UNICODE_CACHE = True
+# The value of the IS_PY2 macro is received from the build script
+IF IS_PY2:
+    DEF UNICODE_CACHE = True
+ELSE:
+    # This codepath is currently unsupported in the python3 version
+    DEF UNICODE_CACHE = False
 
 import collections
 import gzip
 import zlib
 
-from cStringIO import StringIO
-from cpython cimport PyTypeObject, PyUnicode_DecodeUTF8, PyList_Append, PyString_FromStringAndSize
+from cpython cimport PyTypeObject, PyUnicode_DecodeUTF8, PyList_Append
+IF IS_PY2:
+    from cStringIO import StringIO
+    from cpython cimport PyString_FromStringAndSize
+    binary_type = str
+ELSE:
+    from io import BytesIO as StringIO
+    binary_type = bytes
+
 import numpy
 
-cdef extern from "cStringIO.h":
-    struct PycStringIO_CAPI:
-        int cwrite(object o, char * buf, Py_ssize_t len)
-        PyTypeObject * OutputType
-cdef extern from "cobject.h":
-    void * PyCObject_Import(char * module_name, char * cobject_name)
+IF IS_PY2:
+    cdef extern from "cStringIO.h":
+        struct PycStringIO_CAPI:
+            int cwrite(object o, char * buf, Py_ssize_t len)
+            PyTypeObject * OutputType
+    cdef extern from "cobject.h":
+        void * PyCObject_Import(char * module_name, char * cobject_name)
 
-cdef PycStringIO_CAPI *PycStringIO = <PycStringIO_CAPI *> PyCObject_Import("cStringIO", "cStringIO_CAPI")
-cdef PyTypeObject * StringO = PycStringIO.OutputType
+    cdef PycStringIO_CAPI *PycStringIO = <PycStringIO_CAPI *> PyCObject_Import("cStringIO", "cStringIO_CAPI")
+    cdef PyTypeObject * StringO = PycStringIO.OutputType
+ELSE:
+    # The equivalent python3 code has not been written, so for now only reading
+    # is supported in python3
+    pass
 
 # Tag IDs
 
@@ -131,7 +148,7 @@ cdef class TAG_Value:
 
         def __set__(self, val):
             IF UNICODE_NAMES:
-                if isinstance(val, str):
+                if isinstance(val, binary_type):
                     val = PyUnicode_DecodeUTF8(val, len(val), "strict")
             ELSE:
                 if isinstance(val, unicode):
@@ -327,7 +344,7 @@ cdef class TAG_String(TAG_Value):
             return self._value
 
         def __set__(self, value):
-            if isinstance(value, str):
+            if isinstance(value, binary_type):
                 value = PyUnicode_DecodeUTF8(value, len(value), "strict")
             self._value = value
 
@@ -509,7 +526,7 @@ cdef class _TAG_Compound(TAG_Value):
             return data
 
         if isinstance(filename_or_buf, basestring):
-            f = file(filename_or_buf, "wb")
+            f = open(filename_or_buf, "wb")
             f.write(data)
         else:
             filename_or_buf.write(data)
@@ -545,7 +562,7 @@ def load(filename="", buf=None):
     :rtype: TAG_Compound
     """
     if filename:
-        buf = file(filename, "rb")
+        buf = open(filename, "rb")
 
     if hasattr(buf, "read"):
         buf = buf.read()
@@ -807,7 +824,10 @@ def hexdump(src, length=8):
 
 cdef void cwrite(obj, char *buf, size_t len):
     #print "cwrite %s %s %d" % (map(ord, buf[:min(4, len)]), buf[:min(4, len)].decode('ascii', 'replace'), len)
-    PycStringIO.cwrite(obj, buf, len)
+    IF IS_PY2:
+        PycStringIO.cwrite(obj, buf, len)
+    ELSE:
+        raise Exception("[nbt.pyx python3 port] cwrite was not ported to python3")
 
 
 cdef void save_tag_id(char tagID, object buf):