comparators/apk: less kludgy way of detecting APKs

diffoscope/comparators/apk.py

@@ -146,10 +146,18 @@ class ApkContainer(Archive):
         return differences
 
 class ApkFile(File):
-    RE_FILE_TYPE = re.compile(r'^((Java|Zip) archive data|DOS/MBR boot sector).*\b')
+    RE_FILE_TYPE_FALLBACK_HEADER = b"PK\x03\x04"
+    RE_FILE_TYPE = re.compile(r'^(Java|Zip) archive data.*\b')
     RE_FILE_EXTENSION = re.compile(r'\.apk$')
     CONTAINER_CLASS = ApkContainer
 
+    @staticmethod
+    def recognizes(file):
+        if not ApkFile.RE_FILE_EXTENSION.search(file.name):
+            return False
+        return (ApkFile.RE_FILE_TYPE.search(file.magic_file_type) or
+                file.file_header[:4] == ApkFile.RE_FILE_TYPE_FALLBACK_HEADER)
+
     def compare_details(self, other, source=None):
         zipinfo_difference = Difference.from_command(Zipinfo, self.path, other.path) or \
                              Difference.from_command(ZipinfoVerbose, self.path, other.path)

diffoscope/comparators/utils/file.py

@@ -149,6 +149,13 @@ class File(object, metaclass=abc.ABCMeta):
             self._magic_file_type = File.guess_file_type(self.path)
         return self._magic_file_type
 
+    @property
+    def file_header(self):
+        if not hasattr(self, '_file_header'):
+            with open(self.path, 'rb') as f:
+                self._file_header = f.read(16)
+        return self._file_header
+
     @property
     def file_type(self):
         for x, y in (