Currently, using Cinder's backup service with RBD the
backup-create operation gets stuck when logging
('use_json=True' must be set in the config file).
The oslo.log JSONFormatter gets stuck when passing an
RBDVolumeIOWrapper from os-brick. This happens via os-brick's
utils.trace() method which passes a connector containing
{'path': RBDVolumeIOWrapper}.
The oslo.log JSONFormatter format() method calls
oslo_serialization's jsonutils.to_primitive and passes in
this RBDVolumeIOWrapper object.
Therefore the to_primitive method eventually calls
RBDVolumeIOWrapper.read(). In order to fix this the current
path avoids mapping io.IOBase objects and fallback the wrapper
RBD volume object.
Co-authored-by: Eric Harney <eharney@redhat.com>
Closes-Bug: #1908607
Change-Id: I3c416e855cb5f0dc32d14b2749ba92aba8964574
We don't need to worry about versions of Mox released in 2011 any more
https://bugs.launchpad.net/nova/+bug/852095
Change-Id: Ifd2228071c70e516f11eefde29932813b7cf6a97
Signed-off-by: Stephen Finucane <sfinucan@redhat.com>
When serializing an exception return its 'repr'. A 'repr' is a
printable representation of an object.
For example the exception: ValueError("an exception") will be returned
as the string: "ValueError('an exception',)"
Change-Id: Iac9f4624bcc4ff65e27c9ca20c6cbbe9481cf334
For example, to_primitive(fallback=repr) can be used to prevent
serialialization error like "ValueError: Circular reference detected"
when using the JSONFormatter of oslo.log.
If fallback is set, it is also used to convert itertools.count(),
"nasty" objects like types, and to handle TypeError.
Use fallback=six.text_type to convert objects to text.
This patch doesn't change the default behaviour.
Related-Bug: #1593641
Change-Id: Ie0f7f2d09355c3d2a9f7c5ee8f7e02dfea3b073b
We try to raise ValueError directly when can't convert the value into
proimitive in If9e8dd5cc2634168910d5f9f8d9302aeefa16097, but revert it
due to some reasons The better way is that notify consuming projects
with the future API behavior change before the change. It will raise
ValureError in version 3.0.
Partial-Bug: #1593641
Change-Id: I2eff07cd25f0565b380cb6e76628a896c8d0ec61
This reverts commit 3727b2d6e051b7b07680df361f38b9dbb576462c.
It breaks nova's test, need log warning before we make major release.
Change-Id: Ibbfde6207fc8c066155758c7c1aa1521b95d4db1
Currently, to_primitive tries to iterate through all addresses in the
network, because the type doesn't have a special handling that would
short curcuit it, but also has __iter__. This may be detrimental to
performance, up to the point of node crash due to memory exhaustion if
the passed network range is too large (think of 0.0.0.0/0 or even
2001::/64). This behavior also makes it impossible to restore the
original data format (CIDR).
This patch short curcuits the iteration by handling the IPNetwork type
as a special case, same as we do for IPAddress.
Change-Id: I6aecd2d057d282a655ff9e4918c164253142b188
Closes-Bug: #1698355
The problem in the current version of to_primitive function
from jsonutils module is in the situation when the function
doesn't know how to convert an object to primitive. In that
case the function simply returns the same object which causes
the following exception later in json.dumps:
ValueError: Circular reference detected. This exception is not
obvious and is quite misleading. So I think it would be better
to explicitly raise ValueError here.
Closes-Bug: #1593641
Change-Id: If9e8dd5cc2634168910d5f9f8d9302aeefa16097
1.As mentioned in [1], we should avoid using
six.iteritems to achieve iterators. We can use
dict.items instead, as it will return iterators
in PY3 as well. And dict.items/keys will more readable.
2.In py2, the performance about list should be negligible,
see the link [2].
[1] https://wiki.openstack.org/wiki/Python3
[2] http://lists.openstack.org/pipermail/openstack-dev/2015-June/066391.html
Change-Id: Ifbe7929580ec61a2cd50232794c0cbd0cbf37edc
There is a performance impact when using dict() instead of {}
in CPython (up to 6 times longer).
Considering the to_primitive function is recursive this can
have quite an effect.
Measuring:
$ python2.7 -m timeit -n 1000000 -r 5 -v 'dict()'
raw times: 0.24 0.24 0.24 0.239 0.24
1000000 loops, best of 5: 0.239 usec per loop
$ python2.7 -m timeit -n 1000000 -r 5 -v '{}'
raw times: 0.0417 0.0413 0.0407 0.0411 0.042
1000000 loops, best of 5: 0.0407 usec per loop
For more information:
https://doughellmann.com/blog/2012/11/12/
the-performance-impact-of-using-dict-instead-of-in-cpython-2-7-2/
Change-Id: Ia0b5892773a19cbabe40313a3bc788580a943f53
There are some differences in the behaviour of json and msgpack
serializers. Msgpack throws TypeError when it doesn't know
how to handle an object. Json throws ValueError instead. Regarding
the fact that serialization should be configurable (in particular,
serializers should have similar behaviour; right now transition from
json to msgpack leads to errors and fails) this patch proposes to
raise ValueError instead of TypeError on failures so that libraries
already using jsonutils would be able to also work with msgpackutils.
Change-Id: I3d21b7d136e5a426a3c4a70a953c82ddcd6ef5af
Fixes jsonutils.to_primitive so that it can serialize ipaddress
objects, as used by Python 3 or through the backport in
global-requirements.txt. Current support for netaddr is maintained.
Change-Id: I6aa1ecd5be754f61cdda4e47bc0e0180e232b366
Closes-bug: 1590076
OpenStack has ended support for Python 2.6 in all services and
libraries. We no longer need to support a fallback to simplejson on
Python 2.6 so we can remove this fallback logic.
Closes-bug: 1533337
Change-Id: I613541a37229ff2f3ce269cfa7496223d1ff37d0
Document in which version new types and functions were added using
".. versionadded:: x.y". Document changes using
".. versionchanged:: x.y."
For base64 module, add the versionadded tag in the module top
docstring, not on each type/function.
I used "git blame" + "git tag --contains=SHA1" to find these version,
and then I checked manually each version.
Change-Id: I4a891b18064fe7b857a79c57030ff31f7a0370b4
The jsonutils.dumps() function returns bytes on Python 2 and Unicode
oon Python 3. In some cases, we always want bytes. For example, a
HTTP body must be bytes. This function avoids an condition call to
.encode() depending on the type or on the Python version.
For example:
body = jsonutils.dumps(data)
if isinstance(body, six.text_type):
body = body.encode('utf-8')
can be replaced with:
body = jsonutils.dump_as_bytes(data)
Change-Id: Ib9d8f1309982762b54d8a91b1f24f64d0ae6723a
This is a wire format change that the projects are not ready for yet (example Nova).
We should do this as part of a major version bump (2.x) with ample time for folks to move up.
Nova review to prepare the Nova code base for this change:
Ib7e60ed94edd1eb409baf8c29c61237d85944787
This reverts commit 5b0827a3df584b3e26bb44e5358cc28ac4a91636.
Change-Id: Ic4bf047b40c1dad2f8d2bc0afda934e47dba1f9d
Currently, only the dict value are converted with to_primitive().
There's no reason not to do it on keys too to support special types.
Change-Id: I5b447f2fcd814088483f67e54261131cfc87ae79
There's a few code duplications because for some reason the code
handling the recursive part got some conversion code for non-recursive
value.
Let's stack non-recursive value conversion as the top, and then use
"recursive" only where needed.
Change-Id: Ic1ecc76aba5402129a936dfb4649df0806a564a4
Following discussion in change I8b5119e64369ccac3423dccc04421f99912df733
in oslo.utils, we plan to deprecated and remove strtime() which is to be
considered dangerous.
This changes the json serialization code to embed the timezone in the
encoded datetime string if it's available.
Change-Id: I262391987d25b483d2bec4ac3abe94b6e9d032ad
Ensure jsonutils and msgpackutils headers are linkable and
connect into there associated functions and classes so that
clicking on links works (instead of not working).
Change-Id: Ifa28421e7d21a85290ee13d04918325c3f991b92
There's a different behaviour currently between dump() and dumps() as by
default dumps() use the to_primitive function provided by this module,
and dump() does not. This patch fixes that.
Change-Id: I8a116e3777901bc713e9812c1a654a51b43e702a
Move the public API out of oslo.serialization to
oslo_serialization. Retain the ability to import from the old namespace
package for backwards compatibility for this release cycle.
bp/drop-namespace-packages
Change-Id: Ic60f809ea00ac77b0753556c6fd00b97e64d57ff
