Blob Blame History Raw
From 50062c4d8c4108d17b7f12d9518ce883956d3921 Mon Sep 17 00:00:00 2001
From: David Lord <davidism@gmail.com>
Date: Tue, 10 Apr 2018 09:29:48 -0700
Subject: [PATCH] detect UTF encodings when loading json

(cherry picked from commit 0e1e9a04aaf29ab78f721cfc79ac2a691f6e3929)
---
 flask/json.py         | 49 ++++++++++++++++++++++++++++++++++++++++++-
 flask/wrappers.py     | 13 +++---------
 tests/test_helpers.py | 28 ++++++++++++++-----------
 3 files changed, 67 insertions(+), 23 deletions(-)

diff --git a/flask/json.py b/flask/json.py
index 16e0c29..114873e 100644
--- a/flask/json.py
+++ b/flask/json.py
@@ -8,6 +8,7 @@
     :copyright: (c) 2015 by Armin Ronacher.
     :license: BSD, see LICENSE for more details.
 """
+import codecs
 import io
 import uuid
 from datetime import date
@@ -108,6 +109,49 @@ def _load_arg_defaults(kwargs):
         kwargs.setdefault('cls', JSONDecoder)
 
 
+def detect_encoding(data):
+    """Detect which UTF codec was used to encode the given bytes.
+
+    The latest JSON standard (:rfc:`8259`) suggests that only UTF-8 is
+    accepted. Older documents allowed 8, 16, or 32. 16 and 32 can be big
+    or little endian. Some editors or libraries may prepend a BOM.
+
+    :param data: Bytes in unknown UTF encoding.
+    :return: UTF encoding name
+    """
+    head = data[:4]
+
+    if head[:3] == codecs.BOM_UTF8:
+        return 'utf-8-sig'
+
+    if b'\x00' not in head:
+        return 'utf-8'
+
+    if head in (codecs.BOM_UTF32_BE, codecs.BOM_UTF32_LE):
+        return 'utf-32'
+
+    if head[:2] in (codecs.BOM_UTF16_BE, codecs.BOM_UTF16_LE):
+        return 'utf-16'
+
+    if len(head) == 4:
+        if head[:3] == b'\x00\x00\x00':
+            return 'utf-32-be'
+
+        if head[::2] == b'\x00\x00':
+            return 'utf-16-be'
+
+        if head[1:] == b'\x00\x00\x00':
+            return 'utf-32-le'
+
+        if head[1::2] == b'\x00\x00':
+            return 'utf-16-le'
+
+    if len(head) == 2:
+        return 'utf-16-be' if head.startswith(b'\x00') else 'utf-16-le'
+
+    return 'utf-8'
+
+
 def dumps(obj, **kwargs):
     """Serialize ``obj`` to a JSON formatted ``str`` by using the application's
     configured encoder (:attr:`~flask.Flask.json_encoder`) if there is an
@@ -142,7 +186,10 @@ def loads(s, **kwargs):
     """
     _load_arg_defaults(kwargs)
     if isinstance(s, bytes):
-        s = s.decode(kwargs.pop('encoding', None) or 'utf-8')
+        encoding = kwargs.pop('encoding', None)
+        if encoding is None:
+            encoding = detect_encoding(s)
+        s = s.decode(encoding)
     return _json.loads(s, **kwargs)
 
 
diff --git a/flask/wrappers.py b/flask/wrappers.py
index 04bdcb5..3e600fc 100644
--- a/flask/wrappers.py
+++ b/flask/wrappers.py
@@ -144,17 +144,10 @@ class Request(RequestBase):
         if not (force or self.is_json):
             return None
 
-        # We accept a request charset against the specification as
-        # certain clients have been using this in the past.  This
-        # fits our general approach of being nice in what we accept
-        # and strict in what we send out.
-        request_charset = self.mimetype_params.get('charset')
+        data = _get_data(self, cache)
+
         try:
-            data = _get_data(self, cache)
-            if request_charset is not None:
-                rv = json.loads(data, encoding=request_charset)
-            else:
-                rv = json.loads(data)
+            rv = json.loads(data)
         except ValueError as e:
             if silent:
                 rv = None
diff --git a/tests/test_helpers.py b/tests/test_helpers.py
index 9320ef7..9990782 100644
--- a/tests/test_helpers.py
+++ b/tests/test_helpers.py
@@ -21,6 +21,8 @@ from werkzeug.datastructures import Range
 from werkzeug.exceptions import BadRequest, NotFound
 from werkzeug.http import parse_cache_control_header, parse_options_header
 from werkzeug.http import http_date
+
+from flask import json
 from flask._compat import StringIO, text_type
 
 
@@ -34,6 +36,20 @@ def has_encoding(name):
 
 
 class TestJSON(object):
+    @pytest.mark.parametrize('value', (
+        1, 't', True, False, None,
+        [], [1, 2, 3],
+        {}, {'foo': u'🐍'},
+    ))
+    @pytest.mark.parametrize('encoding', (
+        'utf-8', 'utf-8-sig',
+        'utf-16-le', 'utf-16-be', 'utf-16',
+        'utf-32-le', 'utf-32-be', 'utf-32',
+    ))
+    def test_detect_encoding(self, value, encoding):
+        data = json.dumps(value).encode(encoding)
+        assert json.detect_encoding(data) == encoding
+        assert json.loads(data) == value
 
     def test_ignore_cached_json(self):
         app = flask.Flask(__name__)
@@ -85,18 +101,6 @@ class TestJSON(object):
         rv = c.post('/json', data='"foo"', content_type='application/x+json')
         assert rv.data == b'foo'
 
-    def test_json_body_encoding(self):
-        app = flask.Flask(__name__)
-        app.testing = True
-        @app.route('/')
-        def index():
-            return flask.request.get_json()
-
-        c = app.test_client()
-        resp = c.get('/', data=u'"Hällo Wörld"'.encode('iso-8859-15'),
-                     content_type='application/json; charset=iso-8859-15')
-        assert resp.data == u'Hällo Wörld'.encode('utf-8')
-
     def test_json_as_unicode(self):
         app = flask.Flask(__name__)
 
-- 
2.17.1