From b740f2019108b2974fd650cb0aec5b91851eb564 Mon Sep 17 00:00:00 2001
From: Henning Peters <pete@dexterslab.de>
Date: Sun, 6 Mar 2016 09:19:27 +0100
Subject: [PATCH] hash_string() should not depend on python's internal unicode
 representation, also fixes https://github.com/spacy-io/sense2vec/issues/5 for
 py2

---
 spacy/strings.pyx | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/spacy/strings.pyx b/spacy/strings.pyx
index ef8422aa0..d54dcdf1a 100644
--- a/spacy/strings.pyx
+++ b/spacy/strings.pyx
@@ -23,10 +23,8 @@ import ujson as json
 
 
 cpdef hash_t hash_string(unicode string) except 0:
-    # This has to be like this for
-    chars = <char*>PyUnicode_AS_DATA(string)
-    size = PyUnicode_GET_DATA_SIZE(string)
-    return hash64(chars, size, 1)
+    chars = string.encode('utf8')
+    return hash64(<char*>chars, len(chars), 1)
 
 
 cdef unicode _decode(const Utf8Str* string):