From 99eac54b310309702abc9d91f001ce75b70f4377 Mon Sep 17 00:00:00 2001
From: Rob Speer <rob@luminoso.com>
Date: Wed, 4 Jan 2017 18:51:00 -0500
Subject: [PATCH] transliterate: Handle unexpected Russian invasions

---
 tests/test_serbian.py     | 13 +++++++++++++
 wordfreq/transliterate.py | 23 +++++++++++++++++++++++
 2 files changed, 36 insertions(+)

diff --git a/tests/test_serbian.py b/tests/test_serbian.py
index bc64389..7d33367 100644
--- a/tests/test_serbian.py
+++ b/tests/test_serbian.py
@@ -10,3 +10,16 @@ def test_transliteration():
     eq_(tokenize("Pa, ima tu mnogo stvari koje ne shvataš.", 'sr'),
         ['pa', 'ima', 'tu', 'mnogo', 'stvari', 'koje', 'ne', 'shvataš'])
 
+
+def test_actually_russian():
+    # This looks mostly like Serbian, but was probably actually Russian.
+    # In Russian, Google Translate says it means:
+    # "a hundred out of a hundred, boys!"
+    #
+    # We make sure to handle this case so we don't end up with a mixed-script
+    # word like "pacanы".
+
+    eq_(tokenize("сто из ста, пацаны!", 'sr'),
+        ['sto', 'iz', 'sta', 'pacany'])
+
+    eq_(tokenize("культуры", 'sr'), ["kul'tury"])
diff --git a/wordfreq/transliterate.py b/wordfreq/transliterate.py
index 6c5224c..ba676c1 100644
--- a/wordfreq/transliterate.py
+++ b/wordfreq/transliterate.py
@@ -31,6 +31,29 @@ SR_CYRL_TO_LATN_DICT = {
     ord('Ч'): 'Č',   ord('ч'): 'č',
     ord('Џ'): 'Dž',  ord('џ'): 'dž',
     ord('Ш'): 'Š',   ord('ш'): 'š',
+
+    # Handle borrowed letters from Russian
+    ord('Ё'): 'Jo',  ord('ё'): 'Jo',
+    ord('Й'): 'J',   ord('й'): 'j',
+    ord('Ў'): 'U',   ord('ў'): 'u',
+    ord('Щ'): 'Šč',  ord('щ'): 'šč',
+    ord('Ъ'): '',    ord('ъ'): '',
+    ord('Ы'): 'Y',   ord('ы'): 'y',
+    ord('Ь'): "'",   ord('ь'): "'",
+    ord('Э'): 'E',   ord('э'): 'e',
+    ord('Ю'): 'Ju',  ord('ю'): 'ju',
+    ord('Я'): 'Ja',  ord('я'): 'ja',
+
+    # Handle borrowed letters from Ukrainian
+    ord('Є'): 'Je',  ord('є'): 'je',
+    ord('І'): 'I',   ord('і'): 'i',
+    ord('Ї'): 'Ji',  ord('ї'): 'ji',
+    ord('Ґ'): 'G',   ord('ґ'): 'g',
+
+    # Handle borrowed letters from Macedonian
+    ord('Ѕ'): 'Dz',  ord('ѕ'): 'dz',
+    ord('Ѓ'): 'Ǵ',   ord('ѓ'): 'ǵ',
+    ord('Ќ'): 'Ḱ',   ord('ќ'): 'ḱ',
 }