无法pickle <type'function'>:属性查找__builtin __.函数失败

Question

我在我的Django应用程序中收到此错误,但是,它每天只发生一次或更少,并且证明调试非常困难.

Environment:

Request Method: POST

Django Version: 1.3.1
Python Version: 2.6.6
Installed Applications:
['django.contrib.contenttypes',
 'django.contrib.sessions',
 'django.contrib.sites',
 'fimedlabs',
 'data',
 'djcelery']
Installed Middleware:
('django.middleware.common.CommonMiddleware',
 'django.contrib.sessions.middleware.SessionMiddleware',
 'fimedlabs.auth.userMiddleWare')


Traceback:
File "/usr/local/lib/python2.6/dist-packages/django/core/handlers/base.py" in get_response
  178.                 response = middleware_method(request, response)
File "/usr/local/lib/python2.6/dist-packages/django/contrib/sessions/middleware.py" in process_response
  36.                 request.session.save()
File "/usr/local/lib/python2.6/dist-packages/django/contrib/sessions/backends/db.py" in save
  57.             session_data = self.encode(self._get_session(no_load=must_create)),
File "/usr/local/lib/python2.6/dist-packages/django/contrib/sessions/backends/base.py" in encode
  93.         pickled = pickle.dumps(session_dict, pickle.HIGHEST_PROTOCOL)

Exception Type: PicklingError at /
Exception Value: Can't pickle <type 'function'>: attribute lookup __builtin__.function failed

我试过这个答案:

如何判断哪个对象属性pickle失败？

通过向实际错误添加self来查看是否会在Django错误中打印任何内容都无济于事.

在哪里可以打印出来自此错误的问题的对象,以便它出现在Django错误页面中？

谢谢!〜马特

编辑:我存储在缓存中的唯一对象是一个用户对象,代码如下:

class user(object):
username = str()
userid = uuid.UUID(int=0)

client = models.Client()
clientid = uuid.UUID(int=0)
clientname = ''

data = models.User()
accesslevel = models.AccessLevel()

active = False
client_active = False
isFimed = False
isFimedAdmin = False
isClientAdmin = False
isFimedManager = False
mysettingsform = None
viewingas = False

menu = []

_exists = False
_authenticated = False

def __str__(self):
    return str(self.__dict__.copy())

def __getstate__(self):
    return self.__dict__.copy()

def __setstate__(self, dict):
    self.__dict__ = dict

def __init__(self, username=None):
    if username:
        self.initialize(username)

def initialize(self, username):
    self.username = username
    model = models.User.objects.filter(username=username).all()
    if len(model) == 1:
        model = model[0]
        self.data = model
        self._exists = True
        self.userid = self.data.id
        self.active = self.data.active
        self.isFimed = self.data.isFimed()
        self.isFimedAdmin = self.data.isFimedAdmin()
        self.isClientAdmin = self.data.isClientAdmin()
        self.isFimedManager = self.data.isFimedManager()
        self.mysettingsform = UserFormSelf(initial={"id":model.id, "username":model.username, "name":model.name, "email":model.email, "phone":model.phone})

        self.accesslevel = models.AccessLevel.objects.filter(id=self.data.accesslevel_id)[:1][0].level
        cli = self.data.client
        self.client = cli
        self.clientid = cli.id
        self.clientname = cli.name
        if cli.active:
            self.client_active = True

        model.lastlogin = datetime.datetime.now()
        model.save()

        self.menu = getMenu(self.data)
    else:
        self._exists = False

def authenticate(self, password):
    self._authenticated = False
    if (self.active == False or self.client_active == False):
        return False
    if self._exists:
        import hashlib
        hash = hashlib.md5('%s%s' % (str(password), self.data.pwsalt)).hexdigest()
        if hash == self.data.pwhash:
            self._authenticated = True
            return True
    return False

def updateUser(self):
    self.initialize(models.User.objects.filter(id=self.userid).get().username)

def mkContext(self):
    c = Context()
    c['menu'] = self.menu
    c['user'] = self
    c['language'] = language
    c['colors'] = colors
    c["isFimed"] = self.isFimed
    c["isFimedAdmin"] = self.isFimedAdmin
    c["isClientAdmin"] = self.isClientAdmin
    c["isFimedManager"] = self.isFimedManager
    c["mysettingsform"] = self.mysettingsform
    return c

编辑:Werkzeug之后的WSGI文件:

import django.core.handlers.wsgi
djangoapplication = django.core.handlers.wsgi.WSGIHandler()
def application(environ, start_response):
    if 'SCRIPT_NAME' in environ:
        del environ['SCRIPT_NAME']
    return djangoapplication(environ, start_response)
# The following lines enable the werkzeug debugger
import django.views.debug
def null_technical_500_response(request, exc_type, exc_value, tb):
    raise exc_type, exc_value, tb
django.views.debug.technical_500_response = null_technical_500_response
from werkzeug.debug import DebuggedApplication
application = DebuggedApplication(application, evalex=True)

Answer 1

在我的情况下(不是与Django相关),multiprocessing.Pool.map当lambda作为目标函数传递时抛出了这个异常.创建一个命名函数并通过initargs参数(而不是通过闭包)传递所需的上下文数据结构解决了这个问题.

总而言之,触发异常的不良用例是:

import multiprocessing as mp
context = some_object
pool = mp.Pool()
worker_func = lambda x: work(x, context)
results = pool.map(worker_func, data_list)

Answer 2

在哪里可以打印出来自此错误的问题的对象,以便它出现在Django错误页面中？

简短的回答 - 没有重新编译cPickle,你不能.

更长的答案:这是引发异常的一段代码:

root $ grep -Hra "attribute lookup" /usr/lib64/ 2>/dev/null | grep -a failed
/usr/lib64/python2.7/lib-dynload/cPickle.so:H?H???P0H?5zM H?=1??M??H??H?ZM ?????H?=?1?H???M??H??H??M ?????H?4M H?5~H???H?????????H?M H?5tH???H?????????H??L H?5?H???qH?????d???H?M H?5NH???SH?????F???H?dM H?5bH???5H?????(???H?=X1??O??H?HC H?5I H?=U1?A??H??L ?vH??H??I???????H???RK??H?=.H????J??H?5$H??H??H?D$?G??H?D$H??tH?H??H??H??gH?DL?ttH?qH?5nH?=kH??1??XK??H?5bH??H??I????F??H?5\L??H???F??H??tH?EH??H??H?E??M???????I?$H??H??I?$?t???I?DL???P0?d???f?H?EH??H??H?E?F???H?H???P0?7???@H?|$H??P0?????H?H???P0?????H?H???P0?k?????UH??SH?H?H: H???tH?;: H???H?H???u?H?[?Ð?H??M??H??attribute deletion is not supportedunsupported pickle protocol: %dargument must have 'read' and 'readline' attributespickle protocol %d asked for; the highest available protocol is %dargument must have 'write' attributeGlobal and instance pickles are not supported.Attempt to getvalue() a non-list-based picklerUnexpected data in internal listBINSTRING pickle has negative byte countno int where int expected in memoCan't pickle %s: import of module %s failedCan't pickle %s: attribute lookup %s.%s failedCan't pickle %s: it's not the same object as %s.%sCan't pickle %s: extension code %s isn't an integerCan't pickle %s: extension code %ld is out of rangecould not convert string to intLONG pickle has negative byte countcould not convert string to floatBINUNICODE pickle has negative byte countunregistered extension code %ld_inverted_registry[%ld] isn't a 2-tuple of stringsA load persistent id instruction was encountered,

如果你看得足够近,那就有一个说法

Can't pickle %s: attribute lookup %s.%s failed

现在,如果您下载Python源代码,您可以轻松找到负责在以下函数static int save_global(Picklerobject *self, PyObject *args, PyObject *name)中引发异常的代码段./Modules/cPickle.c:

klass = PyObject_GetAttrString(mod, name_str);
if (klass == NULL) {
    cPickle_ErrFormat(PicklingError,
                      "Can't pickle %s: attribute lookup %s.%s "
                      "failed",
                      "OSS", args, module, global_name);
    goto finally;
}

因此,调试此错误的最佳方法是以不同方式格式化字符串(可能提供PyString_AS_STRING((PyStringObject *)name),重新编译和安装修改后的Python版本).

是的,我知道这太糟糕了.我自己也遇到了同样的问题.