added source_to_unicode method to deal with different encodings, davidhalter/jedi-vim#48

jedi/api.py

@@ -61,18 +61,14 @@ class Script(object):
     :type col: int
     :param source_path: The path in the os, the current module is in.
     :type source_path: string or None
-    :param source_encoding: encoding for decoding `source`, when it
+    :param source_encoding: encoding for decoding `source`, if it
                             is not a `unicode` object.
     :type source_encoding: string
     """
     def __init__(self, source, line, column, source_path,
                                  source_encoding='utf-8'):
         debug.reset_time()
-        try:
-            source = unicode(source, source_encoding, 'replace')
-            # Use 'replace' over 'ignore' to hold code structure.
-        except TypeError:  # `source` is already a unicode object
-            pass
+        source = modules.source_to_unicode(source, source_encoding)
         self.pos = line, column
         self.module = modules.ModuleWithCursor(source_path, source=source,
                                                             position=self.pos)

jedi/dynamic.py

@@ -42,7 +42,7 @@ def get_directory_modules_for_name(mods, name):
 
     def check_fs(path):
         with open(path) as f:
-            source = f.read()
+            source = modules.source_to_unicode(f.read())
             if name in source:
                 return modules.Module(path, source).parser.module
 

jedi/modules.py

@@ -26,7 +26,7 @@ class Module(builtin.CachedModule):
     """
     def __init__(self, path, source):
         super(Module, self).__init__(path=path)
-        self.source = source
+        self.source = source_to_unicode(source)
         self._line_cache = None
 
     def _get_source(self):
@@ -311,3 +311,33 @@ def detect_django_path(module_path):
         except IOError:
             pass
     return result
+
+
+def source_to_unicode(source, encoding=None):
+    def detect_encoding():
+        """ For the implementation of encoding definitions in Python, look at:
+        http://www.python.org/dev/peps/pep-0263/
+        http://docs.python.org/2/reference/lexical_analysis.html#encoding-\
+                                                                declarations
+        """
+        if encoding is not None:
+            return encoding
+
+        if source.startswith('\xef\xbb\xbf'):
+            # UTF-8 byte-order mark
+            return 'utf-8'
+
+        first_two_lines = re.match(r'(?:[^\n]*\n){0,2}', source).group(0)
+        possible_encoding = re.match("coding[=:]\s*([-\w.]+)", first_two_lines)
+        if possible_encoding:
+            return possible_encoding.group(1)
+        else:
+            # the default if nothing else has been set -> PEP 263
+            return 'iso-8859-1'
+
+    if isinstance(source, unicode):
+        # only cast str/bytes
+        return source
+
+    # cast to unicode by default
+    return unicode(source, detect_encoding(), 'replace')

test/regression.py

@@ -251,10 +251,22 @@ class TestRegression(Base):
         assert len(api.Script(s, 1, 15, '/').get_definition()) == 1
         assert len(api.Script(s, 1, 10, '/').get_definition()) == 1
 
+
     def test_unicode_script(self):
         """ normally no unicode objects are being used. (<=2.7) """
-        s = unicode("import datetime; datetime.d")
-        assert len(self.complete(s))
+        s = unicode("import datetime; datetime.timedelta")
+        completions = self.complete(s)
+        assert len(completions)
+        assert type(completions[0].description) is unicode
+
+        s = utf8("author='öä'; author")
+        completions = self.complete(s)
+        assert type(completions[0].description) is unicode
+
+        s = utf8("#-*- coding: iso-8859-1 -*-\nauthor='öä'; author")
+        s = s.encode('latin-1')
+        completions = self.complete(s)
+        assert type(completions[0].description) is unicode
 
     def test_multibyte_script(self):
         """ `jedi.Script` must accept multi-byte string source. """