Backport PR #6081: don’t modify dict keys while iterating through them

in `jsonutil.rekey`, which could prevent some keys from being modified during the iteration.

closes  #5901

Fixes 'unorderable types' failure on Python 3 in test_client:
...

IPython/utils/jsonutil.py

@@ -1,243 +1,239 b''
 """Utilities to manipulate JSON objects."""
 
 # Copyright (c) IPython Development Team.
 # Distributed under the terms of the Modified BSD License.
 
 import math
 import re
 import types
 from datetime import datetime
 
 try:
     # base64.encodestring is deprecated in Python 3.x
     from base64 import encodebytes
 except ImportError:
     # Python 2.x
     from base64 import encodestring as encodebytes
 
 from IPython.utils import py3compat
 from IPython.utils.py3compat import string_types, unicode_type, iteritems
 from IPython.utils.encoding import DEFAULT_ENCODING
 next_attr_name = '__next__' if py3compat.PY3 else 'next'
 
 #-----------------------------------------------------------------------------
 # Globals and constants
 #-----------------------------------------------------------------------------
 
 # timestamp formats
 ISO8601 = "%Y-%m-%dT%H:%M:%S.%f"
 ISO8601_PAT=re.compile(r"^(\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2})(\.\d{1,6})?Z?([\+\-]\d{2}:?\d{2})?$")
 
 # holy crap, strptime is not threadsafe.
 # Calling it once at import seems to help.
 datetime.strptime("1", "%d")
 
 #-----------------------------------------------------------------------------
 # Classes and functions
 #-----------------------------------------------------------------------------
 
 def rekey(dikt):
     """Rekey a dict that has been forced to use str keys where there should be
     ints by json."""
-    for k in dikt:
+    for k in list(dikt):
         if isinstance(k, string_types):
-            ik=fk=None
+            nk = None
             try:
-                ik = int(k)
+                nk = int(k)
             except ValueError:
                 try:
-                    fk = float(k)
+                    nk = float(k)
                 except ValueError:
                     continue
-            if ik is not None:
-                nk = ik
-            else:
-                nk = fk
             if nk in dikt:
                 raise KeyError("already have key %r"%nk)
             dikt[nk] = dikt.pop(k)
     return dikt
 
 def parse_date(s):
     """parse an ISO8601 date string
     
     If it is None or not a valid ISO8601 timestamp,
     it will be returned unmodified.
     Otherwise, it will return a datetime object.
     """
     if s is None:
         return s
     m = ISO8601_PAT.match(s)
     if m:
         # FIXME: add actual timezone support
         # this just drops the timezone info
         notz, ms, tz = m.groups()
         if not ms:
             ms = '.0'
         notz = notz + ms
         return datetime.strptime(notz, ISO8601)
     return s
 
 def extract_dates(obj):
     """extract ISO8601 dates from unpacked JSON"""
     if isinstance(obj, dict):
         new_obj = {} # don't clobber
         for k,v in iteritems(obj):
             new_obj[k] = extract_dates(v)
         obj = new_obj
     elif isinstance(obj, (list, tuple)):
         obj = [ extract_dates(o) for o in obj ]
     elif isinstance(obj, string_types):
         obj = parse_date(obj)
     return obj
 
 def squash_dates(obj):
     """squash datetime objects into ISO8601 strings"""
     if isinstance(obj, dict):
         obj = dict(obj) # don't clobber
         for k,v in iteritems(obj):
             obj[k] = squash_dates(v)
     elif isinstance(obj, (list, tuple)):
         obj = [ squash_dates(o) for o in obj ]
     elif isinstance(obj, datetime):
         obj = obj.isoformat()
     return obj
 
 def date_default(obj):
     """default function for packing datetime objects in JSON."""
     if isinstance(obj, datetime):
         return obj.isoformat()
     else:
         raise TypeError("%r is not JSON serializable"%obj)
 
 
 # constants for identifying png/jpeg data
 PNG = b'\x89PNG\r\n\x1a\n'
 # front of PNG base64-encoded
 PNG64 = b'iVBORw0KG'
 JPEG = b'\xff\xd8'
 # front of JPEG base64-encoded
 JPEG64 = b'/9'
 # front of PDF base64-encoded
 PDF64 = b'JVBER'
 
 def encode_images(format_dict):
     """b64-encodes images in a displaypub format dict
 
     Perhaps this should be handled in json_clean itself?
 
     Parameters
     ----------
 
     format_dict : dict
         A dictionary of display data keyed by mime-type
 
     Returns
     -------
 
     format_dict : dict
         A copy of the same dictionary,
         but binary image data ('image/png', 'image/jpeg' or 'application/pdf')
         is base64-encoded.
 
     """
     encoded = format_dict.copy()
 
     pngdata = format_dict.get('image/png')
     if isinstance(pngdata, bytes):
         # make sure we don't double-encode
         if not pngdata.startswith(PNG64):
             pngdata = encodebytes(pngdata)
         encoded['image/png'] = pngdata.decode('ascii')
 
     jpegdata = format_dict.get('image/jpeg')
     if isinstance(jpegdata, bytes):
         # make sure we don't double-encode
         if not jpegdata.startswith(JPEG64):
             jpegdata = encodebytes(jpegdata)
         encoded['image/jpeg'] = jpegdata.decode('ascii')
 
     pdfdata = format_dict.get('application/pdf')
     if isinstance(pdfdata, bytes):
         # make sure we don't double-encode
         if not pdfdata.startswith(PDF64):
             pdfdata = encodebytes(pdfdata)
         encoded['application/pdf'] = pdfdata.decode('ascii')
 
     return encoded
 
 
 def json_clean(obj):
     """Clean an object to ensure it's safe to encode in JSON.
 
     Atomic, immutable objects are returned unmodified.  Sets and tuples are
     converted to lists, lists are copied and dicts are also copied.
 
     Note: dicts whose keys could cause collisions upon encoding (such as a dict
     with both the number 1 and the string '1' as keys) will cause a ValueError
     to be raised.
 
     Parameters
     ----------
     obj : any python object
 
     Returns
     -------
     out : object
 
       A version of the input which will not cause an encoding error when
       encoded as JSON.  Note that this function does not *encode* its inputs,
       it simply sanitizes it so that there will be no encoding errors later.
 
     """
     # types that are 'atomic' and ok in json as-is.
     atomic_ok = (unicode_type, type(None))
 
     # containers that we need to convert into lists
     container_to_list = (tuple, set, types.GeneratorType)
 
     if isinstance(obj, float):
         # cast out-of-range floats to their reprs
         if math.isnan(obj) or math.isinf(obj):
             return repr(obj)
         return float(obj)
     
     if isinstance(obj, int):
         # cast int to int, in case subclasses override __str__ (e.g. boost enum, #4598)
         if isinstance(obj, bool):
             # bools are ints, but we don't want to cast them to 0,1
             return obj
         return int(obj)
 
     if isinstance(obj, atomic_ok):
         return obj
 
     if isinstance(obj, bytes):
         return obj.decode(DEFAULT_ENCODING, 'replace')
 
     if isinstance(obj, container_to_list) or (
         hasattr(obj, '__iter__') and hasattr(obj, next_attr_name)):
         obj = list(obj)
 
     if isinstance(obj, list):
         return [json_clean(x) for x in obj]
 
     if isinstance(obj, dict):
         # First, validate that the dict won't lose data in conversion due to
         # key collisions after stringification.  This can happen with keys like
         # True and 'true' or 1 and '1', which collide in JSON.
         nkeys = len(obj)
         nkeys_collapsed = len(set(map(unicode_type, obj)))
         if nkeys != nkeys_collapsed:
             raise ValueError('dict cannot be safely converted to JSON: '
                              'key collision would lead to dropped values')
         # If all OK, proceed by making the new dict that will be json-safe
         out = {}
         for k,v in iteritems(obj):
             out[unicode_type(k)] = json_clean(v)
         return out
 
     # If we get here, we don't know how to handle the object, so we just get
     # its repr and return that.  This will catch lambdas, open sockets, class
     # objects, and any other complicated contraption that json can't encode
     return repr(obj)

IPython/utils/tests/test_jsonutil.py

@@ -1,144 +1,151 b''
 # coding: utf-8
 """Test suite for our JSON utilities."""
 
 # Copyright (c) IPython Development Team.
 # Distributed under the terms of the Modified BSD License.
 
 import datetime
 import json
 from base64 import decodestring
 
 import nose.tools as nt
 
 from IPython.utils import jsonutil, tz
 from ..jsonutil import json_clean, encode_images
 from ..py3compat import unicode_to_str, str_to_bytes, iteritems
 
 
 class Int(int):
     def __str__(self):
         return 'Int(%i)' % self
 
 def test():
     # list of input/expected output.  Use None for the expected output if it
     # can be the same as the input.
     pairs = [(1, None), # start with scalars
              (1.0, None),
              ('a', None),
              (True, None),
              (False, None),
              (None, None),
              # complex numbers for now just go to strings, as otherwise they
              # are unserializable
              (1j, '1j'),
              # Containers
              ([1, 2], None),
              ((1, 2), [1, 2]),
              (set([1, 2]), [1, 2]),
              (dict(x=1), None),
              ({'x': 1, 'y':[1,2,3], '1':'int'}, None),
              # More exotic objects
              ((x for x in range(3)), [0, 1, 2]),
              (iter([1, 2]), [1, 2]),
              (Int(5), 5),
              ]
     
     for val, jval in pairs:
         if jval is None:
             jval = val
         out = json_clean(val)
         # validate our cleanup
         nt.assert_equal(out, jval)
         # and ensure that what we return, indeed encodes cleanly
         json.loads(json.dumps(out))
 
 
+def test_rekey():
+    # This could fail due to modifying the dict keys in-place on Python 3
+    d = { i:i for i in map(str, range(128)) }
+    d = jsonutil.rekey(d)
+    for key in d:
+        nt.assert_is_instance(key, int)
+
 
 def test_encode_images():
     # invalid data, but the header and footer are from real files
     pngdata = b'\x89PNG\r\n\x1a\nblahblahnotactuallyvalidIEND\xaeB`\x82'
     jpegdata = b'\xff\xd8\xff\xe0\x00\x10JFIFblahblahjpeg(\xa0\x0f\xff\xd9'
     pdfdata = b'%PDF-1.\ntrailer<</Root<</Pages<</Kids[<</MediaBox[0 0 3 3]>>]>>>>>>'
     
     fmt = {
         'image/png'  : pngdata,
         'image/jpeg' : jpegdata,
         'application/pdf' : pdfdata
     }
     encoded = encode_images(fmt)
     for key, value in iteritems(fmt):
         # encoded has unicode, want bytes
         decoded = decodestring(encoded[key].encode('ascii'))
         nt.assert_equal(decoded, value)
     encoded2 = encode_images(encoded)
     nt.assert_equal(encoded, encoded2)
     
     b64_str = {}
     for key, encoded in iteritems(encoded):
         b64_str[key] = unicode_to_str(encoded)
     encoded3 = encode_images(b64_str)
     nt.assert_equal(encoded3, b64_str)
     for key, value in iteritems(fmt):
         # encoded3 has str, want bytes
         decoded = decodestring(str_to_bytes(encoded3[key]))
         nt.assert_equal(decoded, value)
 
 def test_lambda():
     jc = json_clean(lambda : 1)
     nt.assert_is_instance(jc, str)
     nt.assert_in('<lambda>', jc)
     json.dumps(jc)
 
 def test_extract_dates():
     timestamps = [
         '2013-07-03T16:34:52.249482',
         '2013-07-03T16:34:52.249482Z',
         '2013-07-03T16:34:52.249482Z-0800',
         '2013-07-03T16:34:52.249482Z+0800',
         '2013-07-03T16:34:52.249482Z+08:00',
         '2013-07-03T16:34:52.249482Z-08:00',
         '2013-07-03T16:34:52.249482-0800',
         '2013-07-03T16:34:52.249482+0800',
         '2013-07-03T16:34:52.249482+08:00',
         '2013-07-03T16:34:52.249482-08:00',
     ]
     extracted = jsonutil.extract_dates(timestamps)
     ref = extracted[0]
     for dt in extracted:
         nt.assert_true(isinstance(dt, datetime.datetime))
         nt.assert_equal(dt, ref)
 
 def test_parse_ms_precision():
     base = '2013-07-03T16:34:52'
     digits = '1234567890'
     
     parsed = jsonutil.parse_date(base)
     nt.assert_is_instance(parsed, datetime.datetime)
     for i in range(len(digits)):
         ts = base + '.' + digits[:i]
         parsed = jsonutil.parse_date(ts)
         if i >= 1 and i <= 6:
             nt.assert_is_instance(parsed, datetime.datetime)
         else:
             nt.assert_is_instance(parsed, str)
 
 def test_date_default():
     data = dict(today=datetime.datetime.now(), utcnow=tz.utcnow())
     jsondata = json.dumps(data, default=jsonutil.date_default)
     nt.assert_in("+00", jsondata)
     nt.assert_equal(jsondata.count("+00"), 1)
     extracted = jsonutil.extract_dates(json.loads(jsondata))
     for dt in extracted.values():
         nt.assert_is_instance(dt, datetime.datetime)
 
 def test_exception():
     bad_dicts = [{1:'number', '1':'string'},
                  {True:'bool', 'True':'string'},
                  ]
     for d in bad_dicts:
         nt.assert_raises(ValueError, json_clean, d)
 
 def test_unicode_dict():
     data = {u'üniço∂e': u'üniço∂e'}
     clean = jsonutil.json_clean(data)
     nt.assert_equal(data, clean)