[ovs-dev] [PATCH 2/2] JSON serialization via Python's json lib

Terry Wilson twilson at redhat.com
Fri May 6 13:18:52 UTC 2016


There is no particularly good reason to use our own Python JSON
serialization implementation when serialization can be done faster
with Python's built-in JSON library.

A few tests were changed due to Python's default JSON library
returning slightly more precise floating point numbers and returning
'0.0' for 1e-9999 where the in-tree version returns '0'.
---
 python/ovs/json.py | 106 +++++------------------------------------------------
 tests/json.at      |  38 +++++++++++++++----
 2 files changed, 40 insertions(+), 104 deletions(-)

diff --git a/python/ovs/json.py b/python/ovs/json.py
index f1a6499..793ac17 100644
--- a/python/ovs/json.py
+++ b/python/ovs/json.py
@@ -12,11 +12,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from __future__ import absolute_import
+import functools
+import json
 import re
 import sys
 
 import six
-from six.moves import range
 
 try:
     import ovs._json
@@ -25,112 +27,24 @@ except ImportError:
 
 __pychecker__ = 'no-stringiter'
 
-escapes = {ord('"'): u"\\\"",
-           ord("\\"): u"\\\\",
-           ord("\b"): u"\\b",
-           ord("\f"): u"\\f",
-           ord("\n"): u"\\n",
-           ord("\r"): u"\\r",
-           ord("\t"): u"\\t"}
-for esc in range(32):
-    if esc not in escapes:
-        escapes[esc] = u"\\u%04x" % esc
-
 SPACES_PER_LEVEL = 2
-
-
-class _Serializer(object):
-    def __init__(self, stream, pretty, sort_keys):
-        self.stream = stream
-        self.pretty = pretty
-        self.sort_keys = sort_keys
-        self.depth = 0
-
-    def __serialize_string(self, s):
-        self.stream.write(u'"%s"' % ''.join(escapes.get(ord(c), c) for c in s))
-
-    def __indent_line(self):
-        if self.pretty:
-            self.stream.write('\n')
-            self.stream.write(' ' * (SPACES_PER_LEVEL * self.depth))
-
-    def serialize(self, obj):
-        if obj is None:
-            self.stream.write(u"null")
-        elif obj is False:
-            self.stream.write(u"false")
-        elif obj is True:
-            self.stream.write(u"true")
-        elif isinstance(obj, six.integer_types):
-            self.stream.write(u"%d" % obj)
-        elif isinstance(obj, float):
-            self.stream.write("%.15g" % obj)
-        elif isinstance(obj, six.text_type):
-            # unicode() on Python 2, or str() in Python 3 (always unicode)
-            self.__serialize_string(obj)
-        elif isinstance(obj, str):
-            # This is for Python 2, where this comes out to unicode(str()).
-            # For Python 3, it's str(str()), but it's harmless.
-            self.__serialize_string(six.text_type(obj))
-        elif isinstance(obj, dict):
-            self.stream.write(u"{")
-
-            self.depth += 1
-            self.__indent_line()
-
-            if self.sort_keys:
-                items = sorted(obj.items())
-            else:
-                items = six.iteritems(obj)
-            for i, (key, value) in enumerate(items):
-                if i > 0:
-                    self.stream.write(u",")
-                    self.__indent_line()
-                self.__serialize_string(six.text_type(key))
-                self.stream.write(u":")
-                if self.pretty:
-                    self.stream.write(u' ')
-                self.serialize(value)
-
-            self.stream.write(u"}")
-            self.depth -= 1
-        elif isinstance(obj, (list, tuple)):
-            self.stream.write(u"[")
-            self.depth += 1
-
-            if obj:
-                self.__indent_line()
-
-                for i, value in enumerate(obj):
-                    if i > 0:
-                        self.stream.write(u",")
-                        self.__indent_line()
-                    self.serialize(value)
-
-            self.depth -= 1
-            self.stream.write(u"]")
-        else:
-            raise Exception("can't serialize %s as JSON" % obj)
+dumper = functools.partial(json.dumps, separators=(",", ":"),
+                           ensure_ascii=False)
 
 
 def to_stream(obj, stream, pretty=False, sort_keys=True):
-    _Serializer(stream, pretty, sort_keys).serialize(obj)
+    stream.write(dumper(obj, indent=SPACES_PER_LEVEL if pretty else None,
+                        sort_keys=sort_keys))
 
 
 def to_file(obj, name, pretty=False, sort_keys=True):
-    stream = open(name, "w")
-    try:
+    with open(name, "w") as stream:
         to_stream(obj, stream, pretty, sort_keys)
-    finally:
-        stream.close()
 
 
 def to_string(obj, pretty=False, sort_keys=True):
-    output = six.StringIO()
-    to_stream(obj, output, pretty, sort_keys)
-    s = output.getvalue()
-    output.close()
-    return s
+    return dumper(obj, indent=SPACES_PER_LEVEL if pretty else None,
+                  sort_keys=sort_keys)
 
 
 def from_stream(stream):
diff --git a/tests/json.at b/tests/json.at
index 32d7fff..57a97b0 100644
--- a/tests/json.at
+++ b/tests/json.at
@@ -41,6 +41,12 @@ m4_define([JSON_CHECK_POSITIVE],
    JSON_CHECK_POSITIVE_PY([$1 - Python3], [$2], [$3], [$4],
                           [$HAVE_PYTHON3], [$PYTHON3])])
 
+m4_define([JSON_CHECK_POSITIVE_PY23],
+  [JSON_CHECK_POSITIVE_PY([$1 - Python2], [$2], [$3], [$4],
+                          [$HAVE_PYTHON], [$PYTHON])
+   JSON_CHECK_POSITIVE_PY([$1 - Python3], [$2], [$3], [$4],
+                          [$HAVE_PYTHON3], [$PYTHON3])])
+
 m4_define([JSON_CHECK_NEGATIVE_C],
   [AT_SETUP([$1])
    AT_KEYWORDS([json negative])
@@ -216,15 +222,23 @@ JSON_CHECK_POSITIVE(
 # It seems likely that the following test will fail on some system that
 # rounds slightly differently in arithmetic or in printf, but I'd like
 # to keep it this way until we run into such a system.
-JSON_CHECK_POSITIVE(
-  [large integers that overflow to reals], 
+JSON_CHECK_POSITIVE_C(
+  [C - large integers that overflow to reals], 
   [[[9223372036854775807000, -92233720368547758080000]]],
   [[[9.22337203685478e+21,-9.22337203685478e+22]]])
+JSON_CHECK_POSITIVE_PY23(
+  [large integers that overflow to reals], 
+  [[[9223372036854775807000, -92233720368547758080000]]],
+  [[[9.223372036854776e+21,-9.223372036854776e+22]]])
 
-JSON_CHECK_POSITIVE(
-  [negative zero],
+JSON_CHECK_POSITIVE_C(
+  [C - negative zero],
   [[[-0, -0.0, 1e-9999, -1e-9999]]],
   [[[0,0,0,0]]])
+JSON_CHECK_POSITIVE_PY23(
+  [negative zero],
+  [[[-0, -0.0, 1e-9999, -1e-9999]]],
+  [[[0,0,0.0,0.0]]])
 
 JSON_CHECK_POSITIVE(
   [reals], 
@@ -237,10 +251,14 @@ JSON_CHECK_POSITIVE(
 # It seems likely that the following test will fail on some system that
 # rounds slightly differently in arithmetic or in printf, but I'd like
 # to keep it this way until we run into such a system.
-JSON_CHECK_POSITIVE(
-  [+/- DBL_MAX],
+JSON_CHECK_POSITIVE_C(
+  [C - +/- DBL_MAX],
   [[[1.7976931348623157e+308, -1.7976931348623157e+308]]],
   [[[1.79769313486232e+308,-1.79769313486232e+308]]])
+JSON_CHECK_POSITIVE_PY23(
+  [+/- DBL_MAX],
+  [[[1.7976931348623157e+308, -1.7976931348623157e+308]]],
+  [[[1.7976931348623157e+308,-1.7976931348623157e+308]]])
 
 JSON_CHECK_POSITIVE(
   [negative reals], 
@@ -250,10 +268,14 @@ JSON_CHECK_POSITIVE(
   [negative scientific notation],
   [[[-1e3, -1E3, -2.5E2, -1e+3, -125e-3, -3.125e-2, -3125e-05, -1.525878906e-5]]],
   [[[-1000,-1000,-250,-1000,-0.125,-0.03125,-0.03125,-1.525878906e-05]]])
-JSON_CHECK_POSITIVE(
-  [1e-9999 underflows to 0],
+JSON_CHECK_POSITIVE_C(
+  [C - 1e-9999 underflows to 0],
   [[[1e-9999]]],
   [[[0]]])
+JSON_CHECK_POSITIVE_PY23(
+  [1e-9999 underflows to 0],
+  [[[1e-9999]]],
+  [[[0.0]]])
 JSON_CHECK_NEGATIVE([a number by itself is not valid JSON], [1],
                     [error: syntax error at beginning of input])
 JSON_CHECK_NEGATIVE(
-- 
1.8.3.1




More information about the dev mailing list