Merge pull request #88 from nahuelhds/feature/encoding-autodetection

Encoding autodetection

diffengine/__init__.py

@@ -37,6 +37,7 @@
 
 from diffengine.exceptions.webdriver import UnknownWebdriverError
 from diffengine.exceptions.twitter import ConfigNotFoundError, TwitterError
+from diffengine.text import to_utf8
 from diffengine.twitter import TwitterHandler
 from diffengine.exceptions.sendgrid import (
     ConfigNotFoundError as SGConfigNotFoundError,
@@ -169,7 +170,7 @@ def get_latest(self):
             logging.warn("Got %s when fetching %s", resp.status_code, self.url)
             return None
 
-        doc = readability.Document(resp.text)
+        doc = readability.Document(to_utf8(resp.text))
         title = doc.title()
         summary = doc.summary(html_partial=True)
         summary = bleach.clean(summary, tags=["p"], strip=True)

diffengine/text_builder.py → diffengine/text.py

@@ -58,3 +58,17 @@ def build_with_default_content(diff):
         text = text[0:225] + "…"
     text += " " + diff.url
     return text
+
+
+def to_utf8(text):
+    for encoding in ["latin1", "ascii"]:
+        try:
+            result = text.encode(encoding).decode("utf8", "strict")
+            break
+        except (UnicodeEncodeError, UnicodeDecodeError):
+            result = None
+
+    if result is None:
+        return text
+
+    return result

diffengine/twitter.py

@@ -3,7 +3,7 @@
 
 from datetime import datetime
 
-from diffengine.text_builder import build_text
+from diffengine.text import build_text
 from diffengine.exceptions.twitter import (
     AlreadyTweetedError,
     ConfigNotFoundError,

test_diffengine.py

@@ -33,7 +33,7 @@
     AlreadyEmailedError as SGAlreadyEmailedError,
     ArchiveUrlNotFoundError as SGArchiveNotFoundError,
 )
-from diffengine.text_builder import build_text
+from diffengine.text import build_text
 from diffengine.exceptions.twitter import (
     ConfigNotFoundError,
     TokenNotFoundError,
@@ -619,8 +619,8 @@ def get_mocked_diff(with_archive_urls=True):
 
 class TextBuilderTest(TestCase):
     @patch("logging.warning")
-    @patch("diffengine.text_builder.build_with_lang")
-    @patch("diffengine.text_builder.build_with_default_content")
+    @patch("diffengine.text.build_with_lang")
+    @patch("diffengine.text.build_with_default_content")
     def test_build_with_default_content_when_no_lang_given(
         self, mocked_build_with_default_content, mocked_build_from_lang, mocked_warning
     ):
@@ -635,8 +635,8 @@ def test_build_with_default_content_when_no_lang_given(
         mocked_build_from_lang.assert_not_called()
 
     @patch("logging.warning")
-    @patch("diffengine.text_builder.build_with_lang")
-    @patch("diffengine.text_builder.build_with_default_content")
+    @patch("diffengine.text.build_with_lang")
+    @patch("diffengine.text.build_with_default_content")
     def test_build_with_default_content_when_lang_is_incomplete(
         self, mocked_build_with_default_content, mocked_build_from_lang, mocked_warning
     ):
@@ -656,8 +656,8 @@ def test_build_with_default_content_when_lang_is_incomplete(
         mocked_build_from_lang.assert_not_called()
 
     @patch("logging.warning")
-    @patch("diffengine.text_builder.build_with_lang")
-    @patch("diffengine.text_builder.build_with_default_content")
+    @patch("diffengine.text.build_with_lang")
+    @patch("diffengine.text.build_with_default_content")
     def test_build_with_lang_when_lang_given(
         self, mocked_build_with_default_content, mocked_build_from_lang, mocked_warning
     ):
@@ -678,7 +678,7 @@ def test_build_with_lang_when_lang_given(
         mocked_build_with_default_content.assert_not_called()
         mocked_build_from_lang.assert_called_once()
 
-    @patch("diffengine.text_builder.build_with_lang")
+    @patch("diffengine.text.build_with_lang")
     def test_default_content_text(self, mocked_build_from_lang):
         diff = get_mocked_diff()
         type(diff.new).title = "Test"
@@ -760,3 +760,16 @@ def test_lang_content_text(self):
         self.assertEqual(
             text, "change in the URL, the title and the summary\n%s" % diff.url
         )
+
+
+class EncodingTest(TestCase):
+    def test_utf8_do_nothingg(self):
+        text_utf8 = "Me preocupa más la parte futbolística"
+        result = to_utf8(text_utf8)
+        self.assertEquals(result, text_utf8)
+
+    def test_latin1_to_utf8(self):
+        text_latin = "Me preocupa más la parte futbolística"
+        text_utf8 = "Me preocupa más la parte futbolística"
+        result = to_utf8(text_latin)
+        self.assertEquals(result, text_utf8)